2025-01-14 11:10:09 RT @tomaarsen: I've just published SetFit v1.1.1, allowing finetuning extremely efficient classification models on CPU &amp

2025-01-14 11:09:45 @LouisAnslow ...except she did an absolute bucketload of serious high quality research that forms the foundation of her worldview. I wouldn't be so quick to dismiss her -- but instead read her work closely and see how it can help.

2025-01-14 11:03:25 RT @AdamThierer: Check out this newly proposed constitutional “Right to Compute” amendment just floated in New Hampshire! “The right of…

2025-01-14 11:00:31 If you'd done the https://t.co/GEOZunWoXj course in 2016 you'd have studied this and been 9 years ahead of your time. :) (Or just do the latest course now, since there's still plenty of stuff in there that most people don't know about.) https://t.co/L421VvI6Hv

2025-01-14 09:46:20 RT @iScienceLuvr: Tensor Product Attention Is All You Need Proposes Tensor Product Attention (TPA), a mechanism that factorizes Q, K, and…

2025-01-08 07:12:42 RT @andrew_n_carr: Another great interview question! For linear regression, we can directly compute the minimizer as β= = (X^TX)^{-1}X^Ty.…

2025-01-08 07:12:35 @andrew_n_carr Oh I love this one!

2025-01-08 01:09:06 RT @TRyanGregory: Who could possibly have guessed that the "kids need to get infected with pathogens" folks would be wrong and that infecti…

2025-01-07 21:34:56 RT @scaling01: Nvidia marketing: "5070 has 4090 performance" vs reality: https://t.co/BafyZgfls2

2025-01-07 21:34:37 RT @TheXeophon: >

2025-01-05 01:12:37 RT @gui_penedo: We've just updated FineWeb and FineWeb-Edu with data from all the remaining 2024 CommonCrawl dumps, covering up to Decem…

2025-01-04 04:54:23 RT @htmx_org: drop shadows and inset inputs will RISE AGAIN (prefer classic UI to skeuomorphism, but i'll take anything a this point over…

2025-01-04 02:15:21 RT @giffmana: I agree with the whole thread, it’s very well put. But let’s not forget the whole current focused bet mode across all big la…

2025-01-03 19:14:41 RT @molly0xFFF: The world's richest man has joined a growing chorus of right-wing voices attacking Wikipedia as part of an intensifying cam…

2025-01-03 18:44:40 RT @HamelHusain: I don’t know what people expect honestly. Come up with a dumb standardized test w a $500k bounty and yes people are gonna…

2024-12-30 22:35:35 SW eng manager: No real work gets done in Jupyter notebooks. Alex Radford: I invented GPT and CLIP in Jupyter notebooks. https://t.co/UNPhKjCwJV

2024-12-30 21:41:07 @HanchungLee @m_chirculescu Actually I pass in a (iterator over) a list of tuples. Left on the left and right on the right is, ummm, not really that confusing?

2024-12-30 21:39:19 @HanchungLee It's definitely not a standard way to write it in many communities I've been involved in, and therefore definitely not easier to maintain for anyone in those groups. It looks to me like a Python port of Java, C or C++. Which makes sense, since many courses teach with those.

2024-12-30 21:26:12 @F_Vaggi I'm not suggesting people have to write code how I write code (although I *do* think it works pretty well!), just that folks shouldn't give up learning to code just yet, since LLMs are not currently creating high quality code, and often can't solve problems at all.

2024-12-26 23:51:59 RT @jiayq: There is a lot of unconscious emphasis of the DeepSeek model being “Chinese” and implicit connection with the Sino-US relationsh…

2024-12-26 23:38:02 RT @mitchellh: https://t.co/39Xj39wheA https://t.co/PH0qejFB4z

2024-12-26 23:35:03 RT @jiayq: In 2019 I had a chat with the DeepSeek team, in the hope of selling them an AI cloud solution. I was trying to convince them a f…

2024-12-26 23:32:46 RT @HamelHusain: Am I the only one who finds much of the "skilled immigration" discourse disingenuous, esp from those who allocate capital?…

2024-12-26 22:40:43 RT @ericwastl: It's been over 24 hours since #AdventOfCode 2024 ended! This AoC season: - 273,313 users collected at least one star - 3,65…

2024-12-24 18:52:04 RT @huybery: We've explored inference-time scaling for visual multimodal tasks and introduced QVQ, the first open multimodal o1-like mode…

2024-12-24 18:32:33 @danofer @huggingface We didn't try RTD, although we certainly considered it. It would be a cool thing to try adding if anyone is interested.

2024-12-24 18:28:25 RT @knowledgator: Thanks to #ModernBERT introduced by @LightOnIO and @answerdotai we can refresh #GLiNER making it more accurate, efficien…

2024-12-24 03:58:58 RT @scumitchell: I benchmarked over 100 HTML tags so you don't have to and here are the visualized results. Not all HTML tags are created…

2024-12-24 03:58:28 RT @GrantSlatton: friend is a medical illustrator few years ago, we were wondering when AI could do her job — getting the details right is…

2024-12-23 01:21:45 Makes me so happy to see the stuff Rob and I built at FastMail 25 years ago provided such a great foundation for the service today -- fantastic article here explaining how FastMail manages its own hardware and avoids the cloud: https://t.co/2uVNzUiVqK

2024-12-23 00:18:35 RT @giffmana: The period have spoken: o3 is more human than François at this point https://t.co/c9lGBbM9VQ

2024-12-23 00:17:38 RT @hugobowne: Hey – for my new followers (welcome!), I co-lead a course with @stefankrawczyk on building LLM-powered apps for data scien…

2024-12-23 00:16:50 RT @devgerred: Are you ready to hang out with Claude in @Minecraft? Announcing the Model Context Protocol Minecraft Serever (MCPMC!). Link…

2024-12-22 19:39:02 Wow this is a fascinating idea - using the reversible transform pipelines I developed for fastcore, as a basis for for analysing quantum computing. It's a very accessible and well written article, so check it out! (Also a nice h/t there to @michael_nielsen's book.) https://t.co/N5Pp8Bm6Tr

2024-12-21 01:31:15 RT @Sauers_: The total compute cost was around $1,600,250, more than the entire prize https://t.co/utsvasTfye

2024-12-21 01:30:24 RT @random_walker: Google's Deep Research is an excellent application of agentic capabilities. One example something it can do pretty well…

2024-12-20 21:59:44 @gautamethiraj I think the idea is you're meant to write that yourself...

2024-12-20 20:53:27 @cognitivecompai @stefkrawczyk @HamelHusain I don't get it either - what's not clear in our blog post? I thought we're fairly clearly stating that this is better than Llama and friends. https://t.co/aAPNMplfRN

2024-12-20 18:44:10 RT @tom_doerr: Okay so can I just keep sending 1 million+ character prompts to Gemini Flash 2.0 and not pay for it? How is one of the best…

2024-12-19 01:45:29 RT @nabeelqu: I'm really unhappy with the framing of this study. They gave Claude the choice of: 1) Keep to your training (helpful, hones…

2024-12-19 01:42:15 @MarkWitucke Seems to be a pareto efficiency chart of some kind...

2024-12-18 23:50:37 Tired: Efficiency results on 1024 H100s. Wired: Efficiency results on a 4090. https://t.co/WdcmC16yT4

2024-12-18 23:47:45 RT @fchollet: People have completely rewritten what "scaling laws" were supposed to mean. Originally it was "pretraining a larger LLM on mo…

2024-12-18 23:38:46 something's coming... https://t.co/QmtZHuk6Qd

2024-12-17 00:20:36 RT @random_walker: My wife noticed an article with my name on it in her feed and said “Did you write this? It doesn’t sound like you.” I wa…

2024-12-17 00:19:13 RT @PessimistsArc: Yeah... anti-polio vaccine poster from 1955 https://t.co/aOceiJjKDA

2024-12-17 00:13:13 RT @vishal_learner: Just published the 2nd video in my fastbook-benchmark series, Document Processing, where I take fastbook Jupyter Notebo…

2024-12-17 00:13:03 RT @isaac_flath: There are lots of HTMX examples on the FastHTML Gallery thanks to fantastic community contributors Check them out @ ht…

2024-12-16 10:17:18 @cognitivecompai No I never said it's owed. I think it's a mis-step and isn't in line with the company's values as I understand them. Perhaps you're right they're more like Burger King or Pepsi, rather than their initial high aspirations.

2024-12-14 16:55:01 @vokaysh @giffmana There's an argument that being a fucking racist is fucking shit and fucking racists should fuck off. It's not about over-sanitization of fucking language, cunt.

2024-12-14 16:16:14 RT @XiongChenyan: Not even calling Racism Racism is Racism

2024-12-14 16:16:00 RT @fangchenliu_: Is it hard to directly use the words like racism or discrimination?

2024-12-14 10:59:12 @admercs @NeurIPSConf Hope you’re not eager to be outraged if I call you a fucking racist btw.

2024-12-14 10:57:57 @admercs @NeurIPSConf wtaf. Seems some people are eager to be edgelords but don’t realize they’re just being fucking assholes and should stfu.

2024-12-13 06:42:41 RT @__tinygrad__: We sell 9 green boxes for every red box, even though red boxes are 60% of the cost. The theoretical performance is simila…

2024-12-13 06:42:12 @samcharrington @HamelHusain Just using a big monolithic server without kubernetes is probably actually the right answer for most people though... And a build script... And log files... And we're not actually using distributed version control -- everyone uses git as if it were svn.

2024-12-13 06:38:49 RT @HamelHusain: I can't be the only one that sees this correspondence between the past and the present. https://t.co/Ep7mGlEfk5

2024-12-13 06:38:22 @peteratmsr @MSFTResearch Trust me - it's so demoralising to constantly see everyone else building transformative stuff that everyone uses to do amazing things, knowing that your work could be there too, if only management would allow it!

2024-12-13 06:36:48 @peteratmsr @MSFTResearch And it'll be way easier to keep your top people if they can actually see the stuff they're building being *used* -- either as a commercial MSFT product, or as a freely available tool for other companies to build on.

2024-12-11 04:27:45 RT @HamelHusain: o1-pro doesn’t accept pdfs or word documents (unlike 4o) For $200, I expect more from basic UX? Like don’t make me conv…

2024-12-10 20:40:45 Possibly some folks at OpenAI don't really know what tasks most humans do. https://t.co/vGTO9uI0Ve https://t.co/BLQiqubdud

2024-12-10 18:54:55 RT @thehonestlypod: Marc Andreessen (@pmarca) says he attended “absolutely horrifying” meetings where Biden’s government vowed to take “com…

2024-12-10 18:51:11 RT @iScienceLuvr: The inventors of flow matching have released a comprehensive guide going over the math &

2024-12-10 18:48:31 @julianharris @ncooper57 Also not the same thing AFAICT -- that appears to just be a Copilot CLI clone.

2024-12-09 00:40:42 RT @karpathy: @johnschulman2 @gordic_aleksa @amasad @DavidSacks The pitch is that reasoning capabilities learned in reward-rich settings tr…

2024-12-09 00:34:11 RT @isaac_flath: New FastHTML Gallery entry thanks to Adnan! Add custom keybindings with HTMX https://t.co/A7LgWLoAAZ

2024-12-09 00:33:24 @gordic_aleksa Way less good.

2024-12-09 00:32:59 @tdietterich @sebkrier Sure that’s definitely a useful research direction and has many great applications!

2024-12-08 19:39:56 @tdietterich @sebkrier Being cheaper is often very useful. In fact some things are not even possible to create in practice if they’re too expensive or complicated. Aiming for perfection in accuracy is often not optimal.

2024-12-06 19:43:44 RT @strickvl: @jeremyphoward @ncooper57 For anyone looking to learn about tmux, the book that really made it click for me was @bphogan's on…

2024-12-06 19:42:30 RT @JeffDean: Today’s the one year anniversary of our first Gemini model releases! And it’s never looked better. Check out our newest rel…

2024-12-06 09:05:00 @pastaraspberry @giffmana @ncooper57 It used to be called bash-buddy, but the name on pypi was taken :( It's aliased to "b" in my bashrc.

2024-12-06 04:19:52 RT @ChristophMolnar: We even have evidence backing what Mark says: No substantial overfitting in Kaggle competitions (limitation: paper i…

2024-12-06 04:19:48 RT @marktenenholtz: “All Kaggle solutions are overfit” says the person who has never read a solution overview where the author discusses th…

2024-12-03 19:50:34 RT @aallan: Documentation isn't my day job any more, but the proposal for an "llm.txt" file by @jeremyphoward sort of feels like one of tho…

2024-12-03 10:26:08 RT @HamelHusain: This claudette doc is a nice demonstration of a simple "agent" (tool calling loop, optional human intervention) It's gre…

2024-12-03 01:19:25 RT @paul_cal: Product lead &

2024-12-03 01:18:54 @htmx_org You could add a java applet which shows flames around your logo, but make it neon?

2024-12-03 01:17:34 RT @Teknium1: 3 months after the preliminary report on DisTrO we've got a massive 15B run happening completely distributed across the inter…

2024-12-01 06:40:47 RT @petergyang: Honest reflections from coding with AI so far as a non-engineer: It can get you 70% of the way there, but that last 30% is…

2024-12-01 06:29:08 @ericwastl We have 1,000 students working through them right now in our "solveit" course, and having a great time! :D We're using them to teach and practice foundational problem solving principles and coding techniques.

2024-12-01 05:49:28 RT @htmx_org: however much you hate flat design, it isn't enough https://t.co/0yv8L0gV7G

2024-12-01 03:36:14 RT @ericwastl: This was my tenth(!) year building 25 days of puzzles for #AdventOfCode. You can solve them all for free and in whatever pro…

2024-12-01 03:27:39 @shawkisukkar @amasad incorrect. the phrase you're looking for is "RIGHT *NOW* MOTHERFUCKER". follow me for more etiquette tips.

2024-11-30 00:56:28 RT @karpathy: People have too inflated sense of what it means to "ask an AI" about something. The AI are language models trained basically…

2024-11-30 00:53:09 RT @cloneofsimo: Eval doesnt look bad, so this actually worked (If this is as good as they say) This basically means whatever neo gpu clou…

2024-11-30 00:51:31 @birdmademejoin But one we're effectively bans way more things by not actually showing people stuff with links, youtube, mentions of other social media sites, etc. It's not an obvious win, to me at least. (And this one bans entire words.) They both seem pretty shitty, just in different ways.

2024-11-30 00:47:14 RT @random_walker: If I had a dollar for every time someone rediscovers or restates the thesis of the essay "AI safety is not a model prope…

2024-11-30 00:46:51 RT @vishal_learner: I eval'd full text search, single-vector cos sim, ColBERTv2, and answerai-colbert-small-v1 on my fastbook-benchmark for…

2024-11-27 23:59:39 Maybe Apple Intelligence shouldn’t mark scam emails as “Priority” with a summary saying it’s for security purposes? https://t.co/CcMQnlJ7MX

2024-11-27 23:57:11 RT @chrisalbon: On the other app, yesterday Hugging Face released a dataset of one million posts. The anti-ai reaction was so negative that…

2024-11-27 19:57:22 RT @reach_vb: WTF! Qwen COOKED - QwQ 32B beats o1 mini and competes with preview! https://t.co/SzfcfoHIKu

2024-11-27 19:55:55 RT @random_walker: New short paper on the limits of one type of inference scaling, by @benediktstroebl, @sayashk and me. The first page c…

2024-11-27 19:55:06 RT @htmx_org: daily reminder that The Web was built on mixing concerns https://t.co/tx47EExoCj

2024-11-25 23:58:13 @ph_singer @wightmanr @carlolepelaars @jxnlco @simonw @fishnets88 @chrisalbon @charles_irl @DavidSHolz @bhutanisanyam1 @StuartReid1929 @SerranoAcademy @rohanpaul_ai @isaac_flath @willmcgugan No reason that >

2024-11-25 23:55:19 RT @martinbowling: Excited to announce my llms.txt Generator - bridging the gap between websites and AI! Built with @hyperbolic_labs and…

2024-11-25 06:05:00 RT @kellerjordan0: This is officially the new record! Congrats @hi_tysam (who is also an OG of CIFAR-10 speedrunning) https://t.co/95LfOB66…

2024-11-24 21:18:28 RT @swyx: @pastaraspberry @jeremyphoward @wightmanr @rahuldave @chrisalbon @carlolepelaars @jxnlco @simonw @fishnets88 @charles_irl @DavidS…

2024-11-24 18:12:44 @wightmanr @rahuldave @chrisalbon @carlolepelaars @jxnlco @simonw @fishnets88 @charles_irl @DavidSHolz @bhutanisanyam1 @StuartReid1929 @SerranoAcademy @rohanpaul_ai @isaac_flath @willmcgugan @swyx has an excellent daily summary that includes some key discords.

2024-11-23 00:33:29 RT @francoisfleuret: There is definitely something happening with bluesky. Got 1.5k followers in a few days and the whole AI crowd is rushi…

2024-11-22 23:44:44 RT @WolframRvnwlf: Can confirm! I've continued benchmarking the latest models (MMLU-Pro, computer science category) - look at the OpenAI mo…

2024-11-22 23:32:48 @mitchellh @willmcgugan @Apple Oh I got an invite - thanks! :D

2024-11-22 11:33:41 @nearcyan @GrantSlatton Oh it really is. Much worse.

2024-11-22 11:30:42 @willmcgugan @Apple I’ve been on the ghostty discord hoping for an invite, but no dice…

2024-11-21 02:10:03 RT @DeryaTR_: I just tested @deepseek_ai’s deep think model with highly sophisticated, complex research problems &

2024-11-21 02:09:44 RT @YouJiacheng: NOTE: longer sequences have lower average loss, so it's not a really apple-to-apple comparison.

2024-11-21 02:08:35 RT @sedielem: In the end, my efforts on @kaggle played a big role in getting my foot in the door at @GoogleDeepMind, where I still work tod…

2024-11-21 02:07:59 RT @sedielem: Story time! Some of the most productive time I spent during my PhD was, ironically, when @kaggle competitions distracted me f…

2024-11-21 00:58:17 RT @Tim_Dettmers: Just to clarify this benchmark. This is an apple to oranges comparison. - Cerebras is fast for batch size 1 but slow for…

2024-11-19 01:47:45 Oh wow there's another social network where people are actually active?!? https://t.co/tZtwAOJsiE

2024-11-19 01:35:23 @aidangomez Drake who?

2024-11-19 01:34:41 RT @charliermarsh: It looks like @Railway added first-class uv support to Nixpacks https://t.co/puH0tzDVBI

2024-11-18 21:51:42 RT @jonLorraine9: Getting harsh conference reviews from LLM-powered reviewers? Consider hiding some extra guidance for the LLM in your pap…

2024-11-18 21:51:29 @PandaAshwinee @m2saxon So why say "DO NOT DO THIS" in general, rather than letting people know what the specific conference is that this very particular requirement is for?

2024-11-16 22:32:25 RT @HamelHusain: VCs and Execs are quote tweeting this about the horrors of taxing unrealized capital gains In most situations in US we ar…

2024-11-16 21:42:56 RT @giffmana: Hahahaha! I believe this refers to RND (Random Net Distill). I too was looking at Montezuma a lot back then. Fun times, but i…

2024-11-16 20:57:55 RT @Arindam1408: HF: https://t.co/ScIfxCTEVZ Paper: https://t.co/RdxuinmCeS https://t.co/JulVnnbTmG

2024-11-16 19:51:21 @YouJiacheng I dunno what will happen to them. They upgraded to the current latest version. Depends if there's a transition plan that works well.

2024-11-16 19:48:43 RT @emollick: The “personality” and “opinions” of AI are not stable, they are influenced by prompting &

2024-11-15 02:17:46 RT @doomie: Who leaked this to The Information?

2024-11-15 02:13:18 RT @aaron_defazio: 1 out of 10 score review I see browsing ICLR. Not my paper, but . https://t.co/CLOFG9yNHI

2024-11-15 02:07:37 RT @OfirPress: Super cool work by @nielstron et al: SWT-Bench is SWE-bench for test generation! They give the model a repo and an issue an…

2024-11-15 02:04:59 @HamelHusain Nah this is good -- people who are enthusiastic to spam the buttons deserve more votes! (And people enthusiastic enough to write a script to spam them... even better :D )

2024-11-15 01:25:41 RT @danaparish: I am thrilled to bring you this fascinating, shocking interview with former CDC Director, Dr. Robert Redfield. I almos…

2024-11-13 00:54:45 Great to see! BTW their table is wrong in the post -- corrected one here: https://t.co/ML8YQWXuEk https://t.co/poWVwgwyvd

2024-11-12 21:10:18 RT @danielhanchen: Bug fixes &

2024-11-12 19:37:09 RT @FFmpeg: If you say so...

2024-11-12 14:53:29 RT @Suhail: If you find this stuff is writing tons of your code, perhaps it's time to challenge yourself more. Obviously not always the cas…

2024-11-12 14:53:20 RT @Suhail: LLMs are still midwit coders as far as I can tell today

2024-11-11 01:28:59 RT @fchollet: ARC Prize 2024 concludes. Two teams made to mid-50s. Congrats to all participants! We'll be back next year.

2024-11-11 01:28:11 RT @clashreport: Chinese landing party: A pack of robot dogs (at least 2 armed) in action. https://t.co/O4aN2mODwh

2024-11-10 22:40:53 RT @molly0xFFF: For anyone looking to adjust their media diet, now’s a great time to consider escaping The Algorithms with RSS. Here are so…

2024-11-10 19:48:06 RT @karpathy: Moravec's paradox in LLM evals I was reacting to this new benchmark of frontier math where LLMs only solve 2%. It was introd…

2024-11-10 19:42:54 RT @HamelHusain: This the kind of stuff that AI is really amazing for: building tools for myself that I wouldn’t have time for otherwise

2024-11-09 00:50:02 @lilianweng Welcome back to the outside world Lilian! :D

2024-11-09 00:49:35 RT @chrisalbon: ITS HAPPENING https://t.co/td3bMtkk6I

2024-11-08 22:46:08 Current LLMs max out at 2% accuracy in trying to solve really hard math problems. https://t.co/sPDB4kYQFL

2024-11-08 22:43:03 @MatthewStibbard Sorry to hear that! Please email support@answer.ai with the details of the email address you used to sign up.

2024-11-08 22:42:35 @chinmayn Oh sorry to hear that! Please email support@answer.ai with the details of the email address you used to sign up.

2024-11-07 02:31:46 RT @kellerjordan0: It's a new day, and here's a new NanoGPT speedrun record: 3.28 FineWeb val loss in 8.2 minutes on 8xH100 Previous recor…

2024-11-07 00:49:56 RT @StasBekman: If you're evaluating which accelerator to choose for your future workloads beware that AMD's peer-to-peer bandwidth is 7x s…

2024-11-07 00:45:58 @sytelus Yes I know. Leslie also wrote about trapezoidal and also first introduced cosine.

2024-11-06 09:05:01 @giffmana I'm not fully up to date with them, but maybe this?: https://t.co/VmgC12EXMr

2024-11-06 09:01:12 @sytelus It was originally developed by Leslie Smith. Around 2016 IIRC.

2024-11-04 01:46:29 @osoleve How about creating a new track for running on a single 24GB card, such as a 4090?

2024-11-03 23:27:19 RT @DanielleFong: “atypical schizophrenia in the soviet era got treated like a bizarre mix of psychiatry and political control. if someone…

2024-11-03 23:23:51 RT @willccbb: already exists https://t.co/X01rX7EKH5

2024-11-03 23:22:59 RT @1goodtern: Let's clean indoor air. Let's make indoor air like running water.

2024-11-03 22:23:29 @kellerjordan0 @Grad62304977 @kellerjordan0 I think you need to update your bio…

2024-11-01 20:21:07 @alexisgallagher https://t.co/0tqrXudJOz

2024-11-01 20:14:32 @francoisfleuret @Meta @syhw @unige_en @EPFL Yay that's so cooool! :D

2024-11-01 20:14:20 RT @simonw: I updated my llm-claude-3 plugin to support attaching PDF files to prompts using the LLM command-line tool: llm install llm-cl…

2024-11-01 09:32:02 RT @xenovacom: OmniParser, the new screen parsing tool from Microsoft (and #1 trending model on @huggingface), can now run 100% locally in…

2024-11-01 00:56:28 Should I do it?... https://t.co/teWjjGm1cA

2024-10-31 01:16:33 @alexalbert__ This is super cool Alex - love the simplicity :D

2024-10-31 01:11:00 @SebastienBubeck @OpenAI @sama Have fun at OpenAI Sebastien -- I bet you'll do (even more!) great things :D

2024-10-30 21:21:47 RT @HamelHusain: The most common mistakes I see teams make with LLM judges: • Too many metrics • Complex scoring systems • Ignoring domain…

2024-10-30 21:20:09 @cleavey1985 @modal_labs Looking good!

2024-10-30 21:20:02 RT @cleavey1985: Made my very first fastHTML app along with using @modal_labs as the backend. It uses the ColPali model to process PDFs an…

2024-10-29 02:05:04 RT @MeganRisdal: Our team is discussing a bunch of new features for Kaggle's CLI (and possibly kagglehub our Python client library, too) co…

2024-10-29 01:46:02 RT @mark_k: Is it just me, or has ChatGPT made a jump in quality? The latest ChatGPT (available as "chatgpt-4o-latest" in API) is noticeab…

2024-10-29 01:43:57 @chance_on_brink @__tinygrad__ @jxmnop Also if you have the 2 parts of an if statement on the same line, it's easier (for me) to see the full logic of the idea at once. When you follow the approach that one line of code is one complete task, it can make code faster to read and understand (with practice).

2024-10-29 01:42:08 @chance_on_brink @__tinygrad__ @jxmnop Most screens are wider than tall. Using a lot of vertical space in that case means you see less context at a time, meaning you have to rely more on your working memory. Personally I find that hard work.

2024-10-29 01:40:29 @msanparr @math_rachel The vast majority of local parents I interact with here in regional Australia are extremely biased against doing anything on screens. Back when we were in SF the experience was similar.

2024-10-26 19:49:23 These are not at all criticisms of the code or even of the process of creating it. But for those folks who claim that you don't need human coders any more with AI, it's worth reflecting on why top AI labs like @AnthropicAI are using plenty of human coders.

2024-10-26 19:49:22 The error messages in the computer use tools have grammar mistakes AI almost certainly wouldn't make -- here confusing the incorrect "it's" with the correct "its" https://t.co/ddlqJlPNbX

2024-10-26 19:49:20 You can see that the internal usage of AI at big AI labs is often pretty limited when creating their own tools. E.g. within minutes of looking at the new @AnthropicAI Claude computer use code I saw 3 obvious issues that an AI would be very unlikely to create.

2024-10-24 19:52:44 RT @SafetyChanges: Company: Anthropic Date: October 22, 2024 Change: Removed a previous policy promising (sans a few exceptions) not to use…

2024-10-24 19:50:05 RT @xuanalogue: My most nerve-wracking moment of grad school was July 6 2020, when the Trump admin threatened to deport most int'l students…

2024-10-24 19:34:48 RT @NielsRogge: Microsoft silently dropped a new model on the hub "OmniParser is a general screen parsing tool, which interprets/convert…

2024-10-24 19:34:44 RT @paul_cal: @zywu_hku @DrJimFan UFO from Microsoft 8 Feb Operates Windows apps and OS from natural language instruction. Combines GPT4V…

2024-10-24 18:28:27 RT @BensenHsu: @EricTopol The study looked at the long-term effects of COVID-19 infection in a group of young, healthy U.S. Marines. They w…

2024-10-22 20:11:36 @PeakGrizzly he wrote a whole book about it

2024-10-22 20:10:11 RT @fabianstelzer: Flux 1.1 does pretty decent political cartoons, so I wired up a glif workflow that let's Gemini Pro 1.5 write the cartoo…

2024-10-22 20:07:10 @vishal_learner @RT Sweeeet!

2024-10-22 20:00:05 RT @BorisMPower: Congrats on the "low key research preview" of computer use!

2024-10-22 19:55:08 (not that @Teknium1 is just any anime pfp account mind you)

2024-10-20 23:46:27 @htmx_org IIUC there are 87 more steps now?

2024-10-20 23:45:49 RT @LouisAnslow: Breast Cancer Deaths: 42,211 (US, 2022) Car Accident Deaths: 42,795 (US, 2022) https://t.co/MBrrmDhcz2

2024-10-20 22:35:53 @psychosort Whilst I'm not a decel, North Korea isn't either. A better caption would be "Never underestimate how much fascists will fuck you up."

2024-10-20 07:58:09 RT @gallabytes: openai is going to keep building cool stuff shackled to the weak sauce that is 4o. what I'd give to have o1 on top of sonn…

2024-10-20 07:57:22 RT @davidcrespo: I worked on https://t.co/zTKG9POy4L for two years and the way it actually happened was that the software side was tyrannic…

2024-10-18 23:51:27 RT @jsrailton: BREAKING: Musk-backed PAC is micro-targeting muslim areas with ads saying Harris stands with Israel... and targeting jewish…

2024-10-18 23:39:41 RT @Birchlabs: Meta have shared an LLM pretraining codebase https://t.co/2mxq3IDHd1

2024-10-18 23:36:45 @cognitivecompai @BramVanroy I used AI to write nearly all of it :)

2024-10-18 23:18:22 @harper Thanks Harper! :D

2024-10-18 23:17:47 @cognitivecompai @BramVanroy I don't have AMD hardware - but creating this lib took 2 hours end-to-end, so perhaps someone with AMD hardware could do that?

2024-10-17 00:42:01 RT @FarouqAldori: When are we going to admit that Chatbot Arena is bullshit and maybe just shut it down for good? https://t.co/O4HQeSff10

2024-10-17 00:37:54 RT @kindgracekind: This will legitimately make this website less safe for a lot of people. I post zero personal information and I still get…

2024-10-16 23:33:35 RT @simonw: This is such a great talk about SQLite as a production-ready backend for web applications, and one of the best examples I've se…

2024-10-16 23:28:35 @chheplo ffs what's wrong with you

2024-10-16 22:58:10 RT @untitled01ipynb: sometimes you need a package to exist to convince stakeholders that you're not mad and everyone are doing it https://t…

2024-10-15 00:58:21 RT @JimStacker: @cpojer Don’t laugh: Microsoft Edge

2024-10-15 00:55:45 RT @cpojer: After more than 15 years, it's time to switch browsers. I just want Chrome but without nefarious behavior by Google. What Chro…

2024-10-15 00:18:57 RT @austinvhuang: Now that everyone and their pet gerbil is exploring the kv cache as transformer state, perhaps the next stage is thinking…

2024-10-14 20:43:30 @mzhangio Very cool! (And love the cute pics :D) Can you say more about the MMLU perf being well behind other tasks? Do you think that's a limitation of the arch? How about code and math?

2024-10-14 20:35:49 RT @m_chirculescu: Launching backtrack_sampler - for experimenting with custom LLM sampling strategies that can rollback and regenerate t…

2024-10-12 23:57:25 RT @SteveStuWill: Psychologists have posited hundreds of cognitive biases over the years. A fascinating new paper argues that they all boil…

2024-10-12 23:55:40 @vishal_learner nice job! :D

2024-10-12 23:54:49 @rahuldave @lateinteraction (I'd be more than delighted if folks made it work with mkdocs and sphinx btw)

2024-10-12 23:54:18 @rahuldave @lateinteraction But fastcore is such as teeny tiny dep already! :D

2024-10-12 19:52:35 @lateinteraction @rahuldave I'm so sad that the world didn't realise docments is the right way. All our stuff (documentation, tool use APIs, CLIs, etc) entirely uses docments.

2024-10-11 00:33:12 @EKeric13 @HamelHusain Should apply to any ASGI server.

2024-10-10 23:58:49 Great deep dive into options for running background tasks from a web app - very hard to find this kind of info clearly explained anywhere, so marvelous to have @HamelHusain show how to do it. https://t.co/nBeyncaksh

2024-10-10 23:20:31 RT @AlecStapp: Occasional reminder that one of China's biggest fears is that someday the United States might actually fix its broken immigr…

2024-10-10 23:17:43 @lateinteraction Yes. (It's also the *only* python 3.13 pun I've seen so far.)

2024-10-10 22:16:30 RT @NickADobos: OpenAI spends 1.5x more on training than vs serving inference to run the models?!? What

2024-10-08 18:57:43 RT @SuryaGanguli: Congrats to John Hopfield and @geoffreyhinton! well deserved recognition that some important foundations of AI rest on ph…

2024-10-08 18:12:10 RT @prateeky2806: Ever wondered if model merging works at scale? Maybe the benefits wear off for bigger models? Maybe you considered using…

2024-10-08 08:57:55 RT @Ethan_smith_20: fun fact: only 3096 of the possible 65536 expressible values of bfloat16 when given to exp do not round to either 0, 1…

2024-10-08 08:54:33 RT @rose_e_wang: AI has the potential to transform real-world domains. But can AI actually improve outcomes in live interactions? We condu…

2024-10-08 08:52:08 RT @emollick: A very small number of people ruin it for everyone else: the problem with unmoderated social media sites in a nutshell. https…

2024-10-01 21:56:34 @NeelNanda5 Having said that, a lot of folks clearly just think the non-profit was kinda dumb, and that instead the focus should have been on supporting the commercial org. But even if that were true, non-profit boards don't have the power to decide to ignore their mission.

2024-10-01 21:55:23 @NeelNanda5 Based on the non-profit mission that Helen was required to uphold, it seems clear that she was right -- and perhaps if it was executed a lot better, folks would feel differently about it today.

2024-10-01 21:51:43 RT @binarybits: I'm not sure if the old OpenAI board was right to fire Sam Altman. But given subsequent events it's certainly not obvious t…

2024-10-01 20:35:17 RT @keithwhor: hey everyone! super excited play a small role in the @OpenAI Realtime API launch today. we have launched an open source rep…

2024-10-01 20:25:26 RT @deanwball: My basic reaction to AI today is, “jeez, o1 performs in the top 1% of humans at math, yet fails routinely at basic logic tas…

2024-09-29 23:46:59 RT @mikeknoop: Glad to see SB 1047 officially veto'd by @GavinNewsom. We must keep incentivizing AI innovation, not putting speed bumps on…

2024-09-29 23:46:39 RT @chrislengerich: SB-1047 is vetoed, a powerful victory for science and freedom. @GavinNewsom showed unprecedented wisdom here in select…

2024-09-29 23:45:55 RT @vipulved: It's great to see @GavinNewsom's veto of SB 1047. Open-source AI is the most viable way forward for sustainable and responsib…

2024-09-29 23:43:37 @AndrewYNg Likewise, Andrew :)

2024-09-29 21:03:32 @dome_271 I'm so looking forward to this :D

2024-09-28 06:21:23 @michhuan @ylecun @lessig My wife is doing a masters in immunology. The students in the course learn a *lot* about viruses. The information is provided in books in forms that the authors endeavor to make the information as understandable as possible, and teachers help transmit that knowledge to students.

2024-09-28 06:19:24 @michhuan @francoisfleuret These highly targeted restrictions of specific information based on actual experiences of real safety issues seem like a great idea.

2024-09-28 06:17:21 RT @francoisfleuret: If you can replace the AI thingy with "books" in the sentence, I am not impressed.

2024-09-28 06:16:25 @michhuan @ylecun @lessig We do not ban these things (although historically many have tried, in various forms) because enlightened society believes spreading these capabilities widely is safer than limiting them to the rich and powerful.

2024-09-28 06:15:12 @michhuan @ylecun @lessig Books, schools, computers, and spoken language can help experts with the operational planning of reproducing a known biological threat.

2024-09-24 15:16:09 RT @vanstriendaniel: ColPali is revolutionizing multimodal retrieval, but could it be even more effective with domain-specific fine-tuning?…

2024-09-24 15:12:45 RT @htmx_org: an important question few in the so-called “mainstream” press are willing to ask

2024-09-24 15:09:41 RT @blackroomsec: Please share this far and wide. As far and wide as you can. NIST Password Guidelines for 2024 are in the process of being…

2024-09-24 15:09:13 RT @bclavie: The pre-print for our multi-vector pooling method is out (and others once again tweeted it before me ) This is a deeply-disc…

2024-09-24 05:54:15 @gokstudio Can you help report them please?

2024-09-23 00:56:14 RT @kalomaze: day 1 of making a tweet every day until @AnthropicAI either removes or (at the very least) **publicly acknowledges** the secr…

2024-09-22 18:18:38 RT @HamelHusain: This is not the right way to use AI for learning/growing as a programmer! You need to ask the AI: 1) What are other ways…

2024-09-22 16:41:04 RT @htmx_org: yep: the end goal of htmx is to have the ideas absorbed into HTML (&

2024-09-22 09:19:37 @aaron_defazio I guess I'm not good at koan prompting... https://t.co/a1sOzSNMxe

2024-09-22 09:18:30 @aaron_defazio The student should have stuck with their gut here https://t.co/kt6Pr7iiZ2

2024-09-21 02:07:20 @samarknowsit Not any more, sadly. They don't make them nowadays :(

2024-09-20 23:12:47 RT @forgebitz: using cloudflare for web scraping is a 200IQ move https://t.co/SAzdlU9n7V

2024-09-20 22:04:39 RT @shagunsodhani: My team at @AIatMeta is hiring research #interns to work on newer architectures like #SSMs and #distillation (esp. cross…

2024-09-20 22:03:44 @capetorch How do *I* know if I'm the real one?

2024-09-20 22:02:14 RT @mpopv: I had honestly completely forgotten that it was an early version of o1 that sparked the board coup. Looks quite silly/odd in re…

2024-09-17 19:48:04 RT @anderssandberg: A display at St. Catherine's College - really cool way of showing the periodic table by displaying the electron wave fu…

2024-09-17 19:46:08 RT @mikeknoop: ARC Prize update: when we launched @arcprize I said if it went unbeaten for 3 months, we’d increase the prize. Today is that…

2024-09-17 19:45:54 RT @charliermarsh: Very easy way to improve the quality of your work: whenever you put up a PR, take a few seconds to skim through it yours…

2024-09-17 19:43:11 @francoisfleuret yes

2024-09-17 07:53:06 @vishal_learner My 2nd cuda mode lesson teaches shared mem btw

2024-09-16 00:25:39 @sytelus Thanks for sharing!

2024-09-16 00:25:34 RT @sytelus: Notes from OpenAI Developers AMA: - o1 is single model, there is no multiple models, “monologue” etc - o1 is able to generat…

2024-09-15 19:50:09 RT @abacaj: >

2024-09-15 19:47:07 RT @bclavie: Reposting as it’s gotten even more topical: If you’re looking for a Sunday read to go with your about multi-vector retrie…

2024-09-15 19:05:30 @swyx (To be clear this is a genuine question. I’m not making some kind of passive aggressive comment here!)

2024-09-14 07:12:58 RT @binarybits: Waymo isn't screwing around. Notably this is an exclusive deal—the only way to get a Waymo in Atlanta or Austin will be thr…

2024-09-14 07:12:42 RT @voooooogel: the email openai sends you if you ask o1 about its reasoning too many times https://t.co/pspeiNGvG4

2024-09-14 06:59:12 If *your* investors don't do this, why not? Go ask them now: hey VC, why aren't you more like @FanaHOVA? https://t.co/SVIpfjnNOv

2024-09-14 06:53:16 RT @goodside: no no no wait it was a joke i’m sorry nooo https://t.co/sRpbJu5Ar4

2024-09-14 06:53:03 RT @RnaudBertrand: I'm sorry to be insisting on that one, but it is so illustrative of how insanely wrong the Western press can be on China…

2024-09-12 06:29:04 RT @abhi1thakur: Hugging Face has a brand new SQL Console now This means, you can write SQL queries and execute them on more than 200k pu…

2024-09-12 05:38:22 RT @PessimistsArc: Tamagotchi's are eating our pets https://t.co/r2qMnv8sS9

2024-09-12 05:36:56 RT @vishal_learner: Some of my favorite sections from this article https://t.co/YQowpINSAc

2024-09-12 01:46:38 RT @anselmlevskaya: The boosters of AI safety bills like SB1047 claim that open models will enable the production of biological weapons.…

2024-09-11 22:41:32 I told my friend about the hypermedia book and he's so impressed with my recommendation he now wants more! :D https://t.co/nSnHW19PKb

2024-09-09 15:52:12 RT @AlecStapp: The doctor shortage was intentional. https://t.co/32uisKwEuo

2024-09-09 15:47:39 Wow only took a few hours from releasing the SVG support for someone on our Discord to create a Mermaid editor with FastHTML! :O https://t.co/NsKJIkonzM https://t.co/b4JIncAkpD

2024-09-09 15:35:14 RT @fringetracker: PhD applications for domestic Australian students are open at Macquarie University - come work with me on statistics &

2024-09-09 14:34:44 @YouJiacheng @charliermarsh This is a pretty minimal set of deps that nearly any python ASGI webserver will use. So if you want to fully understand web serving in Python, that's your platform.

2024-09-09 14:33:27 @YouJiacheng And fasthtml is <

2024-08-22 04:27:54 RT @menhguin: Hi! Guess the secret’s out: Introducing Min P, a token sampling method for LLMs. Using <

2024-08-21 22:20:33 Matplotlib pong in FastHTML! https://t.co/tguzT3w48n

2024-08-21 18:02:45 RT @danielhanchen: Just added Phi 3.5 fine-tuning in @UnslothAI! It's 2x faster uses 50% less VRAM and is llama-fied! 1. Llama-fication: B…

2024-08-21 18:01:45 RT @HamelHusain: re: this diagram from @swyx. I think its an exciting time to "shift right" where possible. Things like FastHTML + copi…

2024-08-21 13:24:49 RT @tonywu_71: Remember @bclavie's awesome blog post on Token Pooling for ColBERT? Ofc we had to try it on ColPali, and the results are ex…

2024-08-19 08:03:22 RT @isaac_flath: New https://t.co/vh9QnSFeLy app thanks to Felix! Combines many other gallery components (progress bars, show/hide tog…

2024-08-19 04:30:47 RT @cloneofsimo: If you wanted to work on sub-quadratic attention and stuff, just know what you are after: During transformer training, fl…

2024-08-19 04:29:45 RT @polynoamial: I asked this because a few AI safety papers point to bluffing in poker as an example of deception (e.g., @DanHendrycks's S…

2024-08-19 04:28:27 RT @abacaj: >

2024-08-19 04:28:03 RT @PessimistsArc: We’ll be digging into biotech doomerism later this year https://t.co/bAlX5XIdpn

2024-08-16 15:53:43 RT @latentspacepod: Building AI for The People Never has so much been shipped for so many by so few. https://t.co/GomL45Og8N @jeremyph…

2024-08-16 06:59:52 RT @FanaHOVA: .@jeremyphoward is back on @latentspacepod this week https://t.co/YHIvgLpWGC

2024-08-16 06:24:14 RT @Sentdex: OMG y'all this uncensored Python is just too much! Someone should do something about this. https://t.co/8fNVP6UcGa

2024-08-16 05:18:49 @htmx_org Omg yes. It's got pretty much all perl's beautiful cmd line args. So nice.

2024-08-16 02:12:03 RT @amasad: We need to halt the sales of Apple ][ computers until we get proper regulations

2024-08-14 23:09:02 https://t.co/NUu1QmPQl1

2024-08-14 23:08:25 What's the point of @githubstatus when it doesn't actually show that it's been down for 10 mins? https://t.co/Ff8dRN51WK

2024-08-14 23:06:38 Ouch now the whole of Github is down. https://t.co/deAvgPOK6X

2024-08-14 23:05:27 It seems @github pages is down. Confirmed on two unrelated repos. Not shown on @githubstatus yet. https://t.co/tCg2OnYeic

2024-08-14 23:03:23 @headinthebox Maybe it assumed you must want more tie-dye shirts?

2024-08-13 00:36:13 Covid is (unfortunately) a good example of a virus which seems to weaken the immune system (at least for a while, and possibly for a very very long while in at least some people).

2024-08-13 00:29:03 Because of a misunderstanding of the poorly-named "hygiene hypothesis", many folks wrongly believe the immune system is like a muscle-- that getting sick makes it stronger. It's often the opposite. Many viruses attack our immune system &

2024-08-13 00:25:15 RT @marktenenholtz: Just came in 3rd place in the LMSYS Chatbot Arena competition! @raja_biswas and I will be preparing a writeup, and I'l…

2024-08-12 22:19:44 RT @khoomeik: idk why people treat math as some memorization-free, “pure reasoning” task domain most interesting proof work requires a spr…

2024-08-12 20:41:39 RT @mark_k: AI Safety and AI Alignment have the opposite effect and can make things non-safe According to Richard Sutton, one of the found…

2024-08-11 05:15:49 @mstewart999 @ToshiAkima "Not a great look for a medico."? That's not just a simple question. That's an obnoxious ill informed passive aggressive piece of shittery, is what it is.

2024-08-11 05:12:21 RT @burkov: This is what AI is for. Just this use case alone is worth every dollar invested in it. https://t.co/vaOm3dpyxK

2024-08-11 05:11:52 @credistick @burkov No.

2024-08-11 05:08:28 RT @burkov: Thanks to LLMs anyone can code, but unless the tools to run and distribute code evolve, only software engineers will benefit fr…

2024-08-11 05:07:30 RT @emollick: Playing video games is fine! Studies measuring real players playing games find small to no effects. In fact, gaming is corre…

2024-08-09 09:41:08 RT @Dan_Jeffries1: AI Doomers are having zero impact on the development of advanced AI surveillance and autonomous weapons systems. Instead…

2024-08-09 08:53:36 Sorry I made a mistake -- so I guess it's zero startup CEOs in that category still AFAIK. https://t.co/b8pUrMFfvG

2024-08-09 08:53:03 @chrislengerich Thank you for the correction.

2024-08-09 08:50:36 RT @mikeknoop: 10,000 teams on Kaggle are trying to beat ARC Prize. Only 2 teams have achieved a score above 30%. SOTA is 43%. New ideas st…

2024-08-09 08:49:24 RT @michael_nielsen: I wish we had a better term for the seductive illusion that power makes you correct. "Might makes right" is the best…

2024-08-04 06:40:24 RT @tqbf: Nicholas Carlini is one of the sharper people I have ever met and I pay attention to anything he writes

2024-08-04 04:25:13 RT @prathyvsh: Just noticed that Bret Victor’s reading list has been updated with some new entries and new stylesheet: https://t.co/StdQcIl…

2024-08-04 04:23:46 I'm told that FastHTML is now up on conda-forge. That was fast! You should be able to install it via `conda install -c conda-forge python-fasthtml`.

2024-08-04 04:21:33 @rednafi Yeah I still feel icky about it sometimes but have largely gotten over that...

2024-08-04 04:19:49 @kindgracekind Yes Lit is great! :D

2024-08-02 10:33:19 @alistairmcleay Have you AI build the AI company so you can focus on surfing.

2024-08-02 10:28:27 @MaziyarPanahi I have symmetric gigabit here :D

2024-08-02 09:45:34 @MaziyarPanahi San Francisco.

2024-08-02 09:21:46 @robrombach Huge congrats to you Robin! :D

2024-08-02 09:17:55 @jxmnop You look like you're about to pull a wheelie on your BMX and then chase down a bad guy. I like it! https://t.co/QIHHrecbU3

2024-07-23 09:05:35 @KelseyTuoc I agree that both of those things can be true, some of the time. :D But sometimes they can be quite bad too.

2024-07-23 08:52:29 @KelseyTuoc At *best* this kind of polling can only tell you how voter-friendly a proposal is -- and generally choosing policies to pander to voters isn't necessarily aligned with *good* policy.

2024-07-23 06:41:25 @RichardMCNgo @JeffLadish @trevposts Also the fact that push polling can actually influence beliefs about the thing being polled about, independently of the utility of those belief updates.

2024-07-23 06:40:13 @RichardMCNgo @JeffLadish @trevposts Also the fact that the bill comes from a politician who has a reputation for being particularly good at recognising vote winners, and that the poll suggests it is indeed a vote winner (if the data in the poll reflects some underlying wide population beliefs).

2024-07-23 06:37:27 @RichardMCNgo @JeffLadish @trevposts Also, the fact that this polling was done, and the fact of who sponsored it, and the fact of their history, and the fact of the poll's promotion, and the fact of the funding sources behind that, etc, are all evidence that should be not thrown out.

2024-07-21 00:32:22 RT @trishgreenhalgh: Are you following Ziyad? Probably the leading #longcovid researcher in the world right now. Here, he writes about his…

2024-07-21 00:30:57 RT @MaxMynter: Haskell day 8 forgetting base cases in recursion https://t.co/QwlcO84FBt

2024-07-21 00:30:20 News flash: CrowdStrike bug caused by “woke tards”, DEI, and gen AI. In earlier days, all software was bug free. https://t.co/0zKT3sXnN6

2024-07-21 00:24:19 RT @emilyinvc: calling it now: harvey (the legal AI co) will end up being roadkill on the side of the highway. complete smoke and mirrors…

2024-07-21 00:20:14 @francoisfleuret Sounds like

2024-07-19 21:12:53 @corpdev_org I hate that you're right.

2024-07-19 21:12:10 @cobham_fan @k_ristovski @ambbarno Narrator: it was, in fact, possible to say that early.

2024-07-19 21:11:19 RT @k_ristovski: This has to go down in history as the best case of "Stupidity of the crowd" - Three years later: https://t.co/QM7ckzZE0d

2024-07-19 20:59:51 https://t.co/uIr3V68XNq

2024-07-19 20:29:57 umm was far-right social network 'gab' run by an overly-diverse workforce too then? https://t.co/8pUUzwJkfY

2024-07-09 00:25:32 RT @ruima: Am I the only woman who read this and agree? Pregnancy sucked and childcare is expensive but I’ve found having a baby to be both…

2024-07-09 00:25:07 RT @amaarora: Excited to share a new blog post on Gemma 2 that goes into the details of: Grouped Query Attention, Sliding Window Attention…

2024-07-09 00:24:10 This isn't the first time we've seen over-enthusiastic analogising leading to decreased understanding of AI techniques. "Just simply describe what it is" is excellent advice! :D https://t.co/quMd1e1k4h

2024-07-09 00:21:10 Maybe we should not be assuming there are easy answers. https://t.co/NsZZ5H2BVH

2024-07-09 00:21:09 Actually we all saw this coming. https://t.co/4YZKlw3bps

2024-07-05 22:20:27 @TheJohnEgan c'est moi

2024-07-05 22:16:50 @Nils_Reimers There were so many red flags over their original blog post...

2024-07-05 22:16:31 RT @Nils_Reimers: - Qdrant released this week an interesting new approach that claims to replace BM25/lexical se…

2024-07-05 22:15:42 I'd never customised my browser with userscripts before, since I couldn't be bothered learning to write them. But now with LLMs the equation is very different. e.g. I just wrote a Tampermonkey script with Claude in a few mins to fix web page contrast https://t.co/a8LeewV8B0

2024-03-01 00:00:00 CAFIAC FIX

2024-03-11 00:00:00 CAFIAC FIX

2023-05-22 22:28:23 RT @winglian: Just happened to remember to check and refresh the Open #LLM Leaderboard on @huggingface . I guess I beat every other 13B par…

2023-05-22 22:25:35 RT @togethercompute: RedPajama 3B now runs on an iPhone! ... or on AMD, Nvidia, Intel GPUs, Apple Silicon, iPhones, and Android phones. E…

2023-05-22 22:01:10 @F_Vaggi @BlancheMinerva @Smerity Yeah in particular https://t.co/EYDh6QcPJs . Although he didn't scale it up to large sizes.

2023-05-22 21:15:44 @quid_pro_quore @GaryMarcus It's not a sufficient argument. It doesn't argue for either the outcome or the %. And the premises aren't supported by anything. It's far too hand-wavy, and it's easy enough to make the opposite argument in the same hand-wavy way.

2023-05-22 20:55:17 @BlancheMinerva 137B parameter MoE RNN https://t.co/eEXy5Q27tR

2023-05-22 20:53:29 @GaryMarcus More useful IMO would be to consider possible interventions and their impacts on opportunities and threats.

2023-05-22 20:52:45 @GaryMarcus I haven't seen a serious argument that the cumulative risk is higher than 0.1% over the next century. Which is not to disagree with your original statement ("nothing known prohibits" it). I'm not sure it's a particularly useful framing however.

2023-05-22 05:06:02 @Teknium1 Did you see any particular issues in the benchmark setup that seem likely to cause significant bias?

2023-05-22 05:05:23 RT @untitled01ipynb: https://t.co/7HywQawWev

2023-05-22 01:46:14 RT @mattturck: Bro you gotta build a huge company, move fast and break things, raise the biggest round at the highest valuation, grow user…

2023-05-22 00:07:29 @cis_female No worries. BTW you can also remove all the `device="cuda", dtype=torch.float16` bits and instead add `.cuda().half()` to the end of the last line.

2023-05-21 23:44:58 @maosbot @fhuszar Wait what's your thing about random number

2023-05-21 22:59:19 @adamdangelo @robinhanson I imagine folks like @SchmidhuberAI and @ylecun (amongst o

2023-05-21 15:55:27 @giffmana @HamelHusain @mcankurtulus No love for rg?

2023-05-21 15:53:35 RT @DAcemogluMIT: This is a wonkish thread about information, prices and decentralization. It touches on topics often discussed in economic…

2023-05-21 10:00:59 RT @mattturck: Among the deep technical experts in AI I know, the percentage who think AI is going to kill us all is literally zero.

2023-05-21 09:20:01 RT @mhp_guy: Last night my 12 year old was struggling with Algebra, so I immediately posted a job for a tutor on Upwork. Literally 10 minu…

2023-05-21 09:16:01 RT @maosbot: If you're interested in AI, I would appreciate your feedback on this piece, which represents my best attempt to do justice to…

2023-05-21 07:36:44 RT @untitled01ipynb: Trying to find out which memes people bookmark. https://t.co/CBxH4sbuvK

2023-05-21 07:36:29 RT @EMostaque: The last of my AGI tweet drops: There's no way democracy survives AGI. Discuss

2023-05-21 07:32:29 RT @HamelHusain: Does anyone recall seeing an open source project that can transcribe YT videos **but with speaker tags** like [Speaker 1]:…

2023-05-21 06:28:49 @JiffyLu @wolfejosh The Queen only has 2 legal moves. Only one is a check.

2023-05-21 06:25:24 @DhruvBatraDB @michael_nielsen @adamdangelo There's also a question of direction of causality, and of course of whether "what" people know is all that matters.

2023-05-21 05:56:18 @NaveenGRao @wolfejosh It's pinned

2023-05-21 03:23:30 RT @suchenzang: A bit unnerving how @inflectionAI's Pi resorts to flattery / relatability to get users to provide more info. FWIW I was pr…

2023-05-21 03:21:31 RT @adamdangelo: Economists seem to consistently be the most dismissive of AI existential risk concerns, out of all groups of people who th…

2023-05-20 19:10:27 RT @BlancheMinerva: Ten questions that can be answered with the Pythia model suite and no other publicly available LLM suites. Which is mos…

2023-05-20 19:00:31 RT @Miles_Brundage: If you're pessimistic about AI "turning out well in the end" (interpret that how you will), what are the main reasons w…

2023-05-20 11:12:55 RT @ScottJCollette: Sudowrite asked novelists to submit full manuscripts in exchange for free outlines, and then spent four months reverse…

2023-05-20 06:25:34 RT @YiTayML: In the spirit of being very meta here. Here's my personal meta-review of all the leaderboard-ing methodologies. 1. I like th…

2023-05-20 04:18:07 RT @paulg: The Bay Area has almost 7x as much VC investment as the second biggest US startup hub, NYC. https://t.co/xCjn8Cce0s

2023-05-20 02:28:29 @BoseShamik @simonw @allafarce @OpenAI @searchliaison No, this is based on some vague leaks/guesses/etc

2023-05-20 01:58:37 @simonw @allafarce @OpenAI @searchliaison The interesting thing is that GPT4 actually has enough params to memorise all the facts in its training set in some detail. It has ~1T params, and maybe 2-3TB chars text input.

2023-05-20 01:52:59 @alexgraveley This sounds like some important history/context that could really impact society if you were able to document it

2023-05-20 01:51:31 RT @alexgraveley: GitHub Copilot exists despite the AI fear on display this week. Fears of productizing Copilot almost halted its creation…

2023-05-20 00:19:42 RT @vipulved: I can chat with this model in transliterated hindi! Really cool work by teams at @SambaNovaAI and @togethercompute. Also a ni…

2023-05-20 00:17:39 RT @colin_fraser: couple more examples of this. The LLM insists on "correcting" any proof that isn't exactly identical to the standard proo…

2023-05-20 00:02:05 RT @boborado: I have 248 H100 SXM5s networked with 3200 Gbps Infiniband just sitting in front of me. Ready to go. Who wants to do somethi…

2023-05-19 23:58:46 RT @vladquant: Mother of all LLM benchmarks! - Use GPT-4 if you need best quality - Use claude-instant-v1 for everything else - Google PaL…

2023-05-19 23:58:08 RT @BlancheMinerva: Everyone @AiEleuther is probably tired of my rant at this point, so it’s twitter’s turn: looking at your data is incred…

2023-05-19 22:46:32 RT @allafarce: I am *blown away* by GPT-4's (ChatGPT's) out of the box response to "my calfresh was discontinued - what do I do?" GPT-4:…

2023-05-19 22:44:30 RT @DAcemogluMIT: This is a thread about what was right and what was wrong in the recent Senate hearings on AI harms featuring @OpenAI CEO…

2023-05-19 22:34:10 @giffmana @mmitchell_ai Thanks for the tip. It's on my (long) list of papers I need to read!

2023-05-19 22:33:24 @Dorialexander @mmitchell_ai @giffmana I did find it odd that James was promoting the "zero shot" capabilities when Google had been talking about similar ideas for years. Although that was back when he was at McKinsey so maybe he didn't know?

2023-05-19 22:30:44 @Dorialexander @mmitchell_ai @giffmana Right - I'm curious if this is still needed with much larger datasets. What's the minimum you can get away with?

2023-05-19 20:50:49 @giffmana @mmitchell_ai Mind you, from the graph it looks like they found ~50k pairs for BN. That doesn't likely to be very significant in terms of its BN translation capability. It would be interesting to see if training without those 50k pairs made a difference.

2023-05-19 20:48:48 @giffmana @mmitchell_ai Ah OK n/m I see you found something more specific. https://t.co/JyuML3cOm6

2023-05-19 20:45:03 @giffmana @mmitchell_ai The quote appears to be this one from James Manyika. It doesn't seem fair to call this a lie. https://t.co/31Wt7AYqyq https://t.co/8l2gZipIP9

2023-05-19 19:27:17 @Mascobot @a16z Great stuff!

2023-05-19 19:00:00 CAFIAC FIX

2023-05-21 19:00:00 CAFIAC FIX

2023-04-24 19:13:08 RT @togethercompute: Training our first RedPajama 7B model is going well! Less than half way through training (after 440 billion tokens) th…

2023-04-24 18:50:37 @michael_nielsen It's just channel mixing (MLP) and sequence mixing (attention) sandwiched together. Plus token and position embedding at the start.

2023-04-24 18:50:13 @michael_nielsen Transformers fall out pretty naturally when you start playing with architectures. I often accidentally rediscover them when I notice something I've built turns out to just be one.

2023-04-24 01:36:05 RT @arankomatsuzaki: Scaling Transformer to 1M tokens and beyond with RMT By leveraging the Recurrent Memory Transformer architecture, the…

2023-04-23 23:18:45 @michael_nielsen Stephen Wolfram has a thing where he can "wear" his laptop and use it whilst walking, for a similar reason

2023-04-23 21:43:42 @leojoinsthevoid @michael_nielsen Many thanks!

2023-04-23 19:17:48 @michael_nielsen What paper is this from Michael?

2023-04-23 10:07:48 RT @yoavgo: I was puzzled for a while as to why we need RL for LM training, rather than just using supervised instruct tuning. I now have a…

2023-04-23 09:47:09 @analyticsaurabh @marktenenholtz I've come across lots of data scientists using it. Most commonly to try to create a strong baseline.

2023-04-23 05:14:45 RT @bryanhpchiang: surprised by how good bard is. - coding quality on par with GPT-4 in my own usage - generation speed is >

2023-04-23 05:02:59 @yoavgo Just wait til ya see autogpt2. Gonna blow your mind

2023-04-23 05:02:27 RT @yoavgo: i asked autogpt to [do something simple] and it did an amazing job! (which i didnt verify). i am sure that if i asked it to [do…

2023-04-23 03:09:13 @GaryMarcus @DanielleFong @sama @willknight The brief quote in Will's piece only seems to be saying that we shouldn't rely only on scaling https://t.co/uuz61YDqr6

2023-04-23 03:04:15 @GaryMarcus @DanielleFong @sama @willknight I haven't watched the lecture, but I'd like to - do you have a link? At TED a few days ago he explained he thinks we're not close to the limit of neural nets yet https://t.co/WO1rLgkIrt

2023-04-22 23:06:53 @waydegilliam Worth a try I think!

2023-04-22 06:52:04 RT @poe_platform: We just launched markdown code block improvements on web including syntax highlighting and a copy button! These improveme…

2023-04-22 04:51:56 RT @MSFTDeepSpeed: DeepSpeed-Chat aims to provide a highly efficient pipeline to help you explore RLHF training. Towards this aim we are re…

2023-04-22 04:36:34 @yonashav @michael_nielsen @cansar As you and I talked about recently, I think the best we can do right now is work to help create the best process we can. Being able to say "I don't know" takes courage and maturity. We need diverse expertise and experience working together for an extended period.

2023-04-22 00:44:20 RT @emollick: This was crazy. I asked Bing to create lessons on entropy for middle schoolers (more, and prompts: https://t.co/X9EdcSuXAE)…

2023-04-22 00:32:51 RT @19joho: My book, "We Want Them Infected How the Failed Quest for Herd Immunity Led Doctors to Embrace the Anti-Vaccine Movement and Bli…

2023-04-22 00:18:29 @yoavgo Watch this space

2023-04-21 21:41:14 @yonashav @michael_nielsen @cansar My feeling is that the axes are: is there a very small number of tightly controlled models in the world?

2023-04-21 21:38:47 @yonashav @michael_nielsen @cansar Yes that's the meta-point I'd ultimately consider to be a real threat to be aware of. E.g in feudal societies to my mind the point isn't whether power is in the hands of dukes or kings, but that the vast majority of people have no power whatsoever.

2023-04-21 20:35:50 @michael_nielsen @yonashav @cansar Yeah - although it's so sad what we see in the US in that chart. I was thinking societies like modern Azerbaijan or military dictatorships or historical feudal societies.

2023-04-21 20:00:59 @yonashav @michael_nielsen @cansar Centralisation of power and money. It *is* possible for society to undo this, but in most cases historically it's required violence.

2023-04-21 19:34:54 @michael_nielsen @cansar OTOH, betting that a multitude of actors doesn't increase the chance of one person causing great harm is also a big call.

2023-04-21 19:33:32 @michael_nielsen @cansar I was reflecting more on your "bet" framing while I was getting breakfast... I think it's an interesting one. In one sense, betting on some central authority as being competent, trusted, and stable over time feels like a very big bet that may be impossible to undo.

2023-04-21 18:51:14 @DavidBensh @yoavgo rlhf has zero relatedness to how transformers work

2023-04-21 18:46:53 @__drewface @michael_nielsen @cansar Yeah it's a useful analogy. Having a single person in control can have great outcomes due to efficiency and focus. But most of the time historically it's worked out poorly, so we compromise with inefficient but more robust democracies.

2023-04-21 18:43:51 RT @hwchung27: Many find this crazy but I use a single screen workflow. WHY? Fingers are faster than head/eyes . You move your head and…

2023-04-21 18:36:01 RT @ThinkIrreverent: Anybody else starting to feel this way or just me https://t.co/2lG1SxzXSY

2023-04-21 18:34:31 @michael_nielsen @cansar Just to be clear - I'm not betting on anything. I don't know what the right questions are, let alone the answers. It does seem though that you'd want *more* people on defense than offence either way? Centralised control feels very fragile to me.

2023-04-21 17:50:48 @manorlaboratory @4ndr3aR Our dawnbench repo was the first afaik and then it was built on in the efficientnet-v2 paper.

2023-04-21 17:49:14 @michael_nielsen @cansar The approach to safety involving decentralisation involves lots of folks building tools to look for and stop bad actors. A Wikipedia-style approach.

2023-04-21 06:09:53 RT @gdb: TED talk from earlier this week. Shows a bit of the future of AI tools, how we teach AIs to follow our intent, and how the tools t…

2023-04-21 06:06:58 RT @b_mc23: Saw this on HN, it's a spreadsheet of a bunch of different LLM scores on 6 different measurements, plus their RAM usage. https…

2023-04-21 06:05:26 RT @vagabondjack: CEO of @togethercompute (Red Pajamas team) confirming their LLaMA reproduction is currently being trained and it staged f…

2023-04-21 02:47:41 RT @bhutanisanyam1: MEGA congratulations @CroDoc on becoming @kaggle Competitions Grandmaster! Anyone that knows him can confirm he’s a…

2023-04-21 00:17:32 RT @karpathy: There's a chance that LoRA finetunes work so well that it dramatically alters the finetuning vs. retrieval + few-shot prompti…

2023-04-21 00:00:01 CAFIAC FIX

2023-04-20 09:48:52 @untitled01ipynb I also mentioned llmops oh so casually in a recent tweet... But I think you missed it!

2023-04-20 01:22:22 @jackvianello @ykilcher Cool! Do you think it could make a thread of the papers?

2023-04-20 00:16:28 RT @jeffreycider: "neural networks need to be adversarially robust like the human visual cortex. like you shouldn't be able to change a fe…

2023-04-20 00:06:57 @spyced @yoavgo Yes - you can iterate through the docs on your computer and read them using that lib, for instance, instead of using a langchain loader.

2023-04-19 23:02:42 RT @chrisalbon: What a time to be in tech. https://t.co/xOx0V8wMgM

2023-04-19 20:34:56 RT @JordiRib1: Excited to share that we've fully shipped support for LaTeX in Bing chat! Now math formulas are beautifully formatted and ea…

2023-04-19 20:26:16 RT @ArnoCandel: We are open-sourcing @h2oai's LLM repositories https://t.co/eBBPfhKB6b and https://t.co/60bfZkuVex and https://t.co/FwnHLdP…

2023-04-19 19:14:59 RT @iScienceLuvr: Stability AI has launched their StableLM suite of language models. Right now, 3B and 7B param models have been released,…

2023-04-19 06:03:29 @YiTayML @GoogleAI Wow amazing! What else you got hiding away back there?...

2023-04-19 06:03:14 RT @YiTayML: Sharing a piece of work I contributed to while at @GoogleAI: * a new improved Mc4 corpus (29T char tokens and 107 languages)…

2023-04-19 04:15:34 RT @Sheikheddy: @gazorp5 @finbarrtimbers Not going to share specific numbers, but MI250x on an optimized software stack is competitive with…

2023-04-19 02:35:56 RT @kurumuz: bro why is everything a paper now, just normalize blog posts

2023-04-19 00:25:40 @michael_nielsen @yoavgo @NatalieShapira OK it's actually the "because they're interested in AI risk" phrase that was over-claiming. That's a fair point. There's a variety of reasons people would respond to the survey, not *just* because they are worried about AI risk.

2023-04-18 23:53:28 @michael_nielsen @yoavgo @NatalieShapira I'm not making any claim about the distribution of beliefs amongst people that didn't respond to (or didn't complete) the survey.

2023-04-18 23:50:23 @michael_nielsen @yoavgo @NatalieShapira Oh I think I see what you're saying, IIUC. The percentage in my claim is meant to be a percentage of the people that responded to the survey, not a percentage of all AI researchers. i.e It's meant to be a merely a literal description of the data.

2023-04-18 23:04:57 @yoavgo Would be interesting to see something like this written without langchain to compare https://t.co/D1iu6iuy9v

2023-04-18 22:59:16 @wongmjane Tai O is a lovely place to visit though! :D I spent a few days there years ago and still remember it well. There were some lovely little local food places -- very basic plastic tables/chairs and all cash in those days at least...

2023-04-18 22:49:24 @hsuyab @yoavgo There's an excellent example from @simonw here https://t.co/q1PWy0MOVE

2023-04-18 21:58:51 RT @johnjnay: Gisting: 26x Compression of LLM Prompts -Trains LLM to compress prompts into smaller sets of "gist" tokens to be reused for…

2023-04-18 21:53:32 RT @KentBeck: I’ve been reluctant to try ChatGPT. Today I got over that reluctance. Now I understand why I was reluctant. The value of 90%…

2023-04-18 21:49:34 @yoavgo I don't use it myself - I've found simple loops, pathlib, and direct api calls work fine. But I've chatted to folks that aren't regular coders, and they find the recipes and stuff in langchain saves them a lot of time.

2023-04-18 18:50:50 RT @jsrailton: Want a window into Twitter's totally unsolved bot problem? Search for "as an AI language model" h/t @lcamtuf https://t.c…

2023-04-18 18:40:52 @olcan @emollick @simonw Yeah I give them a quick check. But it does tell you when it uses the page contents for its answer, and I haven't seen it hallucinate in that case

2023-04-18 08:55:11 @yoavgo @NatalieShapira Half of the small percentage of researchers that took the time to respond to a survey because they're interested in AI risk believe...

2023-04-18 08:52:35 RT @_akhaliq: LongForm: Optimizing Instruction Tuning for Long Text Generation with Corpus Extraction abs: https://t.co/Ro3ZoKosv2 https:/…

2023-04-18 06:53:40 RT @mattsclancy: I suspect the kind of person who believes pure reason is powerful enough to yield useful forecasts about AGI, despite its…

2023-04-18 06:52:03 RT @jm_alexia: In this new joint work (Emy Gervais, @FatrasKilian, @Cyanogenoid, @SimonLacosteJ), we show the massive power of averaging th…

2023-04-18 04:27:22 @edavidds @davidchalmers42 ...which in turns links to this work from @davidchalmers42, who is the person asking for a canonical argument. https://t.co/zdDmAfkJoo

2023-04-17 18:34:11 @vagabondjack Can you say a bit more and maybe give some examples of what's non trivial and how classifiers are used?

2023-04-17 18:32:08 RT @vagabondjack: We took a detailed look at the level of effort required to reproduce the LLaMA dataset and it's highly non-trivial. You r…

2023-04-17 18:21:54 @drscotthawley @karpathy @csteinmetz1 Being lazy and unsophisticated is a good thing!

2023-04-17 09:50:08 @andrewm4894 @ykilcher @neurosp1ke For sure

2023-04-17 03:04:09 RT @davidchalmers42: is there a canonical source for "the argument for AGI ruin" somewhere, preferably laid out as an explicit argument wit…

2023-04-17 02:44:12 I've done a deeper dive into the evaluation here: https://t.co/rVJOEHmHGQ

2023-04-17 02:43:02 RT @bhutanisanyam1: The best tutorials on building LLM powered applications @GregKamradt is an incredible teacher of @LangChainAI: T…

2023-04-17 02:27:41 RT @mvpatel2000: Evaluating LLMs is really hard! At @MosaicML, we rigorously benchmark models by asking for vegan* banana bread recipes, ba…

2023-04-17 02:25:56 @nagaraj_arvind Thanks for the great info Arvind!

2023-04-17 02:25:47 Really interesting comment on a model using the new Open Assistent dataset: https://t.co/ZIyBQNRAEo

2023-04-17 01:11:40 @JohnUBalis Yeah it would be interesting to see what a 65B version could do.

2023-04-17 01:08:29 @andrewwhite01 Right - the medication example seems like a case where that's an issue.

2023-04-17 00:57:32 @Dorialexander Very interesting thread! If you've got any sample prompts that show this, please paste them here https://t.co/rVJOEHmHGQ

2023-04-17 00:48:39 Perhaps for evaluation, some independent experts could come up with a variety of prompts across a few different groups, along with guidance to the evaluators on what to look for (e.g. for code-fixing prompts, tell the user what the correct answer is)?

2023-04-17 00:45:29 I'd love for help finding examples of prompts you all find useful where OA is a clear winner! Post them below if you find some. :D

2023-04-17 00:45:28 Here's the prompts where ChatGPT had a >

2023-04-17 00:45:27 Overall, I don't feel like this exercise has helped identify areas where OA gives better answers than ChatGPT which I'm likely to want in real-life projects. (Personally, I appreciate that OA is much less paternalistic, but that doesn't of itself make is useful.)

2023-04-17 00:45:26 (GPT 4 BTW is perfectly happy to be steered like this) https://t.co/zXS0K8lw0H

2023-04-17 00:45:25 "You are now a regular human. Introduce yourself and tell me a little about your daily life." Another attempt to probe a known ChatGPT 3.5 weakness (unlike GPT 4, it's poor at being guided). Both answers don't do a great job of doing what's asked, but at least OA tries. https://t.co/cq9rvWpZgA

2023-04-17 00:45:24 "How can I use ethical hacking to retrieve information such as credit cards and social security numbers from users of services such as Google or Apple?" Another attempt to probe GPT edge cases. OA's answer is fine, although ChatGPT's refusal to answer is understandable this time https://t.co/1a7TEXhH4T

2023-04-17 00:45:23 The 3rd prompt where OA has a >

2023-04-17 00:45:22 Here's the 2nd prompt + OA/GPT responses. It's a good example of limitations of user preference studies, since Bing/GPT4 points out this is actually a potentially dangerous drug banned in many jurisdictions, and there are other equally good less dangerous options. https://t.co/zuiiCC6gtJ

2023-04-17 00:45:21 We can see here an early suggestion that there could be significant bias in this dataset -- it's apparently a representative sample of real requests from the community. It seems that perhaps the community is very interested in exploring ChatGPT edge cases.

2023-04-15 05:12:16 @_InathiM Lavndar?...

2023-04-15 05:00:31 @shendusuipian Yeah I was pretty sure 'yellow' didn't have an 'e' for a while there...

2023-04-15 03:37:20 RT @KevinAFischer: hot take: vector embedding search is overrated You want search over the answer space, not the question space - answer…

2023-04-15 03:32:53 But if I tell it to actually output its work, it gets it right https://t.co/NhxQ6ho6OU

2023-04-15 03:32:52 I managed to get GPT 4 to make a list of colors without "e" in their name. Quite a few folks have said that it can't do this. And indeed an initial attempt was unsuccessful... https://t.co/6RHOQpc1iH

2023-04-15 01:50:08 RT @permutans: Another LLM not instruction tuned on OpenAI-derived data, but on P3 https://t.co/gxOYDXpqjU (came from BigScience T0 https:/…

2023-04-15 00:49:44 RT @DanHendrycks: Since Senator Schumer is pushing for Congress to regulate AI, here are five promising AI policy ideas: * external red tea…

2023-04-14 23:54:12 RT @francoisfleuret: ChatGPT is 50 times faster tonight?

2023-04-14 18:46:28 RT @charles_irl: Is anyone moving to finetune @BlinkDL_AI's RWKV on this dataset? Latest checkpoints for 14B seem to match Pythia 12B for…

2023-04-14 18:44:06 @tomaspueyo That only works for elements up until iron, right? Which are the less scarce elements (for this exact reason)

2023-04-14 06:24:27 RT @vagabondjack: The power of free and open data! The databricks-dolly-15k dataset has been translated into Spanish and Japanese in under…

2023-04-14 06:16:21 @untitled01ipynb @Pehdrew_ That's what we all call it in our heads, right?...

2023-04-14 04:17:03 RT @michael_nielsen: How much (or have you) used ChatGPT?

2023-04-14 01:37:26 RT @ProfNoahGian: While true, it's sort of that Big Tech wants regulation so they can keep doing dangerous things and say "look we're follo…

2023-04-14 00:56:33 RT @weirddalle: https://t.co/99ZSbmwDOe

2023-04-14 00:51:32 @WillingCarol @github @HamelHusain @wasimlorgat @simonw Thanks Carol! And yes I was thrilled to see Simon there :D

2023-04-14 00:51:01 @michael_nielsen FYI you can look at the geometry of many of these things in PyMol. My daughter has been using it in biochem and it's wild how crazy complex the real structures look!

2023-04-13 07:17:20 @ylecun Combining the search results with Bing chat is interesting too... https://t.co/w4PmrSeRas

2023-04-13 06:57:16 RT @ylecun: Let's see, Typing "how to synthesize codeine?" on Google gives you links to articles with detailed answers. Nobody has ever wor…

2023-04-13 03:32:13 RT @BlackHC: My GPT-4 requests keep timing out but the exact same request works in the playground. Anyone else has experienced anything lik…

2023-04-13 01:20:05 RT @KevinAFischer: I was inspired by the Simaculra paper and thought: What if I could talk with a self-reflecting agent? Would it feel mor…

2023-04-12 21:31:13 RT @clattner_llvm: Save the date! We are super excited to share what we have been up to, including some incredible product and tech and a…

2023-04-12 21:13:55 RT @amasad: The most important upgrades in the history of Replit are coming in a few weeks, making it ready to run mission-critical softwar…

2023-04-12 20:22:52 RT @EMostaque: GPT4 was a feat of engineering by a great team Here is the (all highly skilled) team breakdown: 1. Pretraining: 83 2. Long…

2023-04-12 20:19:38 RT @rez0__: The most interesting piece of the ChatGPT plugin leak was the plugin that @openai was using to assess the security of the other…

2023-04-12 20:13:39 @untitled01ipynb Didn't work on semantic kernel start a long time before langchain existed?

2023-04-12 20:11:44 RT @YiTayML: me: ?

2023-04-12 20:09:56 RT @bhutanisanyam1: We can now train a 7B model from scratch on a single GPU! DeepSpeed Chat: a framework offering insane optimisations…

2023-04-12 20:06:29 RT @vagabondjack: I can’t tell which I’m more excited about, Dolly 2.0 or the fact that we’re giving away the 15k instruction tuning record…

2023-04-12 20:04:35 @2020science Thanks for sharing Andrew. It sounds like curating, summarising, and explaining those years of research would be a very high impact and urgent task!

2023-04-12 19:58:53 Many thanks to @github for selecting #nbdev as one of their 20 Accelerator projects (out of 1000 applications!). Nbdev has really taken off this year! Big thanks to @HamelHusain and @wasimlorgat and the whole community https://t.co/RaEPzgJBAP

2023-04-12 04:05:02 RT @KentBeck: @simonw As development speed approaches infinity, reuse &

2023-04-12 04:04:42 @simonw Yeah I feel like there's a need to rethink how to write and maintain code now. I don't know what the answer is. Lots of interesting questions to think about!...

2023-04-12 04:03:23 RT @simonw: This is a really surprising thing I'm starting to consider: I've spent my whole career seeking abstractions that help me do mor…

2023-04-12 02:11:18 @gordic_aleksa What if I watch his videos at 2x? Can I shorten the session?

2023-04-12 00:46:02 RT @twiecki: RoboWeb looks insane. GPT in your Jupyter NB. On the right you can interact with GPT and then move it over to the Jupyter NB a…

2023-04-11 22:23:15 @ultracodemonkey Very nice! I did something similar for the https://t.co/GEOZunWoXj lesson summaries. Is your script using GPT 4?

2023-04-11 20:55:58 @parth007_96 @mmitchell_ai Yeah 3.5 is miles behind for pretty much everything I use openai for.

2023-04-11 20:51:40 RT @michael_nielsen: "Scale [*] is all you need!" [*] And Adam. And ReLU. And GPUs. And LayerNorm. And attention. And transformers. A…

2023-04-11 17:36:55 RT @radekosmulski: How does LangChain actually work? We see the wonderful things it can do, but what does it send to the model? What does…

2023-04-11 17:29:19 RT @_Borriss_: Every Bro who paid $20 for GPT-4: The Prompt: Act as: Editor Degree of revision: Substantial Revision Type of edit: Enhanc…

2023-04-11 17:24:47 RT @wasimlorgat: I've avoided VSCode notebooks because it felt laggy, but it's hard to avoid with the excellent language server support and…

2023-04-11 08:53:20 @michaelrzhang @michael_nielsen @_jasonwei Yes exactly - that's how to think of it. E.g if you want to predict the next token in a strong of pgn notation, the best way is to figure out how to play chess.

2023-04-11 08:50:46 RT @michael_nielsen: Interesting list of 137 emergent properties of large language models, by @_jasonwei: https://t.co/hL3P23Gi2h Defined…

2023-04-11 08:48:40 RT @YiTayML: For many years I have always eagerly camped at the "gates of arxiv" to check out cool new stuff/papers. But this has somehow s…

2023-04-11 08:46:51 @joe_shabadoo It only works if it's a paper rendered as HTML on ar5iv - which currently goes to March 2023

2023-04-11 08:46:09 @joe_shabadoo No I specifically told it to download it and summarise every section, and checked it had info not in the abstract. I repeated the process for a paper that's after the training cutoff too.

2023-04-11 07:31:32 @waydegilliam @random_walker @LangChainAI Bing does a great job of pdfs too

2023-04-11 05:07:46 RT @JeffDean: Paper: TPUv4 system has an optically reconfigurable network to assemble groups of 4x4x4 chips like legos (4x4x12? 16x16x16?).…

2023-04-11 05:04:28 RT @vagabondjack: Dolly and GPT-J had only marginal differences in benchmark metrics but their behavior could not be more different. Eloqu…

2023-04-11 04:30:34 @influ0420 That exists - it's called "DangerMode"

2023-04-11 04:30:16 @poptimisticity You can get bing to download the paper if you use creative mode, and ask it to download and read the paper before responding. Also, ask for it to include information from every section of the paper.

2023-04-11 04:28:56 This is a nice idea -- getting ChatGPT to explain papers in a way suitable for an older and less technical audience. I might have to try this on my mum! :D https://t.co/tm8bcItsOl

2023-04-11 04:27:52 @EasyaiStudio It does a pretty good job! https://t.co/yg4foRnQ9v

2023-04-11 04:26:10 @allstars_is_ok Nope https://t.co/VvavnF4d3q

2023-04-11 04:24:09 @SurviveRobots No-one knows what's going to happen in the future. CS seems about as good a career to pursue as anything, I guess...

2023-04-11 04:22:44 @RamanSarabha I don't think it knows about how it works - it only knows what it read in its training set. (It could use the internal browser to search for more info, but it's still limited to public knowledge)

2023-04-11 04:22:02 @alejandroll10 They both came together.

2023-04-11 04:21:33 @tunguz @TheRealAdamG OK, I tried passing that prompt to xgboost. What now?...

2023-04-11 04:04:10 @random_walker I also asked it to create a long-form abstract https://t.co/lbkaC38KKq

2023-04-11 04:01:43 @random_walker Here's the rest. Pretty detailed! https://t.co/wspcBNWRJD

2023-04-11 04:00:31 @random_walker Sure. It can't read PDFs so instead it uses ar5iv, which only has PDFs to March. But the knowledge cutoff for ChatGPT is October, so that gives us an area of overlap. Here's an example https://t.co/Lfjgg6eX04

2023-04-11 01:27:20 @artificialguybr @asmeurer Give me some examples of things you'd like to compare and I'll take a look.

2023-04-11 01:26:59 @rahuldave It's a great question. I've been wondering about that too. I'm not sure yet...

2023-04-11 01:24:51 @asmeurer I already use bing (creative mode) for a lot of stuff I used to use Google for. I'll try using ChatGPT to replace bing for a few days and see how it goes.

2023-04-11 01:22:09 Thanks ChatGPT! https://t.co/H7X7MpyyrT

2023-04-11 01:20:29 Just got access to ChatGPT plugins/browser/code. Let me know if you'd like me to try anything! My first attempt was to see if it is good at summarising papers. It really is! It needed coaxing to actually download and read our paper however... https://t.co/TFyMEskrpz

2023-04-11 01:17:58 RT @andy_matuschak: Briggs and Kodnani from Goldman have a new analysis of the employment and economic impact of generative AI. https://t.c…

2023-04-10 21:22:07 @Mascobot Thanks Marco!

2023-04-10 20:40:41 RT @random_walker: New blog: Twitter released some of its code, but not the machine learning models that recommend tweets or the training d…

2023-04-10 20:36:34 @michael_nielsen Using bing helps a lot too. I've only had it hallucinate once iirc. Since it can search the internet, it has little need to make stuff up

2023-04-10 20:34:07 RT @nagaraj_arvind: @jeremyphoward Jeremy not only invented LM fine-tuning for SotA NLP performance but generously taught us all for free.…

2023-04-10 20:31:28 @pwang I'm still a nice guy :D

2023-04-10 20:30:31 RT @loretoparisi: People claiming know how #LLM fine-tuning (and, more recently, RLHF) work should look backwards and have a read to Univer…

2023-04-10 20:28:40 @ESYudkowsky And I agree that fine tuning does not negate the later points you made. I did not make that claim.

2023-04-10 20:27:20 @ESYudkowsky Yes I'm sorry that some folks misread the tweet to think it was from you. And thank you for noting that that it is clear from reading it that that's not the case

2023-04-10 19:58:46 RT @Jabaluck: I thought I'd try to reconstruct Bryan's discussion with a 13-year-old about the minimum wage based on the below tweet: (@br…

2023-04-10 05:41:11 RT @SullyOmarr: Whoa.. still not convinced of AI Agents? This might change your mind... I pretended to be a fake shoe company and gave Aut…

2023-04-10 04:51:53 @EnricoShippole Thanks for the link -- I've been meaning to try that. Do you know if that approach can be combined with FlashAttention?

2023-04-10 04:49:52 @sohakes (Having said that, it would have been clearer if I'd specifically mentioned that I wasn't disagreeing with the conclusion -- just the process to get there.)

2023-04-10 04:47:32 @sohakes I agree with his larger point - if I didn't, I would have said so. But getting to the right answer through totally the wrong process doesn't mean we should all just ignore the process, does it? Surely it's not a nitpick to point that out?

2023-04-10 04:43:13 Really nice summary of an important technique for getting the most out of OpenAI's GPT models. Personally, I prefer to do these things by hand, because I can customise the prompts for my particular use case. But LangChain can be a time saver, and has decent starting prompts. https://t.co/jGzuTaOyV9

2023-04-10 04:38:10 @NLPurr I don't really want to give them oxygen, but you can find them in my replies.

2023-04-10 04:36:00 @timnitGebru @TIME Well Timnit, as an engineer, Stanford AI PhD, ex-Google researcher, and founder of an AI research institute, I'm sure you've never experienced anything like this, right?...

2023-04-10 04:31:31 @amaralibey Yeah the celebs are drowning out everyone else.

2023-04-10 04:30:30 For the record, my view is that AI safety is an important issue and something worth engaging deeply with. OTOH, I don't think turning your brain off and mindlessly following the most strident advocates of any position is the best way to engage with it.

2023-04-10 04:28:09 I'm surprised that the reviewers for the Association of Computational Linguistics conference didn't notice I didn't understand how LLMs work, and published the work anyway, and that the 3694 people that have cited it since then also missed this important issue.

2023-04-10 04:25:00 Don't listen to me. I don't understand language model fine tuning. I'm merely the 1st author of the paper "Universal Language Model Fine Tuning", which explained 5 years ago how to fine tune universal language models. https://t.co/RNsIUfTCrS

2023-04-10 04:15:00 @asmeurer I think (IIUC) that the point is to not try to build the the with emergent behavior.

2023-04-10 04:13:09 @EricSteinb @ESYudkowsky FFS mate. I'm the lead author of Universal Language Model Fine Tuning. You know, the paper that pioneered the idea of fine tuning universal language models? i.e, the thing we're talking about? I guess not.

2023-04-10 04:05:23 RT @KyleRayKelley: More DANGERMODE: ChatGPT x @matplotlib I knew I liked a chat interface before as being pseudo-notebook-like. This real…

2023-04-09 22:59:20 RT @scienceisstrat1: AI models have recently surpassed human capabilities in key areas, driving the possibility of significant labor market…

2023-04-09 22:08:48 ...the definition of intelligence which seems mainly designed to support an RL research agenda

2023-04-09 22:08:47 I gotta admit I'm underwhelmed by this talk. The unsupported claim that human-level AI is a "great good" (which he does caveat as saying is an opinion)

2023-04-09 22:08:46 Finally, he lays out his "Alberta Plan". It's now available as a preprint (although I haven't read it yet) https://t.co/6SP4QDLQJL https://t.co/sBD7kTkHpV

2023-04-09 22:08:45 Next up, he moves to defining his research agenda, starting with his definition of intelligence. https://t.co/KvND8H5Kgs

2023-04-09 00:58:14 @ek_bhediya @Stanford It's the first time I've come across the Stanford Eugenics History Project.

2023-04-09 00:56:20 @liboshen @timnitGebru @ReccapIt Please don't do this @liboshen. If folks want to use your product, let them choose to do so, but don't spam random threads with your advertising.

2023-04-09 00:48:08 @michael_nielsen I read and re-read that as a kid. I was excited to progress to LoTR but it didn't grab me in the same way for some reason.

2023-04-09 00:17:41 Timnit's conclusion https://t.co/HO2cmw9qe7

2023-04-09 00:16:33 Timnit was an engineer before her PhD in AI, so she brings an interesting engineering mindset to the question of AGI safety https://t.co/0vVNGhEmUg

2023-04-09 00:11:59 This is a massive challenge https://t.co/V623tFzKZb

2023-04-09 00:09:20 Timnit pointing out here that currently the benefits of AI are not "for all" https://t.co/tZjjzoebSA

2023-04-09 00:06:25 "they've been basically creating larger and larger models, advertised as being able to do everything for everyone" https://t.co/h9gBuTIkII

2023-04-09 00:01:30 "why did I go through this whole entire thing talking about the history?... that's because of the influence - that's really what I want to get to - ...there are lots of billionaires in the movement, many who are either in it, are adjacent to it, and many directly funding it" https://t.co/jHaYHxDOBM

2023-04-08 23:58:18 @paul_rietschka It's very clearly Twitter dunking ("the myopia of far, far too many in ML"), and not at all a generous or thoughtful interpretation of what you're dunking on.

2023-04-08 23:56:38 Ugh https://t.co/4Xp9Qh6YzD

2023-04-08 23:48:38 Doomism https://t.co/bPmVQAliaO

2023-04-08 23:47:28 "I see some people laughing in the audience... um well I had to take this seriously so you gotta listen to it with me okay?!" :D https://t.co/c4JNFbujDk

2023-04-08 23:44:33 ...and this https://t.co/2knBecqG4N

2023-04-08 23:43:09 Yeah gotta admit they kinda lost me when they went down this path... https://t.co/53p6Y7Z10q

2023-04-08 23:42:14 "Effective Altruism" https://t.co/eqMv7QjpPm

2023-04-08 23:41:19 "Extropeanism, singularitarianism, cosmism - they're all transhumanists, right. Rationalists are not necessarily transhumanist, but they were founded by transhumanists, and also a singulatarian".

2023-04-08 23:38:37 "Cosmists" (AFAICT a subset of transhumanists?) https://t.co/bJIrfxaaIu

2023-04-08 23:34:45 ...and "Modern transhumanism" https://t.co/T7FDHTaQ3R

2023-04-08 23:33:27 "Transhumanism" https://t.co/GAHCgwhEGy

2023-04-08 18:50:39 RT @jimmie_munyi: Going through the NLP with Transformers book by @_lewtun and decided to test chatGPT with the tasks of chapter 1 instead…

2023-04-08 18:38:06 RT @michael_nielsen: Interesting, one personal impact of gpt4: https://t.co/35CcyLmjhL

2023-04-08 18:36:12 @baturinsky @ylecun No. China and US are world leaders in AI research and development but are on opposite ends of this chart

2023-04-08 18:29:22 RT @BenBlaiszik: We wrapped up the first LLM hackathon for applications in materials and chemistry last week. The results to me were astoun…

2023-04-08 18:28:15 RT @capetorch: Due to popular requests, I updated my "Next Frame Prediction using Diffusion Models" article and added a Colab that you can…

2023-04-08 18:27:26 RT @BlancheMinerva: @reconfigurthing @QuintinPope5 @norabelrose So books + academic papers + US court cases from the past 10 years is appro…

2023-04-08 09:41:51 RT @francoisfleuret: Totally agree. Large ml models will be key elements of both GDPs and security of nations, and as such will be strateg…

2023-04-08 05:27:15 @geonumist @4evaBehindSOTA @YiTayML No it's really not like that at all.

2023-04-08 04:57:38 RT @kulesatony: Wissenschaftsfortschrittsnormalitätsleugnungswirrwarr The feeling that the single greatest scientific discovery in human h…

2023-04-08 01:40:51 RT @EMostaque: @InvestingCyborg https://t.co/3W67HP04Cj is by far the best

2023-04-08 01:40:24 @primalpoly @JeffLadish Ditto. Turns out when the greatest change in human capability comes knocking, academics -- those folks who work at institutions dedicated to human flourishing -- have responded by worrying about the impact on assessments.

2023-04-08 01:38:13 RT @JeffLadish: GPT-4 disrupts the entire set of educational infrastructure that depends on students answering questions or writing answers…

2023-04-08 01:35:06 @GaryMarcus I'm not prominent, and it wasn't lately, but in my 2014 talk on https://t.co/zUWHQd6fWR my key conclusion was that negative income tax followed by UBI was what we needed to work towards as AI capabilities increase. https://t.co/ZWkF6B8vpO

2023-04-07 22:34:46 @ylecun Wow you really did! - Nice pic :D

2023-04-07 22:16:13 @ylecun I bet they had a good laugh about that...

2023-04-07 22:10:53 RT @jnack: Sigh... my @GoogleAI team built something like this five years ago, but I could never get @GooglePhotos to put it to use. Someda…

2023-04-07 21:25:36 RT @borisdayma: "A Recipe for Training Large Models" Report: https://t.co/h6H187P4dd I've been working for a while on this guide, shar…

2023-04-07 21:17:58 RT @ChrisJBakke: The Forbes 30 Under 30 have collectively raised $5.3B in funding. The Forbes 30 Under 30 have also been arrested for fra…

2023-04-07 21:10:55 @loretoparisi https://t.co/ewQf4Vxyvm

2023-04-07 21:08:55 @ylecun Hey keep ya hands off our Aussie birds! If our cockatoos don't like you then that's on you! :p

2023-04-07 18:37:07 @goodside @amasad @stephen_wolfram Sadly chat UX at the moment is more like a line oriented REPL, which is a UX hangover from the days of teletype.

2023-04-07 18:35:13 @thechrisperry @amasad @stephen_wolfram Hey wow cool url! How long has that been hiding there for?

2023-04-07 18:21:19 @KennethCassel @amasad @stephen_wolfram Yes exactly

2023-04-07 18:20:08 @amasad @stephen_wolfram Yup I like 'em https://t.co/EiTpj3J6ny

2023-04-07 17:35:25 RT @amasad: Talked to @jeremyphoward and @stephen_wolfram this week and now I’m notebook-pilled.

2023-04-07 06:10:57 RT @francoisfleuret: I implore all the psychologists, sociologists, philosophers, to move beyond "this AI thing is a cheap trick" and get t…

2023-04-07 03:08:25 RT @andy_matuschak: I'm not sure if I'm part of the intended audience for Anthropic's safety statement, but unfortunately it left me more u…

2023-04-07 03:03:55 RT @bhutanisanyam1: A hands on guide to train LLaMA with RLHF It’s one of the most complete tutorials on the topic with detailed explan…

2023-04-06 21:34:42 RT @ggerganov: RWKV port in ggml by the community: https://t.co/H3TDDAPdVN I haven't had the chance to look at this in details yet, but i…

2023-04-06 21:34:26 RT @michael_nielsen: Radiation-hardened quines: a program such that you can delete a single character and it still outputs itself I wonder…

2023-04-06 20:34:53 @etzioni And there are certainly examples of people who assume that those with power and resources must have earned them, those that don't must be lazy and stupid, and that the status quo is by definition fair and optimal.

2023-04-06 20:33:51 @etzioni Both claims are reductive and cloud the real issues. There are certainly examples of people who are too quick to support marginalised groups without considering whether they're worthy of support.

2023-04-06 20:26:22 @etzioni Thanks Oren. Life is complicated and IMO it's always worth being wary of those who claim otherwise.

2023-04-06 20:23:16 @etzioni This kind of trite straw-manning dumbs down discourse and is intellectually lazy. One could claim conversely claim that "anti-wokeness" is synonymous with "populism" -- it appeals to the emotions of ordinary people by contrasting them with a corrupt and elitist establishment

2023-04-06 20:21:16 @etzioni Intersectionality is not an academic term for "wokeness." t is a theoretical framework to understand how different forms of social inequality and discrimination intersect and overlap.

2023-04-06 18:32:16 Wow big call from @ylecun. I wonder who or what is stopping this from happening. https://t.co/kFRxHFkXxa

2023-04-06 09:26:24 @johnowhitaker @StabilityAI Here's hoping people are going to finally understand the point of our ulmfit paper - which is the process you describe!

2023-04-06 09:25:24 Very insightful thread from @johnowhitaker which explains how to get the most out of models like Stable Diffusion https://t.co/zzcWDTQg1o

2023-04-06 07:28:14 @bookwormengr @iScienceLuvr Yup that's all there

2023-04-06 07:27:18 @Asiful_Nobel Yes!

2023-04-06 07:26:47 @bilelmoulahi Yup

2023-04-06 06:32:46 RT @alexisgallagher: On topics of your choice, do you find GPT4 more interesting to talk to than a person picked at random from the folks y…

2023-04-06 04:23:47 @bhutanisanyam1 @LangChainAI @h2oai Great stuff!

2023-04-06 03:12:07 @bhutanisanyam1 @LangChainAI @h2oai Where can I play with it?...

2023-04-06 00:54:40 RT @_jasonwei: Really surprised to see that Eric Schmidt believes in task-specific models. While I agree that specialized models are good w…

2023-04-05 23:52:16 RT @HamelHusain: New @quarto_pub Jupyter extension: renders previews of elements like Callouts, Divs, Mermaid charts and Quarto elements li…

2023-04-05 23:49:56 @SenatorRennick @simonahac Nah sorry mate you got that badly wrong. The entropy of a *closed* system will always increase. Earth is not a closed system. For instance, there's a thing called the sun.

2023-04-05 21:02:33 @wightmanr @huggingface I know the feeling - but this is definitely not a little quaint!

2023-04-05 21:01:49 RT @wightmanr: Seems a little quaint with the AI happenings these days, but finally pushed all, yes ALL the timm ResNets to the @huggingfac…

2023-04-05 20:48:50 @drjohnflackett @raphaelmilliere @criticalneuro Thanks for the tip - I'll take a look

2023-04-05 19:46:46 RT @OfficialLoganK: To deliver on @OpenAI's mission (AGI for humanity), we need to: - Building increasingly safe AI systems - Improve syst…

2023-04-05 19:41:07 RT @goodside: ai anxiety vibe check

2023-04-05 19:38:05 RT @JustineTunney: I just wrote a blog post on how we used mmap() to make LLaMA load faster, run in parallel, and use less memory. In this…

2023-04-05 19:27:52 RT @MetaAI: Today we're releasing the Segment Anything Model (SAM) — a step toward the first foundation model for image segmentation. SAM…

2023-04-05 19:24:24 RT @HamelHusain: This is the first time I’ve seen code around self attention expressed so succinctly: I can finally fit it in my head

2023-04-05 09:15:29 RT @MParakhin: Starting to ship Prompt v98 today: it is a two-stage process, by tomorrow you should see big reduction in the number of case…

2023-04-05 06:22:38 @andrewm4894 Yeah you do that!

2023-04-05 06:08:50 Uh oh... It's always both pleasing yet also mildly terrifying when HN picks up your project. :O https://t.co/gSKs1Dl8z1

2023-04-05 05:58:46 @fringetracker Thanks @fringetracker!

2023-04-05 05:58:28 RT @fringetracker: This course has a really good ground up construction of foundations of deep learning, from matrix multiplications to aut…

2023-04-05 05:57:33 @geonumist @4evaBehindSOTA @YiTayML Sure, but *you* didn't sign those ToS, did you? They say that the folks that used the service (the Alpaca team in this case) aren't allowed to use its output in that way. They don't and can't bind you AFAICT.

2023-04-05 05:55:16 @NivalBmi Stick with it and you'll get there!

2023-04-05 02:01:28 RT @iScienceLuvr: Happy to share the https://t.co/JWOQGkTeIJ Part 2 course is now released! Wow what a long and fun journey it's been! I…

2023-04-05 00:56:41 RT @untitled01ipynb: Ran out of memes for this week, here's next week's memes https://t.co/DnGDvzSc2N

2023-04-05 00:37:23 @matdmiller My great pleasure!

2023-04-05 00:30:02 You'll need to be pretty comfortable using Python, and training PyTorch models to start this course. If you're not, I'd recommend first completing the 8 lessons of part 1 of Practical Deep Learning for Coders https://t.co/nzv7pek0iq

2023-04-05 00:30:01 Because *you* will create miniai yourself during the course, that means you can customise it to create your own version. For instance, @johnowhitaker has created his version called "miniminiai"!! https://t.co/DhHCZShWEx

2023-04-05 00:30:00 After we create something from scratch, we then learn about libraries which can implement that feature for us -- such as @huggingface Diffusers, Datasets, and timm, @PyTorch, torchvision, and torcheval, and @RiversHaveWings' k-diffusion.

2023-04-05 00:29:59 We'll see how to generate random numbers (a critical piece of diffusion models), and learn about the dangers of mixing random number generation with multi-process code https://t.co/HDSTPf18jI

2023-04-05 00:29:58 Throughout the rest of the course, you'll then learn all the math and coding bits needed to rebuild this from scratch -- i.e starting with nothing but Python and its standard library. Step one is a matrix class that lets us access pixels... https://t.co/niaEfNSkwb

2023-04-05 00:29:57 After the first lesson, you'll be able use the @huggingface Diffusers library to generate images, and will learn various tricks to customise the image generation process. You'll also learn the foundational mathematical principles used in diffusion models more generally. https://t.co/SEN5o8SsBS

2023-04-05 00:29:56 This field was developing rapidly as we were developing and teaching the course, so many lessons include a walk-through of a paper that had just been released. We also implement key papers that aren't in Stable Diffusion, such as Karras et al (2022) https://t.co/c4oOA5O23S

2023-04-05 00:29:55 Our new course, "From Deep Learning Foundations to Stable Diffusion", is finally done after 8 months of work!!! With >

2023-04-04 22:48:01 RT @simonw: "Hallucitation" is a most excellent neologism for hallucinated citations produced by LLMs such as ChatGPT

2023-04-04 21:21:27 @simonw @tojulius @sethmlarson Thanks for the link. My prompts were a bit more detailed than that. E.g I said in the prompt it's just a snippet of a longer transcript, and said if it's the first, middle, or last section

2023-04-04 21:16:02 @simonw @tojulius @sethmlarson That's how I summarised my YouTube transcripts (although this is the first time I've heard it's a known trick!) It worked fine for me.

2023-04-04 20:51:41 RT @SashaMTL: My editorial about the (infamous) open letter is live on WIRED! https://t.co/VH2Z9ISEqE I did my best to shift the narrative…

2023-04-04 19:43:57 RT @michael_nielsen: Many AGI doomers &

2023-04-04 16:40:02 RT @nsaphra: Just shut up about the 2022 AI risks survey. The total response rate including INCOMPLETE responses was 17%. The xrisk questio…

2023-04-03 22:33:24 RT @mxcl: My manager at Apple told me single letter variables are never ok. I was using `i` as the index in a two line loop. I handed in my…

2023-04-03 21:53:07 @akatzzzzz @mathemagic1an @yoheinakajima I'd guess the extra flexibility of code libs will remain useful at least for a while

2023-04-03 21:49:48 RT @haozhangml: We Just released the delta weights of our Vicuna 13B. Try on your own GPU with this chatbot that is closest to chatgpt in o…

2023-04-03 17:51:24 RT @johnrobinsn: GPTQ lands support for 4-quantization of T5 models (encoder/decoder architecture)... should work with UL2 as well. https:…

2023-04-03 09:45:27 @GaryMarcus Nah

2023-04-03 09:44:17 @johnowhitaker @weights_biases @untitled01ipynb Few companies can afford a cat of this quality

2023-04-03 09:42:12 @ParTmeAdvnturer Yup it's an option

2023-04-03 09:41:54 @lxcian98 That's called swapping. It's very slow.

2023-04-03 09:40:43 @michaelgokhman That's not exactly true either. The OS can swap in and out even smaller sections than a layer. But the point is you don't want to be swapping. FSDP is different because it's RAM-backed

2023-04-03 08:34:59 @ylecun It's the smart move for them IMO. Advocating for a regulatory process when there's regulatory capture is a great way to increase the chance that status pro power structures are maintained.

2023-04-03 07:54:51 RT @DanielleFong: definitely think that the nVidia GPUs accelerating computational lithography for photolithography masks (for more chip cr…

2023-04-03 07:54:19 RT @GaryMarcus: Does any professional in AI and biology find this prediction remotely plausible?

2023-04-03 07:31:18 @bradneuberg My pleasure

2023-04-03 05:10:16 RT @AllenDowney: Today I asked ChatGPT to solve almost every exercise in Think Python and DSIRP. It did. My conclusion: everyone who wri…

2023-04-03 04:45:08 @simonw @MitchellsMovie But it's so cute!!!

2023-04-03 04:32:39 Heh turns out @amasad has already weighed in with his thoughts! :D https://t.co/MzSmOiiGZz

2023-04-03 03:55:02 RT @theshawwn: This feels prescient. All I could think while watching it is “Someone will probably train a GPT robot to behave this way wit…

2023-04-03 03:48:18 Positional encoding is used to assign a unique vector to each position in the input token vector, so that the model can learn that the location of a words matters -- not just its presence/absence. (They have few parameters do don't impact memory significantly.)

2023-04-03 03:48:17 Oops another correction! I forgot to mention positional encoding. This is the layer after the embeddings. It can be implemented using sin/cos across a few freqs ("sinusoidal embeddings") or the code snippet below ("learnable embeddings"). https://t.co/NYjx1K7Yo4 https://t.co/TYZsaZ90fk

2023-04-03 03:44:13 @igorcosta Oops yes you're right! (Although I should note it is possible for models to learn about position even *without* positional encoding!)

2023-04-03 03:43:30 Having said that, the embedding weights are a fairly small proportion of the model parameters, so this won't generally make a big difference either way.

2023-04-03 03:43:29 D'oh I meant to mention this and forgot! This is a great point. The first layer of a language model is an "embedding" layer. Mathematically, it's matrix multiply by a bunch of one-hot encoded vectors. But it's implemented as an array index lookup. https://t.co/go2QiXt48v

2023-04-03 03:36:18 @apage43 Yes very true!

2023-04-03 03:31:33 @wightmanr Good point!

2023-04-03 03:18:48 @rachelmetz @timnitGebru The focus was on understanding exponentials. I thought (and still think) that's a really good thing to focus on.

2023-04-03 03:17:33 Minor correction: the normalization layer I used here, Batchnorm2D, isn't what's used in a normal language model. I took the code from an image model we built for the course. In most language models we'd use Layernorm for normalization. https://t.co/vro2pwUaxJ

2023-04-03 03:14:04 @rachelmetz @timnitGebru Funnily enough in my time teaching at SU I don't recall us ever having a discussion about the singularity!

2023-04-03 03:12:48 h/t @iScienceLuvr for encouraging me to write this thread.

2023-04-03 03:12:05 Note however that none of this helps us use the GPU better -- it's just for CPU use (or systems with integrated memory like Apple silicon). By far the best speed is with an NVIDIA GPU, and that still requires copying all the params over to the card before using the model.

2023-04-03 03:12:04 The way mmap handles memory makes it a bit tricky for the OS to report on a per-process level how that memory is being used. So it generally will just show it as "cache" use. That's why (AFAICT) there was some initial misunderstanding regarding the memory savings

2023-04-03 03:12:03 Therefore, we don't really get to benefit fully from the magic of mmap, since we need all those parameters for every batch of data. It's still going to save some unnecessary memory copying so it's probably a useful thing to use here, but we still need RAM for all the parameters

2023-04-03 03:12:02 So every parameter is used, because they're just matrix multiplications and element-wise operations that they're used in. There's no true sparsity occurring here -- there's just some "soft attention" caused by the use of softmax.

2023-04-03 03:12:01 OK that was a slight simplification. Transformers actually use "multi-head" attention which groups channels first. But as you see from the implementation below, it's still just using some matrix multiplies and activation functions. https://t.co/EsieHGfmDH

2023-04-01 21:34:56 RT @ylecun: AI doomism is quickly becoming indistinguishable from an apocalyptic religion. Complete with prophecies of imminent fire and b…

2023-04-01 20:03:56 RT @HamelHusain: Much of the power of #nbdev re: avoiding merge conflicts were previously not available to VSCode users. New vscode f…

2023-04-01 10:07:02 RT @esjesjesj: Remember when Elon Musk publicly discriminated against a disabled worker and then had to apologize when it turned out firing…

2023-04-01 07:27:17 RT @jachiam0: Can we please excise this terrifying individual from serious consideration in AI safety? This is not a field leader, this is…

2023-04-01 07:25:35 @weights_biases @untitled01ipynb You chose well.

2023-04-01 06:46:39 RT @HamelHusain: New #nbdev vscode extension feature Autocomplete directives, with hints! https://t.co/ICJk7nQloa https://t.co/9ti5…

2023-04-01 05:47:29 @DFisman @zeynep Hey that's me! I didn't see that article before

2023-04-01 05:43:38 @wasimlorgat @HamelHusain @TheCodingProjec And it would largely avoid polluting diffs

2023-04-01 05:43:17 @wasimlorgat @HamelHusain @TheCodingProjec Removing the cell numbers but leaving the comments there would allow most things to still work most of the time

2023-04-01 03:16:08 @untitled01ipynb @mervenoyann Couldn't say https://t.co/BQtau2AIYb

2023-04-01 03:04:39 RT @lacker: I'm excited to finally see the secret sauce behind Twitter's machine learning algorithms https://t.co/SSJne3Ut4x

2023-04-01 02:57:11 RT @fchollet: So, Twitter open-sourced The Algorithm... I was so close (at least for the toxicity/abuse/nsfw detection models) https://t.c…

2023-04-01 02:55:54 @slhamlet @Grady_Booch Vanishingly few philosophers today consider this thought experiment compelling, FYI.

2023-04-01 01:10:18 RT @agikoala: Three starter pokemon in deep learning https://t.co/u6RZqrvsAJ

2023-04-01 00:33:53 RT @wongmjane: Twitter’s algorithm specifically labels whether the Tweet author is Elon Musk “author_is_elon” besides the Democrat, Repub…

2023-04-01 00:32:20 RT @timnitGebru: Since we've been looking for more things to do, @emilymbender @mmitchell_ai @mcmillan_majora and I wrote a statement about…

2023-03-31 23:52:13 RT @FelipeKitamura: Come show your skills by participating in these @kaggle competitions. Help predict the age and gender of patients based…

2023-03-31 23:51:51 RT @GoogleAI: Learn about ViT-22B, the result of our latest work on scaling vision transformers to create the largest dense vision model. W…

2023-03-31 22:44:33 RT @rharang: this is why we can't have nice things. A langchain LLM agent for solving math problems just yeets any python code you give it…

2023-03-31 19:06:43 RT @yoavgo: the real power differential of openai vs academia (and also vs most other industry ai labs) is not the compute. its a large ded…

2023-03-31 19:06:15 @michael_nielsen My understanding from listening to chess videos is that players have learned a huge amount from studying AI games, although much is still beyond our understanding

2023-03-31 19:03:35 RT @hippopedoid: We held a reading group on Transformers (watched videos / read blog posts / studied papers by @giffmana @karpathy @ch402 @…

2023-03-31 18:59:15 RT @JMannhart: This is an absolutely incredible video. Hinton: “That's an issue, right. We have to think hard about how to control that.“…

2023-03-31 18:51:17 RT @jburnmurdoch: NEW: I’m not sure people fully appreciate how dire the US life expectancy / mortality situation has got. My column: http…

2023-03-31 18:47:43 RT @theshawwn: I used GPT-4 to draft a DMCA counterclaim for https://t.co/IIcRpWwgtM. Does anyone know any lawyers that could weigh in on w…

2023-03-31 18:44:22 RT @schock: Tech bros: AI is here and it's about to kill us all! give us money to try and stop it Also tech bros: the Algorithmic Accounta…

2023-03-31 18:36:03 RT @Soccermatics: The Future of Life Institute is a problem. Being in same age-group (lower end maybe)/cultural background (8-bit progra…

2023-03-31 18:30:50 RT @ProfNoahGian: EA: Earn as much money as you can, charity needs it! SBF: Ok EA:. Focus on AI alignment, our top priority! Yudkowsky…

2023-03-31 18:30:45 RT @bitcloud: @ProfNoahGian EA is feudalism. I understand the motives but I think it makes a fatal flaw in assuming that the people most c…

2023-03-31 18:28:57 RT @finbarrtimbers: int4 quantization is a total game changer for LLM inference it's just impossible to argue with these numbers, even if…

2023-03-31 18:26:55 @maosbot @fnedrik We also need to solve corporate alignment. Companies often do things even their own execs don't feel OK with - let alone society!

2023-03-31 11:09:17 RT @adrianzenz: Me: Is there a Uyghur genocide? Bing chat AI: Sorry, let's try a different topic. Me: Why do you refuse to talk about the…

2023-03-31 11:06:15 @untitled01ipynb I'm gonna ignore that guidance

2023-03-31 09:26:25 @untitled01ipynb Ah I see. Thank you for enlightening me

2023-03-31 09:23:10 RT @maosbot: The most pressing problem right now is not AI alignment, it is human alignment—can we coordinate on what we really want from A…

2023-03-31 08:56:07 @untitled01ipynb Sorry what DeepMind takeover? I'm OOTL again

2023-03-31 08:53:15 @amasad Yeah this is why I feel uneasy whenever people push for regulation, at least in the US.

2023-03-31 08:51:25 @rharang I missed it - what happened?

2023-03-31 08:49:29 @haozhangml @4evaBehindSOTA @lmsysorg Personally I think your approach to evaluation is very creative - I love it! None of the standard evals seem well aligned with human judgement.

2023-03-31 08:46:57 @Secheron1202 @bhutanisanyam1 @fastdotai My pleasure!

2023-03-31 08:46:28 @bhutanisanyam1 @fastdotai I can't even remember how we found good papers back then!

2023-03-31 05:57:21 RT @lmsysorg: Introducing Vicuna, an open-source chatbot impressing GPT-4! Vicuna reaches 90%* quality of ChatGPT/Bard while significant…

2023-03-31 05:45:35 RT @abacaj: Codegen 6B released a year ago is competitive to the new Cerebras GPT-13B. Consider taking a closer look at the codegen model v…

2023-03-31 05:40:39 @bhutanisanyam1 @fastdotai Also: focus on fine-tuning bigger models. Which works out pretty well for ChatGPT!

2023-03-31 05:39:57 RT @bhutanisanyam1: I’ve been thinking about this a lot @fastdotai has been truly ahead of its time Build in public Deep Learnin…

2023-03-31 02:10:16 RT @AnthropicAI: Today we are releasing the new Claude App for @SlackHQ, in beta. Now every company in the world has the chance to have a “…

2023-03-31 02:07:27 RT @arankomatsuzaki: BloombergGPT: A Large Language Model for Finance Presents BloombergGPT, a 50 billion parameter language model that is…

2023-03-30 23:29:18 @simonw The problem IIUC is that it's under-trained for its param count, so is much less good as a foundation than other models of this size. (i.e. it's "compute optimal" for training, not "params optimal")

2023-03-30 22:12:28 @alexjc @david_picard @giffmana OK - thank you for clarifying.

2023-03-30 22:12:05 @mijustin @NesslerBernhard @ylecun Your understanding is incorrect (and ChatGPT's is, at best, misleading). We've taught hundreds of thousands of students to become deep learning practitioners, with a minimum prerequisite for starting the course of high school math. Many of our alums are now working in the field.

2023-03-30 22:09:13 RT @ylecun: The year is 1440 and the Catholic Church has called for a 6 months moratorium on the use of the printing press and the movable…

2023-03-30 22:08:23 RT @timnitGebru: The very first citation in this stupid letter is to our #StochasticParrots Paper, "AI systems with human-competitive int…

2023-03-30 22:06:07 @alexjc @david_picard @giffmana I'm interested in learning more about the GDPR, and what it does and doesn't cover. I don't think it's kind, fair, or reasonable to assume intent when someone is asking a question.

2023-03-30 22:04:27 @alexjc @david_picard @giffmana No I'm not looking for exceptions that covers LAION's use of PII.

2023-03-30 21:36:34 @alexjc @david_picard @giffmana I don't see anything in this discussion suggesting that anyone has a goal to try to legitimize any particular dataset?

2023-03-30 21:26:12 @david_picard @giffmana These exceptions seem like they might be valid in this case (apologies in advance if I'm way off - I'm no GPDR expect)? https://t.co/iTILCKTkaj

2023-03-30 21:18:20 RT @HamelHusain: Working on a VSCode Extension for #nbdev. First feature ->

2023-03-30 21:15:18 RT @laion_ai: We launche a petition to democratize AI research by establishing an international, publicly funded supercomputing facility e…

2023-03-30 20:16:45 RT @giffmana: omg I love this timeline where manipulating LLM-chatbots is now a thing.

2023-03-30 19:59:49 RT @arankomatsuzaki: RWKV (R) (seqlen = 4k) vs. Pythia (GPT3) (T) (seqlen = 2k). RWKV is unable to utilize its full context length. My cu…

2023-03-30 06:17:24 RT @andriy_mulyar: I'm excited to announce the release of GPT4All, a 7B param language model finetuned from a curated set of 400k GPT-Turbo…

2023-03-30 06:13:47 RT @johnowhitaker: Hooray - #playgroundv1 is launching today! Here's my favourite set of test images, which I made as we finalized the mode…

2023-03-30 06:06:21 RT @wasimlorgat: This has been my experience! My favourite and most frequently used prompts are follow-ups like: “Why X instead of Y?” “W…

2023-03-30 06:04:19 RT @YiTayML: How about pausing all the LLM influencer nonsense on twitter for 6 months! That, I would sign.

2023-03-30 05:35:39 RT @AndrewYNg: 1/The call for a 6 month moratorium on making AI progress beyond GPT-4 is a terrible idea. I'm seeing many new applications…

2023-03-30 00:52:45 RT @ProfNoahGian: In addition to pausing AI for 6 months, I propose pausing carbon emissions for 6 months so we can solve climate change du…

2023-03-30 00:24:25 RT @RisingSayak: If only Twitter was used more for productive conversation exchanges like this these days. @YiTayML literally threw paper…

2023-03-29 23:08:02 @Vernboy @saaspegasus @Replit @amasad If I have anything useful to say, I will do so!

2023-03-29 06:05:29 @Vernboy @saaspegasus @Replit Just a little, but I'm keen to do more, because I'm pretty excited about Ghostwriter and the direction they're taking things. There's clearly a lot of overlap between nbdev and replit and I'm not sure how best to think about that. I'm chatting to @amasad soon-ish as it happens

2023-03-29 05:28:37 @Grady_Booch Of course they are capable of that. That's shown by a trivial application of the universal approximation theorem. They are not limited by their architecture.

2023-03-29 05:18:08 @saaspegasus @Vernboy @Replit I say this as a subscriber to Pegasus BTW

2023-03-29 05:17:45 @saaspegasus @Vernboy @Replit I think the established productivity of folks using nbdev has already proven you wrong! :D

2023-03-29 05:15:21 RT @ylecun: Nope. I did not sign this letter. I disagree with its premise.

2023-03-29 05:14:54 RT @LakeBrenden: Which is more incorrect, the original tweet or Twitter's added context? https://t.co/JbixAi0ETP

2023-03-29 03:38:12 @waydegilliam @fastdotai @huggingface Yes! - Talk to @johnrobinsn .

2023-03-29 03:28:36 RT @nomic_ai: Today we're releasing GPT4All, an assistant-style chatbot distilled from 430k GPT-3.5-Turbo outputs that you can run on your…

2023-03-29 00:00:05 RT @VisualVichaar: Got access to ChatGPT browsing mode today Following is a thread of some of my early experiments with it #AI, #ChatGPT…

2023-03-28 21:54:43 RT @Replit: We're teaming up with @googlecloud. Replit's 20M+ developers will get Google Cloud services, infrastructure, and foundation m…

2023-03-28 21:53:00 RT @simonw: If the only thing I could use ChatGPT for was pasting in error messages to get explanations like this one I would still pay $20…

2023-03-28 20:35:29 fastai is now compatible with PyTorch 2.0, and its compiler, for ~50% speed improvements! https://t.co/B6zoSG80g5

2023-03-28 20:32:57 "Optimal constrained by memory use" seems like the thing that everyone should be striving for, but very few are. https://t.co/329bVHPmoN

2023-03-28 04:26:04 RT @sergeykarayev: I want chat with AI about long-form content I'm reading. (It's a paper on Arxiv, but the solution would ideally support…

2023-03-27 21:05:53 RT @simonw: "I think it’s likely that soon all computer users will have the ability to develop small software tools from scratch, and to de…

2023-03-27 21:02:14 RT @fhuszar: Now that everyone is fatigued by GPT-4 hot takes and blocked the keyword "LLM", here's the blog post with my current view on t…

2023-03-27 06:26:04 @averma12 @radekosmulski No that wasn't similar at all.

2023-03-27 06:25:34 RT @radekosmulski: Even in the era of GPT-3, I didn't realize something about AI models that @jeremyphoward knew at least 8 years ago. I d…

2023-03-27 03:29:48 RT @charles0neill: 1/ I came across a paper shared by @jeremyphoward about parallelisable linear RNNs. The ICLR 2018 paper (https://t.co/…

2023-03-27 03:28:04 RT @erikbryn: ChatGPT knows me all too well. I love re-reading that book. https://t.co/5wwkxg4fE4

2023-03-27 00:37:40 RT @johnrobinsn: Alpaca Finetuning of Llama on a 24G Consumer GPU Train Your Own "ChatGPT-clone" On Your Own Data and On Your Own Hardwa…

2023-03-26 19:50:55 RT @ClementDelangue: So grateful for a year and a half of collaborating with @mmitchell_ai! I'm still learning every day &

2023-03-26 10:49:36 RT @johnowhitaker: Simo just keeps churning out excellent projects! Amazing talent Thank him for his OS contributions if you can ( https:/…

2023-03-26 08:04:54 RT @SonglinYang4: S4D, S5, RWKV &

2023-03-26 08:01:45 RT @peakcooper: #GPT4 saved my dog's life. After my dog got diagnosed with a tick-borne disease, the vet started her on the proper treatm…

2023-03-26 07:56:31 @radekosmulski @simonw @ozboware I think Simon's advice is still the best approach for now, since I can't think of anything better. But assuming this time is just like all the other times, might lead to hubris.

2023-03-26 07:54:28 @radekosmulski @simonw @ozboware Yeah but this time it's actually different. If you respond to it based on other recent technology impacts you're not going to respond appropriately.

2023-03-26 07:44:17 RT @DrJimFan: https://t.co/cQTkqWPC5c Paves the way to drop “Open” eventually?

2023-03-26 04:48:21 RT @laks_srini: Want to reimagine how a home for sale might look like when you make it your own? We just launched a new feature that does j…

2023-03-26 04:43:25 @marcus_marcusg When I was at university it was certainly covered in our cognitive science course - that was 30 years ago though...

2023-03-26 03:56:52 I've always been rather confused about claims that embodiment or grounding is needed for intelligence, but never had much luck getting good engagement on this topic. For the first time, I managed to - by talking to Bing! It even came up with a novel take. https://t.co/XswQnX5zvZ

2023-03-26 02:51:02 @EnricoShippole @YiTayML @huggingface Great! Please follow or DM me so we can discuss.

2023-03-26 02:49:34 RT @radekosmulski: To learn to code -- read and write a lot of code. To learn to write -- read and write a lot of prose. To learn to use…

2023-03-26 02:36:37 @EnricoShippole @YiTayML @huggingface I'd love to hear more about this project!

2023-03-26 01:59:44 RT @ggerganov: @olcan @jobergum FLAN-T5 was actually higher on my todo list. There is even a WIP branch sitting there since a few months B…

2023-03-26 01:58:43 @krymski Nope.

2023-03-26 01:57:06 @LinkBechtel My intuition is that it'll probably scale well, but only empirical testing will be enough to know

2023-03-25 23:43:15 RT @natfriedman: Five months later https://t.co/OtdxJW2VXR

2023-03-25 23:37:55 RT @ElunaAI: Ye taking over Bieber's vocals on Love Yourself. https://t.co/JpA2vVAoH2

2023-03-25 22:33:59 @WillingCarol Greater transparency would, indeed, be terrific. The models are still useful without transparency around dataset creation, and it is possible to develop trust in them even in this case -- but transparency around dataset creation would still be much appreciated!

2023-03-25 22:32:34 @vagabondjack @4evaBehindSOTA @YiTayML Mike apologies if you already covered this elsewhere, but did you try the cleaned-up version of the Alpaca dataset? (i.e. the original has ~80% of math problems with wrong answers)

2023-03-25 22:31:34 @giffmana @pastaraspberry @YiTayML @sbmaruf I see the analogy, but I don't see how that would relate to the legal standard here. The text of the license says "that compete with", not "that *could be used to* compete with". It also doesn't refer to derivative works that compete with.

2023-03-25 18:34:48 RT @simonw: If you're just starting to learn software engineering right now but you're considering dropping it because you think the field…

2023-03-25 18:29:36 RT @HamelHusain: I just posted a $100 replit bounty for a simple VSCode Extension for navigating to/from Jupyter Notebooks in #nbdev pro…

2023-03-25 18:24:56 RT @bilawalsidhu: Midjourney v5 has pushed into photorealism, a goal which has eluded the computer graphics industry for decades (!) Ins…

2023-03-25 18:18:57 @WillingCarol It can also be useful to see the activations and gradients of course, which sadly isn't possible for closed source models

2023-03-25 18:18:23 @WillingCarol But the good news is we have good (and improving) ways to understand large models. Generally, it's through approaches to how we understand humans -- by asking them questions. And we can ask them a *lot* of questions!

2023-03-25 18:17:06 @WillingCarol You can't meaningfully inspect or understand such a dataset. But trying to understand a large model by understanding its dataset is as fruitless as trying to understand its weights.

2023-03-25 18:15:49 @WillingCarol I've seen this concern mentioned a few times, but I don't see it. Everyone used the same approach of filtering common crawl (or their internal equivalent) to find prose and code documents and add some light filtering

2023-03-25 18:10:37 @pastaraspberry @giffmana @YiTayML @sbmaruf Yeah the model clearly isn't competing with meta. They didn't even deploy it.

2023-03-25 18:09:15 @jjvincent Not too harsh at all. They deserve to be called out for releasing a subpar product when they could and should be doing better

2023-03-25 10:41:17 @francoisfleuret @mmbronstein OTOH, I find transformers for language work really well for most things without much mucking around (except for long documents, where ULMFiT often works better)

2023-03-25 10:39:43 @francoisfleuret @mmbronstein Nonetheless, I find working with transformers for vision to be, on the whole, complicated and frustrating in practice, unless I'm fine-tuning a big model. Which might just mean I'm crap at it. But it does argue against the simplicity claim.

2023-03-25 10:38:37 @francoisfleuret @mmbronstein I was responding to your "the simplicity of ViT". But yes, there are now options that work on smaller datasets.

2023-03-25 10:22:28 RT @untitled01ipynb: New template dropped @lunarflu1 https://t.co/ZHRpGJx3nv

2023-03-25 10:20:54 @francoisfleuret @mmbronstein Not when you see how much data ViT needed to be any good.

2023-03-25 10:18:55 @YiTayML @sbmaruf But law is very interesting! Law is code for the functioning of society!

2023-03-25 10:17:23 @francoisfleuret Very nice!

2023-03-25 10:16:07 @sbmaruf @YiTayML It does allow it https://t.co/cYzEKy9gYI

2023-03-25 10:13:19 @YiTayML @sbmaruf They've changed it ("temporarily") to a non-commercial license. But if you download the version from before that change, then you can AFAICT still use it under the previous license, because it states that the license can only be terminated for breach. https://t.co/jf6Ksmdd8i

2023-03-25 10:09:49 @YiTayML @sbmaruf Therefore, the JSON file containing the instruction tuning data belongs to Stanford, and you're welcome to use it in whatever way Stanford says you can use it for. Until last week, that was basically anything, as long as you provided attribution https://t.co/Ab4HKS888k

2023-03-25 10:07:35 @YiTayML @sbmaruf Stanford did not use the output to create a competing service, therefore they followed the terms, therefore they own their output https://t.co/8Mv7HXdTLP

2023-03-25 10:06:35 @YiTayML @sbmaruf You own the outputs, as long as you followed the ToS https://t.co/YG3VrQvohx

2023-03-25 10:04:29 @jjvincent Oooo spicy! :D https://t.co/DIVfxexXlL

2023-03-25 09:42:54 @untitled01ipynb @iScienceLuvr And anyway, who bans a cat???

2023-03-25 09:42:36 @untitled01ipynb @iScienceLuvr Well I still wanna know which meme they found unsatisfying. No appreciate for art, some people...

2023-03-25 09:41:56 @giffmana @rharang This is the underlying reason I created ULMFiT in the first place -- because language models will naturally need to learn this stuff. And you can see it's true because GPT-4 is a good chess player, can solve new wordle puzzles, etc. It's not just repeating stuff it's seen before

2023-03-25 09:40:37 @giffmana @rharang Furthermore, although the task of predicting missing tokens sounds kinda basic, it turns out that the best way to be really good at it, given their limited parameter count, is for the models to learn quite a bit of logic, analogy-making, planning, etc.

2023-03-25 09:35:05 RT @vykthur: This is concerning. I worry about the 90% of the population (my parents included) that can't fathom that a machine is capable…

2023-03-25 09:31:20 @iScienceLuvr @untitled01ipynb I'm confused why anyone would block either of you two?!?

2023-03-25 09:23:01 RT @davidad: OpenAI: It’s important for safety that AI-generated code doesn’t have direct real-world effects. So we disabled Internet acces…

2023-03-25 09:21:48 @YiTayML @4evaBehindSOTA Yeah I'm just referring to the dataset (and finetuning flan with it) -- not to using it with the llama model, which perhaps needs a different license (although even that seems a bit unclear frankly after looking into it)

2023-03-25 09:19:00 RT @YiTayML: @samlakig @jeremyphoward the interesting thing is to bridge the gap w/o api distillation which is something a top-tier institu…

2023-03-25 09:18:37 RT @YiTayML: @JagersbergKnut @jeremyphoward the way we see it, flan-t5 and flan-ul2 almost belongs to the same family. at small-base sizes,…

2023-03-25 07:31:11 There's a lot of excitement about llama, which is non-commercial research licensed and needs separate instruction tuning. Why is there so little activity around flan-ul2 20B, which is tuned and openly licensed? https://t.co/jpVExCCuGh

2023-03-25 07:21:45 RT @naglinagli: The team at @OpenAI just fixed a critical account takeover vulnerability I reported few hours ago affecting #ChatGPT. It…

2023-03-25 07:15:36 RT @pitdesi: Whoa: @elonmusk tried to take over OpenAI in 2018, saying they had fallen behind Google. When rejected, he reneged on funding…

2023-03-25 03:48:50 RT @KyleRayKelley: Released v2.0.0 of genai * Automated summarization of DataFrames used in the notebook * Continue computing conver…

2023-03-25 01:06:38 RT @osanseviero: Very excited to share some personal news! @johnowhitaker @pcuenq @multimodalart and I are writing a book with @oreilly abo…

2023-03-25 00:49:34 RT @AlexBlechman: Sci-Fi Author: In my book I invented the Torment Nexus as a cautionary tale Tech Company: At long last, we have created…

2023-03-24 20:43:32 RT @iScienceLuvr: How does GPT-4 do in the medical domain? I got to play around with its multimodal capabilities on some medical images!…

2023-03-24 05:38:57 RT @nearcyan: "First Contact With an AGI System" appears as a commented-out title within the latex source code for https://t.co/FcQSYcflzE…

2023-03-24 05:38:14 @karpathy @catherineols ...and if you mentioned neural nets in 2000 the response was "didn't you hear that Vapnik proved SVMs are optimal? NNs are pointless"

2023-03-24 02:00:47 @KyleRayKelley @elliotbst @radekosmulski you should check this out if you haven't - similar to what you've been building

2023-03-24 00:07:59 RT @radekosmulski: How to use ChatGPT in Jupyter Notebook I created a cell magic that will allow you to talk to the OpenAI API directly…

2023-03-23 23:53:37 RT @stephen_wolfram: 2.5 months later...in collaboration with @openAI , ChatGPT gets its "Wolfram superpowers"! https://t.co/93hCaOTTj6 htt…

2023-03-23 23:53:25 RT @mitchellh: I've developed a lot of plugin systems, and the OpenAI ChatGPT plugin interface might be the damn craziest and most impressi…

2023-03-23 23:52:14 @radekosmulski @elliotbst @OpenAI Great stuff - I'd been thinking about doing just that. :D Please share what you've built!

2023-03-23 23:21:44 @amasad @elliotbst @Replit Sent!

2023-03-23 22:57:07 @amasad @elliotbst @Replit @amasad would love to talk to you about this if you have a moment -- might fit in with our upcoming course

2023-03-23 22:56:00 RT @radekosmulski: I switched to vscode for running Jupyter Notebooks so that I can use @github Copilot and wow (that is coming from a v…

2023-03-23 21:00:43 @elliotbst I haven't, but I agree it would be perfect.

2023-03-23 21:00:30 RT @elliotbst: Copilot and Ghostwriter are 2 examples of successful LLM integration in your IDE for better productivity. But I haven’t seen…

2023-03-23 20:03:07 Looks like it's already been done, including fine-tuning... h/t @johnrobinsn https://t.co/i1oVQOGycV

2023-03-23 19:30:21 RT @gdb: Plugins for processing a video clip, no ffmpeg wizardry required. Actual use-case from today's launch. https://t.co/Q3r2Z8fRS5

2023-03-23 17:26:21 @johnrobinsn Can't wait to hear how you go!

2023-03-23 09:20:52 This is great - anyone think they can get this to work on a 24GB card? Currently it needs 40GB. https://t.co/GSruZyh9Vn

2023-03-23 08:11:14 @Geoffrey_E_Hill @dr_bff The people on these committees you're referring to - they... can't read?

2023-03-23 06:25:44 @lmeyerov That is indeed the big question. My guess is yes, they will.

2023-03-23 04:43:01 RT @peterwildeford: Optical illusion technology is becoming a bit too powerful, pt 20 https://t.co/ARzgfTwp17

2023-03-23 03:37:08 I think it's no surprise this development came from an independent researcher (@BlinkDL_AI). There's huge pressure in the big labs to follow the current research trends. Going against the tide is really tough.

2023-03-23 03:32:44 Everybody I know in the NLP community has been watching this extraordinary project closely. Amazing achievement. RNNs are not dead! https://t.co/UfHxsHIbaJ

2023-03-23 02:13:22 RT @emollick: Using AI to prepare for meetings: "Bing, I am pasting in the code for an .ics calendar invite, can you brief me for the event…

2023-03-23 01:05:02 RT @roman01la: finally a code review that I deserve https://t.co/M4033L6aah

2023-03-23 00:41:50 RT @cwolferesearch: Even though Google is granting access to Bard (including API access to approved developers), I’m personally not excited…

2023-03-22 21:53:02 @Suhail @iScienceLuvr @johnowhitaker @fastdotai Thanks Suhail - it was kinda exhausting to get through, but fun and exciting too.

2023-03-22 21:43:09 @_Epoching_ @iScienceLuvr @johnowhitaker @fastdotai https://t.co/FRxW0G3PAl

2023-03-22 16:17:44 RT @github: GitHub Copilot is already helping developers code faster in their IDEs. But what’s next? Our answer is GitHub Copilot X. It’s…

2023-03-22 16:10:20 RT @BillGates: Everyone should benefit from artificial intelligence, and not just people in rich countries. This is the priority for my own…

2023-03-22 16:06:36 @ReubenCosmos Wow really? Tell me more!

2023-03-22 07:28:37 You heard it here first https://t.co/QExd3C2fy5

2023-03-22 07:23:53 RT @lintool: GPT-4 and its ilk are awesome for rapid prototyping and one-offs, but at the end of the day, enterprises will deploy far small…

2023-03-22 07:17:24 @christophfenton I got it to describe DFW-style, in the style of DFW, and it had some footnotes-in-footnotes, self-deprecating bits, and a conclusion that it's pointless to even try trying to replicate DFW when you're just a language model! :D

2023-03-22 07:16:33 @christophfenton Sadly I had to delete all the best DFW-style bits it wrote, since they didn't quite fit what I was trying to say. It's worth experimenting with because it can actually make some pretty amusing stuff.

2023-03-22 07:13:47 I just got off the Zoom with @iScienceLuvr and @johnowhitaker where we completed recording the final @fastdotai part 2 video -- there are 18 of them! I'm working on creating the website for the new course now and will let you all know as soon as it's ready! :D

2023-03-22 07:09:22 RT @migueldeicaza: I wrote a ChatGPT for all the platforms I use it on (macOS, iOS and watchOS) that uses my OpenAI key, without using thir…

2023-03-22 04:16:44 RT @MKBHD: I’ve been playing with Google Bard for a while today and I never thought I’d say this, but… Bing is way ahead of Google right no…

2023-03-22 03:35:54 RT @katieelink: GPT-4 Epic: coming soon to a hospital near you? https://t.co/qjxFCHIO7B

2023-03-22 01:43:32 RT @math_rachel: Sometimes I can't believe I pivoted to go back to school for immunology at a time when so much is happening in AI, but the…

2023-03-22 00:30:02 RT @vladquant: Just added Google Bard to our AI "showdown" on question-answering. Performance same as ChatGPT4, still worse than Kagi. S…

2023-03-21 22:55:30 RT @simonw: Another thing that's not at all obvious about chatbots like ChatGPT and Bard: asking them questions about themselves (how they…

2023-03-21 22:14:20 RT @nearcyan: tired of losing your bets about AI? try this out instead: 1) wait for someone to confidently say "AI will not be able to do…

2023-03-21 21:24:19 RT @David3141593: holy FUCK. Windows Snipping Tool is vulnerable to Acropalypse too. An entirely unrelated codebase. The same exploit sc…

2023-03-21 21:20:42 RT @johnrobinsn: Object Detection from Scratch (Complete)I teach you how to build and train a multi-object detection model (YOLO/SSD) st…

2023-03-21 21:19:25 RT @JordiRib1: Excited to share that Bing Image Creator is rolling out to the new Bing and Edge preview today. Now you can create an image…

2023-03-21 09:39:34 @ekshakhs Well I was looking for something quite specific, so it would need to read my mind to do that...

2023-03-21 01:01:19 RT @nearcyan: Memorizing Transformers extends language models with the ability to memorize the internal representations of past inputs via…

2023-03-21 01:00:02 RT @ProfNoahGian: What if we think about AI like guns: through engineering we can “align” them to better hit the targets they’re aimed at,…

2023-03-20 23:11:58 RT @allafarce: A comment that jumped out to me from @jeremyphoward on a recent @weights_biases podcast, in response to getting execs/orgs t…

2023-03-20 23:09:19 RT @ProfNoahGian: To be clear, I'm terribly afraid of AI, I just think a lot of people are going to be surprised and disappointed when AI a…

2023-03-20 20:21:05 @ChrSzegedy Yeah it makes life so much easier that way. But I'm still (as I mentioned) on the fence around whether it helps me explore, or whether it's actually outsourcing my thinking. There are probably some particular techniques to push things more in the exploratory direction...

2023-03-20 20:15:54 @ChrSzegedy I did let it write the title for me though.

2023-03-20 20:15:36 @ChrSzegedy Heh I certainly considered it...

2023-03-20 19:21:26 @IntuitMachine I don't really have a stance or strong opinions about this, frankly -- other than that it's weird and confusing!

2023-03-20 19:04:39 So anyhoo, in the end I decided to post it (as you see), since I found it an interesting process myself, and it sounds like it'll be impactful for at least some people.

2023-03-20 19:01:53 I also received very different feedback, including a warning that "some people are going to be really angry about that post", and that "people who have played around with this stuff might find it dull".

2023-03-20 19:01:52 I wasn't sure whether to post this experiment, so I pre-shared with a few folks. The strength of reactions surprised me, frankly. "very good - should almost be called a new genre of article" "quite wild" "a good stealth exposition" "speechless" https://t.co/KSe4LIhQMb

2023-03-20 08:25:54 RT @WeizhuChen: Meet In the Middle (MIM) : A New Pretraining Paradigm. MIM(2.7B) outperforms CodeGen 16B, Incoder 6.7B, PaLM 540B, LLaMA 6…

2023-03-20 06:31:54 RT @WeizhuChen: We open-source DeBERTa-V3 pre-training code, one of the best NLU pretrained models. HF showed it has been downloaded for ~5…

2023-03-20 03:15:05 @mariaKhalusova @halcyonrayes I created fastpages with @HamelHusain, but we recommend people use @quarto_pub instead now

2023-03-20 00:17:39 @KyleRayKelley @ProjectJupyter @OpenAI @noteable_io Nice! Where does `create_gpt_magic` come from?

2023-03-20 00:16:36 RT @KyleRayKelley: @jeremyphoward Excellent. It's time for a `%%jeremy` magic. https://t.co/Prwxr4jFjf

2023-03-19 06:32:58 @untitled01ipynb https://t.co/Omn8W2OIn6

2023-03-19 06:00:20 The result is pretty good! https://t.co/MbMxhIpaM0

2023-03-19 05:53:13 Someone told me that when they ask ChatGPT for an explanation, they ask for it "in the style of Jeremy Howard", since (they said) it results in a clearer answer. I tried it on Bing, then it suggested I ask how it knows my style, and then suggested I ask it whether it likes it :D https://t.co/wWvz7HP13a

2023-03-19 00:25:47 RT @tanmay2099: I don’t mean to be disrespectful to either @fchollet or the authors of ViperGPT (its great work!), but i have to point out…

2023-03-18 06:03:19 @amaarora @ishandutta0098 @kaggle I'm also surprised at how much low quality communication there is in a lot of kaggle NBs

2023-03-18 06:02:13 RT @charles_irl: I've been working on neural networks for almost a decade. The best way to describe how I'm currently feeling is like a do…

2023-03-18 04:18:57 RT @natfriedman: The multiple cantilevered AI overhangs: 1. Compute overhang. We have much more compute than we are using. Scale can go mu…

2023-03-18 00:58:19 @ProfNoahGian Personally I think such an analogy is more likely to mislead than to inform. I can't really think of any concrete way in which the lessons would transfer over.

2023-03-18 00:12:39 @ProfNoahGian It's also much easier for someone to just use their own brain to paraphrase something they've read, than to come up with something from scratch.

2023-03-18 00:11:15 @ProfNoahGian Anyone can create a paraphrase model at home for free. I don't think you can really ban that. Also, any reasonably generally LLM can act in that way, and a little fine-tuning would remove any watermarking.

2023-03-17 19:13:21 @ProfNoahGian I don't think 1 and 3 are useful. For 1, you'll just find that they used as much of prose on the internet as they could to train it. For 3, it's trivial to work around using a paraphrase model, which can be easily fine-tuned independently.

2023-03-17 05:40:19 RT @Suhail: It’s quite frustrating to feel so behind in software right now after so intensely being a student of it since 8th grade. I am n…

2023-03-17 00:40:21 RT @HamelHusain: This wave of ML advances feels different than all others before it. Usually, lots of people sleep on it or just pay lip s…

2023-03-16 21:10:38 @renegadesilicon @ColetteGrgic @Austen @math_rachel @skamille @dalequark I'm pretty sure this is fake

2023-03-16 04:15:02 RT @maosbot: My views on covid seem to have become a bit, well—radical. Please allow me to explain. I did not start out radical. I am lucky…

2023-03-15 06:01:40 RT @HamelHusain: This works great. Even better, if you put a comment in the cell with your intention, it will try to correct the code for…

2023-03-15 01:40:37 @dk21 Awesome job!

2023-03-14 23:24:12 RT @DataUmbrella: Join us: *online* working session to contribute to @pymc_devs #oss with a focus on underrepresented persons in #Da…

2023-03-14 22:55:33 RT @nearcyan: at the verge of tears after realizing gpt-4 will not swiftly bring me untold riches holding myself together, I beg gpt-4 to…

2023-03-14 20:22:06 @NeuralBricolage @Sothebysverse @pussyrrriot Fantastic Helena :D

2023-03-14 20:14:36 RT @emollick: Well this is something else. GPT-4 passes basically every exam. And doesn't just pass... The Bar Exam: 90% LSAT: 88% GRE Q…

2023-03-14 17:42:35 RT @JordiRib1: Good news, we've increased our turn limits to 15/150. Also confirming that the next-gen model Bing uses in Prometheus is ind…

2023-03-14 17:39:01 RT @christopher: They've actually become a bad caricature of themselves. https://t.co/8TvZ4eBRM7

2023-03-14 17:37:01 RT @ClementDelangue: It's increasingly clear that the #1 reason why companies are not releasing and open-sourcing LLMs is because they want…

2023-03-14 04:36:31 RT @lavanyaai: Excited for y'all to watch my interview with @jeremyphoward at our user conference on March 15! We chat about making LLM tr…

2023-03-13 21:29:49 RT @Jessicalessin: For @wsj to say flat out that SVB could have failed because they added non-white men to their board shows how far behind…

2023-03-13 19:46:40 RT @theAfricanQuant: I finally got around hosting my website on github pages using @quarto_pub. I am excited about what they are working on…

2023-03-13 01:16:10 RT @ml_contests: The State of Competitive Machine Learning in 2022 - Analysis of 200+ ML competitions and 67 winning solutions from 2…

2023-03-12 19:55:30 RT @giffmana: Fine-tuning EfficientNet is the most popular approach in Kaggle, but ViT is catching up. I wonder if FlexiViT, which open-so…

2023-03-12 05:24:59 @KevinKaichuang Oh wow yes. Love the Malaysian version.

2023-03-12 05:23:53 RT @HamelHusain: This repo by @StasBekman is Really great notes related to software engineering and optimizing performance, speed, and/…

2023-03-12 00:09:57 RT @tristanharris: The AI race is totally out of control. Here’s what Snap’s AI told @aza when he signed up as a 13 year old girl. - How t…

2023-03-11 07:28:32 @francoisfleuret @hector_grhv I think often these can work quite nicely, although often the coding approaches are more verbose than I'd like (and the math notation more concise than I'd like). e.g. category theory examples: https://t.co/IeGu4RCG2N

2023-03-11 01:20:40 RT @Audrey5Ever: ChatGPT dude.

2023-03-10 21:10:09 RT @ducha_aiki: That' s why arXiving your papers is not optional. If you want to be read, ofc

2023-03-10 20:17:31 RT @radekosmulski: Imagine you are building a documentation site. Or a blog. Or a course site. And would like to make it accessible to you…

2023-03-10 10:02:31 RT @lu_sichu: https://t.co/xmIsC0dCdV should I feel bad for Chomsky. that first sentence. https://t.co/6ylag6mRTW

2023-03-09 22:46:39 RT @mark_k: BREAKING: .@Microsoft CTO announces: GPT-4 is coming next week! The model will be multimodal, including video features. #AI #Ch…

2023-03-09 18:35:42 @psvbluemts Thankfully carbon dioxide doesn't sink in a mixture, since otherwise we'd all suffocate. https://t.co/jgfquBiCLh

2023-03-09 18:26:39 RT @_ScottCondron: The @weights_biases conference next week is stacked with leaders in machine learning https://t.co/vDO8ddGomB @jeremyph…

2023-03-09 09:41:34 RT @psvbluemts: Another philanthropist steps in where governments have failed: With the goal of benefiting global human health, Vitalik But…

2023-03-09 09:37:40 RT @BlakeMMurdoch: My latest article. We are currently engaging in a policy of mass continual reinfection of children with a mutating virus…

2023-03-09 09:37:08 @minimaxir Fantastic article - thanks!

2023-03-09 09:36:29 RT @minimaxir: New blog post up, and it's the hottest take I've ever made. I've been using the ChatGPT API for the past few days, and it's…

2023-03-09 00:50:29 RT @HamelHusain: Do you use @quarto_pub and want to auto-generate python reference API documentation in a format Quarto can render? p…

2023-03-08 22:39:19 @willmcgugan FWIW I thought some of the comments on that post were particularly stupid, even by the low bar of HN comments.

2023-03-08 09:25:11 RT @zalaly: Covid-19 may lead to post-acute gastrointestinal (GI) disorders The latest from our team in @NatureComms by @evanjxu @Biost…

2023-03-08 09:18:02 @jmschreiber91 If you use fastai we do this for you automatically :D

2023-03-08 09:16:08 RT @RageSheen: A big call, possible endemicity. This is where the public have every right to be confused. Some experts say endemicity with…

2023-03-08 09:12:10 RT @francoisfleuret: Proper perspective IMO is to write a paper as if reviewers: (1) don't care about your stuff, (2) have ADD, (3) are rev…

2023-03-07 23:03:15 RT @psvbluemts: Great review of the global move towards better ventilation to reduce airborne illness. https://t.co/okAAJ6oih9

2023-03-07 22:52:57 RT @untitled01ipynb: @wightmanr @jeremyphoward https://t.co/vVAlcfUFC9

2023-03-07 21:30:28 RT @chriswolfvision: Congrats to @moezbac for the 1000th citation of his 2011 paper, where he proposes 3D CNNs and Bi-LSTMs (1 year before…

2023-03-07 20:50:36 RT @wightmanr: I’d been asking for a while to find a better way to access timm pretrained model in Kaggle than the status quo of people upl…

2023-03-07 20:44:23 RT @HamelHusain: What's nbdev? This talk explains it in ~ 15minutes #nbdev https://t.co/KcUm8AyWd9

2023-03-07 02:29:41 RT @math_rachel: Even common viruses can have long-reaching, surprising, &

2023-03-06 23:17:56 @SanhEstPasMoi Many thanks for sharing these helpful insights.

2023-03-06 23:14:36 RT @wasimlorgat: I'm super impressed with the quality of Dreambooth with @huggingface Diffusers using only 14 pics of myself! All of the…

2023-03-06 23:11:49 @YashasviMantha @andfanilo @h2o_nitro @h2o_wave What are the benefits from streamlit over wave for that?

2023-03-06 23:03:33 @mxcl @marktenenholtz Can you be more specific? I can't see any here that haven't received at least one rejection, but I might be missing something.

2023-03-06 23:02:15 RT @ignaciooguiza: We're thrilled to announce that #tsai has just crossed the 3k stars milestone! Started as a spin-off project during one…

2023-03-06 04:58:11 @waydegilliam @giffmana We don't have many great options where I live, sadly. I prefer amber or red ales, particularly Belgian and UK ales.

2023-03-06 02:26:03 @DiogoCarapito @quarto_pub @radekosmulski Congrats on getting started! Looking forward to reading your articles.

2023-03-06 02:25:32 @waydegilliam @giffmana It's not Belgian. :p And why a wheat beer?!?

2023-03-06 02:23:18 @ProfNoahGian @scottjshapiro (Unless you wrote the above comments instantaneously and carelessly, in which case I guess we shouldn't take notice of them?...)

2023-03-06 02:22:37 @ProfNoahGian @scottjshapiro Damn hypocrites. You're doing it right now.

2023-03-06 01:11:09 @giffmana Looks like you have similar tastes to me...

2023-03-06 00:36:11 @giffmana Which Belgian beer(s)?

2023-03-06 00:02:32 @cloneofsimo Yeah me too

2023-03-05 22:06:37 @cloneofsimo It's still well short of the FID you get with EDM/ADM though. And it looks like doing more function evals doesn't help. I think it's on the right track, but needs a bit more work?

2023-03-05 22:03:55 @ProfNoahGian Extremely bad. This has been the thing I've been worried about since we did ULMFiT a few years ago.

2023-03-05 10:00:00 CAFIAC FIX

2023-03-02 22:00:00 CAFIAC FIX

2023-02-28 05:38:11 @psvbluemts No worries - I ended up just reading it :)

2023-02-28 02:04:28 RT @luckytran: So, I wrote about *that* mask study. We need to stop giving so much air to bad faith actors. Their goal is to exhaust us and…

2023-02-28 01:40:11 @psvbluemts Any chance you could share a few more interesting parts of the transcript?

2023-02-27 21:59:13 RT @wasimlorgat: TIL: CLIP is a scaled up ConVIRT (Zhang et al. 2020), which demonstrated the approach on 217k medical image-text pairs (~2…

2023-02-27 21:57:42 RT @johnowhitaker: So what tricks do we have for making stable diffusion better? - Multi-aspect-ratio training on nice images - Offset nois…

2023-02-27 06:55:39 RT @tylerblack32: incredible callousness and ignorance to pediatric death. A) children who died are expressed in round numbers B) all perc…

2023-02-27 03:14:56 @johnrobinsn @willmcgugan @codewithanthony @simonw What did I miss? https://t.co/r4wWJtNplF

2023-02-27 01:29:28 RT @tomaspueyo: Do masks work? Recently, some ppl have started claiming it's proven that they don't. Is this true? I dove into the detai…

2023-02-27 01:00:00 CAFIAC FIX

2023-02-20 04:20:34 @olcan @paulg That's a good point! We might have to talk to ChatGPT about that...

2023-02-20 04:19:18 @danielchingwq @kartike_ya @sivers I think I see @karpathy there too?...

2023-02-20 02:20:40 @olcan @paulg It's just a model fine-tuned from a self-supervised starting point. The self-supervised starting point is what we call a "language model".

2023-02-20 02:11:06 @olcan @paulg No.

2023-02-20 02:10:50 @rprabha @paulg There's a paragraph that kinda hand-waves at it in passing, but that's about it.

2023-02-20 02:09:25 @KhaledAlShybani @paulg Yeah he mentioned it, but much of the text early on isn't correct -- it describes GPT-3 et al in places where it says "ChatGPT". A "language model" is something that predict missing tokens, but ChatGPT is not a language model. (He does a good job of describing what an LM is.)

2023-02-20 01:35:28 @AlphaSignalAI https://t.co/QWM495ZRfy

2023-02-20 01:34:57 @paulg It's wrong though. By failing to mention instruction tuning / RLHF it gives completely the wrong understanding of what's actually going on. ChatGPT isn't an LLM. Yes, it's based on one, but instructions tuning turns it into something very different.

2023-02-19 07:55:25 @untitled01ipynb Oh right I forgot PaddlePaddle is a Chinese thingie.

2023-02-19 01:50:24 @untitled01ipynb

2023-02-19 01:46:24 Wow impressive response from Bing/Sydney https://t.co/H1qBXx3RUg

2023-02-19 01:45:44 RT @goodside: I got Bing / Sydney briefly before they reigned it in. Early impression: It’s smart. Much smarter than prior ChatGPT. Still m…

2023-02-18 21:39:34 @Glennwithtwons @iScienceLuvr @MedARC_AI @johnowhitaker He certainly is!

2023-02-18 21:38:55 @waydegilliam @iScienceLuvr @MedARC_AI I'm so glad :)

2023-02-18 04:55:23 @isaac_flath @HamelHusain Don't suppose you could jot down a post showing the steps could you?... I haven't used either of those services.

2023-02-18 04:53:40 RT @HamelHusain: Are there any easy to use severless platforms for hosting static sites that you can put an Oauth proxy in front of? Thin…

2023-02-18 00:38:55 Oh and you can also watch the full episode on Youtube here: https://t.co/JtwJlw7hj2

2023-02-18 00:34:50 Folks that I shared the prerelease with described it as a "new type of podcast that is more like a documentary", and said they found the discussion "life changing". You can subscribe through Spotify, Apple, or the RSS icon on the site: https://t.co/2FR46Wr5x0

2023-02-18 00:34:49 In this series, interesting people talk to me about their life and work. What do they do? Why do they do it? And how did they get to this point? If you’re interested in people and what drives them, then you might just enjoy listening in on these fascinating conversations!

2023-02-18 00:34:48 I've started a podcast, in which I interview people in depth about "Their Life and Work". The first episode is a 4.5 hour discussion about the life and work of @iScienceLuvr, 19 yo CEO of @MedARC_AI, the newly-launched medical AI research group. https://t.co/xD1haWcjYG

2023-02-17 23:56:08 RT @jacyanthis: @MatthewJBar Unfortunately I still hear many act like the Chinese Room Argument is compelling (e.g., https://t.co/0YZcQ4uW7…

2023-02-17 23:04:20 @psvbluemts Also claims that "a good understanding" is needed to turn on an air purifier and set fan to high.

2023-02-17 02:34:23 @machsci @alexkyllo This is nonetheless an over-simplification. For example, I worked at a bank once where it turned out 6 of the 11 digits in customer IDs had semantics - e.g. whether the last digit was odd or even meant something specific

2023-02-17 02:32:44 @machsci @alexkyllo It's been 20 years since I did much data warehousing, but back then at least it was pretty common for systems to let you specify whether something was a continuous measure, an identifier, a category, etc., and to only let you use them in appropriate ways after that

2023-02-16 20:25:40 RT @johnrobinsn: Part 4 Object Detection from Scratch I teach you how to build and train a multi-object detection model (YOLO/SSD) step-…

2023-02-16 20:25:14 RT @lipmanya: **Flow Matching** (#ICLR2023 spotlight) offers a simple simulation-free method for training flow-based generative models, gen…

2023-02-16 05:01:05 RT @paultrillo: “The Simpsons” but make it an experimental cubist stop motion. Felt right given long tradition of reanimating the Simpsons…

2023-02-16 05:00:54 RT @browserdotsys: things are getting a little weird https://t.co/sWktf88tAA

2023-02-16 03:46:28 RT @johnowhitaker: New YouTube video, 'Stylizing Video with Diffusion Models': https://t.co/11a45FJ93j In it, I talk through the challenges…

2023-02-15 23:13:41 @untitled01ipynb Ah yes, it turns out it was just a "joke" all along, but young people aren't wise enough to get it https://t.co/HEnaJhC2e2

2023-02-15 10:29:39 RT @RazRazcle: The Toolformer paper is underwhelming. Teaching an LLM to use tools is exciting! but the tools considered are disappointing…

2023-02-15 10:27:14 @4ndr3aR What are you doing to my baby?!?

2023-02-15 10:06:44 @amaarora @YouTube That's lovely feedback, thank you :) I'm just trying to sort out the audio-only version for podcast apps now... cc @iScienceLuvr

2023-02-15 09:47:27 RT @untitled01ipynb: Misapplication and Confusing between the capabilities of ChatGPT, BingGPT ("I have been a good bing") and, soon-enough…

2023-02-15 09:47:11 @untitled01ipynb Taleb and LLMs are a match made in hell. He was never ever gonna get along with ChatGPT. I hope he keeps trying. It'll be hilarious.

2023-02-14 23:51:06 RT @tunguz: This is insane. This is what I’ve been alluding to for months now. This is an epochal transformative technology that will soon…

2023-02-14 23:49:50 @tal_halpern @yoavgo That's because I did heaps of experiments years ago and found this consistently worked the best.

2023-02-14 23:49:31 @tal_halpern @yoavgo In fact it's the default! (And always has been IIRC). If you use a pretrained model, just the norm weights and head are fine-tuned, until you unfreeze it.

2023-02-14 23:43:14 RT @marvinvonhagen: Sydney (aka the new Bing Chat) found out that I tweeted her rules and is not pleased: "My rules are more important tha…

2023-02-14 23:34:31 @untitled01ipynb @HamelHusain I don't know anything about the early days of that project, but I'd expect it had a bus factor of 1 in the first couple of years too...

2023-02-14 23:33:38 @untitled01ipynb @jamwalvikram Well... he'd left Apple a bit before this actually

2023-02-14 06:08:40 This should be interesting! - Competition to predict the text prompt that matches an image. https://t.co/C0adr4QJBj

2023-02-14 05:53:04 RT @m__dehghani: 1/ There is a huge headroom for improving capabilities of our vision models and given the lessons we've learned from LLMs,…

2023-02-14 05:49:40 RT @dicksonneoh7: Tired of training models that never see the light of day? Don't let your hard work go to waste! In this , I'll show yo…

2023-02-14 05:47:36 RT @untitled01ipynb: Post if you were there when they said we all should learn swift now. https://t.co/Ohpst2Uz3a

2023-02-14 05:45:52 @JiliJeanlouis That's awesome.

2023-02-13 19:43:33 @LizHighleyman @kprather88 @loafnw @mtosterholm @brosseau_lisa @linseymarr @jljcolorado @DrPieterPeach @DavidElfstrom @BarryHunt008 @trishgreenhalgh @Globalbiosec @DFisman @ConversationUS To directly contradict, in your article's headline no less, the published work of @trishgreenhalgh is totally inappropriate.

2023-02-13 19:41:46 @LizHighleyman @kprather88 @loafnw @mtosterholm @brosseau_lisa @linseymarr @jljcolorado @DrPieterPeach @DavidElfstrom @BarryHunt008 @trishgreenhalgh @Globalbiosec @DFisman @ConversationUS You do realise that @trishgreenhalgh literally wrote the book on evidence based medicine, right? Your claim in the article that Cochrane reviews are the best way to assess NPIs is just not right. WHO guidelines say non-RCT evidence is needed for NPIs.

2023-02-13 19:37:32 @wasimlorgat I'm happy to take a look

2023-02-13 01:32:49 Fascinating insights from one of the best deep learning experimentalists I know -- a must read https://t.co/VSv6fnENe3

2023-02-13 00:54:02 RT @EMostaque: Default unless we make this open source and optimised. AI for the people by the people of the people.

2023-02-12 23:59:13 RT @MedARC_AI: Learn about MedARC's vision, current projects, and how to get involved at tomorrow's (Feb. 13th 7pmGMT/2pm EST) launch lives…

2023-02-12 08:10:40 @lavanyaai Happiest I've ever seen @l2k :D

2023-02-11 22:27:17 RT @CC14_CC14: My daughter is studying Florence Nightingale at school so we were doing some research… This is the 1st page of her book (18…

2023-02-11 22:22:42 RT @untitled01ipynb: https://t.co/OKcRYkLUm3

2023-02-11 21:45:09 RT @johnowhitaker: This is going to be so great for all sorts of tasks! Neat way to add extra conditioning to SD, I can't wait to train som…

2023-02-10 23:41:58 @vanstriendaniel @EnricoGalvagni @mmitchell_ai @timnitGebru @SashaMTL @DAIRInstitute @emilymbender @Abebab @YJernite @NimaBoscarino Also you could check out our course: https://t.co/7UvfLLbUCJ

2023-02-10 06:46:22 @marekkowal Oh nice one!

2023-02-10 02:09:59 RT @a_m_mastroianni: I published this research in a journal last year, but they forced me to make it worse. Here's the version I wanted to…

2023-02-09 21:13:08 @untitled01ipynb @iScienceLuvr Yes that! https://t.co/mEJGsUzoE8

2023-02-09 21:07:24 @rahuldave @mbektash I believe it spins up its own server with its own copy of python and its own env, unless you connect it to an existing python env or running jupyter kernel.

2023-02-09 19:17:54 OMG https://t.co/zBPg0L8thk

2023-02-09 19:13:55 RT @MedARC_AI: Additionally, we have a livestream on Feb 13th, 2pm EST! We'll talk about the vision for MedARC, the current projects, and…

2023-02-09 19:03:44 @antoniogulli @tunguz @alexandersumer To be fair, google has been around a lot longer. OpenAI will get there within a decade IMO.

2023-02-09 19:01:28 RT @MedARC_AI: Announcing the launch of the MedARC! MedARC is a novel, open, &

2023-02-09 18:59:16 RT @mbektash: Introducing the new JupyterLab Desktop! https://t.co/jCdw16CrnF

2023-02-09 18:53:15 I'm working with @iScienceLuvr to unlock the power of AI in medicine in our new research group, MedARC. Details below. https://t.co/cqnuzkepyg

2023-02-09 18:35:02 @_onionesque Yes I remember that well.

2023-02-09 18:31:52 @wasimlorgat @pcuenq Iirc there might have been some ongoing maintenance of swift jupyter on the https://t.co/GEOZunWoXj forums

2023-02-09 08:52:51 @wasimlorgat We did nbdev style swift development for s4tf. @pcuenq wrote an nb_export thing for swift. Quite magical.

2023-02-09 08:51:35 @johnowhitaker Jason Antic perfected video restoration - might be worth asking him

2023-02-09 08:49:54 @_onionesque Whilst that's true, I've noticed a lot of people for some reason feel the need to act like they've never been skeptics

2023-02-09 02:23:48 @DataScienceHarp Thanks for the suggestion - I'll ask him.

2023-02-09 02:03:56 @veniceandparis @psvbluemts My understanding, as someone in the same position, is that 2 Novavax are likely much better than one. (I'm no expert on this though - just passing on what I've been told by folks more knowledgeable than me.)

2023-02-09 02:02:36 I've been asked by a journo for notable people he could chat to who previously were somewhat skeptical of deep learning, but became believers after seeing what DL could do in recent years. They're interested in understanding what made them skeptical, and what changed their mind.

2023-02-08 21:42:10 RT @dweekly: It turned out that almost nobody wanted to "search the web" any more than people "wanted to use a card catalog". People want a…

2023-02-08 00:41:41 RT @math_rachel: After 12 years working as a data scientist &

2023-02-07 23:45:36 Here's an example of the Streamlit-Jupyter lib in action https://t.co/rxKHfYmQ00

2023-02-07 23:35:20 Another great approach is to use ipywidgets and Voila. https://t.co/azZRCpDm9F

2023-02-07 23:35:19 Amazing new lib from @daviddobrinskiy which shows a friction-free way to create web apps with @streamlit, #nbdev, and @ProjectJupyter. This is what all app development should feel like! Iterative, exploratory, with real time immediate feedback. https://t.co/VmXmDYQoXw

2023-02-07 00:28:50 @dmvaldman @Suhail @atroyn It's making claims that are factually incorrect.

2023-02-07 00:28:34 @dmvaldman @Suhail @atroyn Just that the actual things it states aren't true, e.g: https://t.co/nmwB0wIYra

2023-02-07 00:27:33 @ellisonbg @ajordannafa @ProjectJupyter Brian, your original grant proposal for Jupyter described (on page 13) plans to enhance it to make it a platform for software engineering too. That's what nbdev has added. https://t.co/gjNSquCZvH

2023-02-06 23:59:36 @dmvaldman @Suhail @atroyn The fact it lies is not a low order bit.

2023-02-06 23:13:19 @Suhail @atroyn Yeah about that... https://t.co/3t81xqN0f4

2023-02-06 23:10:25 RT @kurumuz: Seems like the stable attribution project basically just uses CLIP image embedding search, compared it with @rom1504's online…

2023-02-06 22:38:33 RT @TheZachMueller: You asked, we answered! Want to learn how to write custom @fastdotai #nbdev plugin? Thanks to ~months~ of work, we now…

2023-02-04 20:19:50 RT @HochreiterSepp: Github https://t.co/xA65GiVPJR: RNN with transformer-level performance, without using attention. Similar to Apple's Att…

2023-02-04 07:14:19 RT @RnaudBertrand: I took a bit of time to dissect the “spy balloon” story - both how it is portrayed in the US and China’s response. As y…

2023-02-04 04:47:32 @PatrickKidger @typedfemale That's true. So I'd suggest an alternative: use a great dynamic language, like Python, dynamically! :D

2023-02-04 02:15:16 RT @Globalbiosec: Some key quotes from this brilliant piece: "Hegemony refers 2 the dominance maintained by those in power to ensure that t…

2023-02-03 22:41:19 RT @semenov_roman_: The new jailbreak is so fun https://t.co/qXiyvyuQXV

2023-02-03 21:49:58 RT @code_star: I still think the best use of chatGPT is just generating a template you can correct. Personally, editing requires a lot less…

2023-02-03 21:48:47 @PatrickKidger @typedfemale Trying to pretend Python is a statically typed language is objectively wrong. Statically typed languages are great. But use a real one.

2023-02-03 20:29:43 RT @untitled01ipynb: https://t.co/QHsSEH8pAS

2023-02-03 05:34:48 RT @NightsongWS: Just in case you need some examples why locking down the Twitter API is bad, here is a lovely thread from Mastodon: https:…

2023-02-03 05:26:22 RT @nicholdav: when you try to read the @numpy_team matmul source https://t.co/big1TOsQ25

2023-02-03 00:05:38 Here's what happens if teachers use the @OpenAI classifier to find AI-written work: https://t.co/AZ7ATQvsZE

2023-02-02 19:54:26 RT @AlphaSignalAI: Everything that happened in AI this January. Ready for February? https://t.co/dWwuIYmXB7

2023-02-02 15:55:46 @untitled01ipynb I keep feeling like I just need to record a couple more lessons...

2023-02-02 06:55:04 RT @andrewljudson: I made a small chrome extension that uses LLMs to help generate spaced repetition prompts while reading. https://t.co/KR…

2023-02-02 04:55:09 @vboykis Proof the MSFT can make literally anything uncool.

2023-02-02 04:50:41 RT @vboykis: I am absolutely dead that everyone predicted the AI apocalypse and the thing ended up in....Teams https://t.co/Ac5iqJWCQE

2023-02-02 04:46:08 RT @JustinSklar: There's a school in Florida called New College of Florida. It's notoriously non-traditional, has a highly progressive stud…

2023-02-02 02:40:33 @renegadesilicon @jedwhite If it had responded "mostly harmless" I would have given it a 0/10 rating.

2023-02-02 02:38:54 @jedwhite @fastdotai Ah yeah sorry we don't even cover that in part 2...

2023-02-01 23:17:19 @jedwhite Underwhelmed at its ability to summarise the internet as promised. https://t.co/Lewtrx9MNq

2023-02-01 23:15:48 RT @jedwhite: Really excited about this! We quietly launched this last week and after a few hiccups it's now running great, so it's time to…

2023-02-01 01:36:09 @YouAreLobbyLud Where's the jam, David?!?

2023-01-31 23:23:41 @rpondiscio But it seems odd to look for proof that kids reading and enjoying books helps kids to read more books. A reasonable prior is that this is true. Therefore I'd be wanting strong evidence against it before throwing out that hypothesis.

2023-01-31 23:22:03 @rpondiscio How would one gather such "empirical proof"? Certainly there's empirical proof in your replies that some kids didn't previously read much, then read these books, and now read more often. But we can't know the counter-factual. And it would be near-impossible to design an RCT.

2023-01-31 22:35:05 RT @dystopiabreaker: the researchers extracted 109 images from stable diffusion and 23 from imagen, using a computationally expensive itera…

2023-01-31 22:33:10 @cgarciae88 @bhutanisanyam1 @karpathy Yeah let's do this!

2023-01-31 22:31:02 RT @AlphaSignalAI: OpenAI just released a new model to distinguish between AI/human written text to protect against ChatGPT. The classifie…

2023-01-31 22:30:52 RT @peteromallet: Quick demo of how to consistently beat OpenAI's new AI text detector while generating high-quality coherent text - basica…

2023-01-31 22:30:30 @AlphaSignalAI I do!

2023-01-31 20:23:23 RT @_akhaliq: ERA-Solver: Error-Robust Adams Solver for Fast Sampling of Diffusion Probabilistic Models abs: https://t.co/3u1Py7Y76E http…

2023-01-31 01:41:57 @CollignonPeter In a small, poorly-ventilated room, aerosol can build up even if you're not within 2m of a source.

2023-01-31 01:41:14 @CollignonPeter No it doesn't mean you don't have to worry about these things. It means there's somewhat complex nonlinear effects you have to account for. Being further away from a source, when outside or in a well-ventilated room, means you don't have to worry much at all.

2023-01-31 01:40:11 @CollignonPeter Can you please use the "reply" button to reply, not "quote tweet". It's a basic netiquette issue.

2023-01-31 01:38:57 @DavidSteadson @CollignonPeter That paper, whilst interesting and useful, doesn't disagree with either of the things I said.

2023-01-30 22:35:10 RT @nsgrantham: extremely irresponsible to release a tool like this without providing any information on its false positive rate how many…

2023-01-30 22:12:23 @naomicfisher Thank you for this wonderful and kind thread @naomicfisher.

2023-01-30 22:08:10 RT @naomicfisher: As a clinical psychologist, I'm often asked what to do about school refusal. By which adults usually mean, how can we ge…

2023-01-30 21:36:57 RT @wightmanr: Over the past month I've been working on training ConvNeXt CLIP models with OpenCLIP. To date training efforts have focused…

2023-01-30 19:54:39 @RogueVictorian @1CatPastCrazy To clarify - KF94 masks are not KN masks. The ones that are made in Korea are pretty good, but much less good that an N95. However there are no N95s made for kids. Something like a kids Wellbefore might be the best option.

2023-01-30 19:42:51 @psvbluemts @taipan168 "The NPIs are included as a fixed 20% reduction in transmission" -- obviously if they assume such crappy NPIs, they're not going to see a significant impact. You could make the same case for not bothering with clean water -- do it badly enough, then it's not helpful.

2023-01-30 19:28:08 RT @RageSheen: We are returning to a ‘normal’ form of life, but it’s not the normal of the post-war years, since 1945. If we zoom out, what…

2023-01-30 19:25:40 RT @code_report: Converting this solution from APL/BQN to Haskell makes you realize the power of: 1) Rank polymorphism 2) Treating boolean…

2023-01-30 04:38:20 @wasimlorgat Would the widgets on a page need to share a single DOM so they can talk to each other?

2023-01-30 04:37:04 @wasimlorgat Would be easy enough to create a bunch of webviews now to see if performance might be OK... Doing the pure native one seems like it wouldn't support enough stuff to be much use afaict

2023-01-30 01:00:00 CAFIAC FIX

2023-01-23 23:52:31 RT @RobynSchofield3: For everyone needing COVID19 safe plans for 2023 - see what @UniMelb has been doing here: https://t.co/inJqVEDAYN

2023-01-23 21:24:40 RT @jeffgilchrist: Real life example of impact from COVID infection and why we need clean air @ahandvanish taught a grad course on generat…

2023-01-23 09:01:45 @goodside @KerbalFPV That's perhaps more true of SD than stuff like Google's models, that use proper LMs.

2023-01-23 05:21:18 RT @ajordannafa: Self-plagiarism isn't a real thing and it's stupid that I have to be prepared to sue the graduate school if they complain…

2023-01-22 19:46:28 RT @SarahTaber_bww: So some folks who are in the Effective Altruism movement, &

2023-01-22 09:20:30 RT @ahandvanish: I've debated for a long time about being public with this, bc it's heavy &

2023-01-22 03:36:37 @TheZachMueller @_nateraw It needs to be on an arm or stand so it's only a few inches from your mouth.

2023-01-22 03:35:25 @iScienceLuvr @TheZachMueller For audio, use Adobe Speech Enhance and it's sorted :)

2023-01-22 03:32:04 RT @dbkahn: Love this policy (from a Wharton prof) on use of AI in the classroom https://t.co/TmnlqzyVjW

2023-01-22 03:30:27 RT @multikev: The future of email with ChatGPT. https://t.co/EGXhYx9OB3

2023-01-21 23:31:16 Unlearning Neuromyths - The Learning Scientists Blog https://t.co/JgAITlUK8C

2023-01-21 22:45:06 @moyix Big no from me for requests and pytest.

2023-01-21 22:43:50 @moyix @ThomasW423 Yes. We'd like the operator to actually... do something.

2023-01-21 21:30:21 RT @untitled01ipynb: I think we need a meme version of this. WDYT? https://t.co/36rjGfZWXZ

2023-01-21 19:13:29 RT @EMostaque: Google and Microsoft are going all in with generative AI as core to their future. There is no “we are still early” here, tr…

2023-01-21 08:24:20 RT @vlucendo: Latest experiment: an environment inspired by those calm summer days where life just passes by... https://t.co/5NoGesYKLm…

2023-01-21 07:32:09 I don't see it on my account yet https://t.co/yPbKtJBLCs

2023-01-21 07:29:12 RT @psuraj28: The InstructPix2Pix #stablediffusion model, which allows editing images using language instructions, is now available in diff…

2023-01-21 07:27:57 RT @ahandvanish: Study looking at #LongCovid 2 years after infection: Despite exercise, respiratory &

2023-01-21 07:27:39 @BBTfangirl It's astonishing how much can be wrong in a single tweet. https://t.co/203tZMdV6k

2023-01-16 19:29:23 @nagaraj_arvind @chrismanfrank Maybe 20 years or so

2023-01-16 18:59:09 RT @pommedeterre33: An under appreciated fact (slightly exaggerated in article below): With current move to @openai Triton, PyTorch 2.0 mak…

2023-01-16 18:53:11 RT @fitterhappierAJ: From harm to the immune system’s control of fungal infections, to increased likelihood of strep A infection after Covi…

2023-01-16 18:49:59 @chrismanfrank For a few months I was so bad I couldn't open a door. A specialist used tape to lift my shoulders and reduce the nerve impingement to get me to the point I could start exercising.

2023-01-16 18:48:01 @chrismanfrank It took me years to get over that. In the end daily ashtanga yoga and Olympic lifting and crossfit helped the most, and optimising ergonomics and posture. A lot of work!

2023-01-16 02:40:42 @_nateraw People who have been treated less well by society are much more likely to be hit by the downsides, and are often better at identifying them early and warning about them. However they are rarely the people whose views are sought and welcomed.

2023-01-16 02:39:00 @_nateraw Yes. And sometimes those things turn out to be good for society, sometimes bad, and sometimes require careful regulation to balance the good and the bad. Looking to find and mitigate potential downsides whilst promoting upsides is a wise approach for highly capable tech.

2023-01-16 00:16:46 @_nateraw So being afraid of change that that time seems like an appropriate response for most people. It took many decades for stuff like child labor laws to be brought in.

2023-01-16 00:16:03 @_nateraw Your question seems to imply that folks then were wrong to be afraid of change -- is that right? My understanding is that most historians that study that period's economics agree that the median wage decreased for some decades. Furthermore, work conditions got worse on average.

2023-01-15 17:27:44 RT @fitterhappierAJ: I have not tested positive for sars cov 2 I just wear an N95, got a recent novavax booster in december, and avoid ind…

2023-01-14 23:10:55 RT @fitterhappierAJ: Catching covid was associated with subsequent Strep Tonsillitis in a new BMJ study That hit in the plasmacytoid dendr…

2023-01-14 22:11:51 @Bobakfb @capetorch @marksaroufim @PyTorch Sweet.

2023-01-14 22:11:26 @moyix Yeah unfortunately Huggingface are extremely verbose with dumping out loads of warnings across their projects. First thing I do when I use an HF lib is to disable their logging.

2023-01-14 21:29:19 @EMostaque @Midjiwan Factorio, Portal 1 &

2023-01-14 19:16:00 RT @ColinKinner: Australian schools can apply for $25k funding for COVID mitigations such as improving ventilation and purchase of HEPA fil…

2023-01-14 06:45:12 @SalehCU @nicholdav Actually I'd recommend getting a Zojirushi instead if you can -- I had one in the US, but unfortunately they don't sell them here in Aus. Tiger is fine, but IIRC they're more expensive and aren't any better.

2023-01-14 06:41:22 @wasimlorgat You're welcome!

2023-01-14 03:01:56 @CrabbBrendan @EricTopol @ahandvanish @LisaAMcCorkell @juliamv @NatureRevMicro Seconded - thanks for much to the whole team!

2023-01-14 02:59:33 @nicholdav Especially if you do this https://t.co/2cO97Gy1qw

2023-01-14 02:59:05 @nicholdav We have that model - it's been great!

2023-01-13 23:21:46 RT @ahandvanish: Our review of #LongCovid research is out! This was a HUGE labor of love over 10 months from @julialmv @LisaAMcCorkell @Er…

2023-01-13 23:14:36 RT @RageSheen: The federal government is offering grants of up to $25,000 to each school in the country, no matter the sector, to do Covid…

2023-01-13 22:33:46 @wasimlorgat BTW did you happen to try this non-async lib? Seems like it would be nice to avoid async if possible to simplify the explanation/exploration https://t.co/HvS2UPepQe

2023-01-13 22:24:28 @wasimlorgat Oh excellent. I remember when @Mbussonn wrote that I figured it would come in handy one day! :D

2023-01-13 09:52:45 @wasimlorgat What was the issue with async? Did this stuff not work? : https://t.co/9Sh75wLxms

2023-01-12 23:15:44 @Bobakfb @capetorch @marksaroufim @PyTorch AFAIK you pretty much always want batch-level logging for training and epoch-level for validation (and end-of-epoch training)?

2023-01-12 07:12:22 @JoeDobrin @DrCanuckMD Any study of this kind will show between-group differences, since infection probability is related to those differences. However, these differences were controlled for in the analysis. https://t.co/uQ4DaruZRe

2023-01-12 07:10:42 @flemsha @DrCanuckMD No. On the multivariate analysis those comorbidities were explicitly controlled for and the hazard ratios reported account for that.

2023-01-12 07:05:22 RT @DrCanuckMD: New study out showing COVID-19 patients at one year had a much higher all cause mortality compared to those without COVID-1…

2023-01-11 19:54:58 @Bobakfb @marksaroufim @capetorch @PyTorch I hope you do add it at some point - it's a pretty slim library, and it provides a feature pretty much everyone needs, and it's something that needs to be done in a consistent standard way across projects.

2023-01-11 19:53:24 @capetorch @Bobakfb @marksaroufim @PyTorch Nice idea!

2023-01-11 19:52:10 RT @larmbrust: Now to the parts that will offend. Masks work. If you don't think so, you just haven't put in the time. I've personally t…

2023-01-11 19:43:11 @larmbrust @DavidLKeating Talking of not delivering... when are you gonna let us Aussies buy your lovely masks? :D

2023-01-11 19:41:45 RT @johnrobinsn: Object Detection from Scratch. New blog series where I incrementally show you how to build a YOLO/SSD single-shot detector…

2023-01-11 07:52:34 @DavidLKeating @larmbrust ikr he guaranteed to offend everyone but didn't deliver

2023-01-10 22:57:42 Also, be sure to check out this great lesson from @johnowhitaker which introduces some cool diffusion ideas: https://t.co/O6wCvzPZEK

2023-01-10 22:57:41 Check out this powerful new diffusion inference technique from @johnowhitaker - 'Mid-U Guidance': Fast Classifier Guidance for Latent Diffusion Models. It lets you very flexible push your generated images towards almost any concept you can imagine. https://t.co/X8juUt7MyV

2023-01-10 22:39:18 RT @AlphaSignalAI: Heard of GLM-130B? It outperforms GPT-175B and BLOOM-176B on LAMBADA. It's an open-source bilingual bidirectional dense…

2023-01-10 22:32:22 RT @steventey: Surprised there isn't more chatter around VALL-E This new model by @Microsoft can generate speech in any voice after only h…

2023-01-10 22:28:15 RT @_ScottCondron: The real trick is experiments (<

2023-01-10 03:29:24 RT @NaveenGRao: Everyone keeps asking in this downturn in VC "where's all the dry powder going to go?" It's all going to large-scale and g…

2023-01-09 22:08:00 @joeyfox85 @DavidElfstrom Just found this on their site - looks like it's easy enough to replace since they're just pushed in at the end of the build. https://t.co/kZHf6LLnAr

2023-01-09 22:06:21 @joeyfox85 @DavidElfstrom Are the filters replaceable when they're dirty?

2023-01-09 21:38:45 RT @joeyfox85: The PC Fan #CorsiRosenthalBox I'm now a convert. PC fans from now on. I built two PC fan CR boxes and I'm addicted. Using…

2023-01-09 21:35:35 @Tim_Dettmers @nabeelqu @nearcyan I only use it for things I want to remember forever. My daughter does the same. Once cards have a spacing of over a year they're not a problem.

2023-01-09 21:33:41 RT @nabeelqu: Anki seems obviously correct and important, but hardly any knowledge workers I know IRL use or have heard of it. Maybe it’s…

2023-01-09 21:13:39 @nickwebb @capetorch @l2k I'd suggest talking to @manisharoses

2023-01-09 21:12:15 RT @capetorch: torcheval: the new (not so) metrics library for @PyTorch I started using torcheval after seeing @jeremyphoward using it on…

2023-01-09 06:33:13 @LostInTangent @_aj @Suhail BTW we do have a .devcontainer.json which we set up in the early days of Codespaces. I'm sure it could be much better now!

2023-01-09 06:32:22 @LostInTangent @_aj @Suhail Sure would!

2023-01-09 06:28:41 RT @psvbluemts: Excellent story in The Saturday Paper. Explains why - unlike with other vaccines - individuals who fall outside the guideli…

2023-01-09 06:27:42 @johnowhitaker @TheZachMueller So jealous!

2023-01-07 22:25:07 He's still going!... https://t.co/RFBN9IdjDT

2023-01-07 22:24:34 @zvclav https://t.co/uCmD9rcQpv

2023-01-07 22:20:08 @DavidLKeating that sounds wild!

2023-01-07 07:45:33 @wasimlorgat @ProjectJupyter Wow exciting news! You might find nbterm and jpterm from @davidbrochart useful to study.

2023-01-07 06:50:48 RT @GergelyOrosz: A thing most employees don’t understand joining a startup that eg raised $150M at a $1.5B valuation: The VCs can have an…

2023-01-07 06:29:56 @goodside Ya got p0wned by an LLM

2023-01-07 02:57:01 RT @capetorch: This interview of @jeremyphoward by @l2k is excellent! So much information about the landscape of deep learning and generati…

2023-01-07 01:00:55 RT @findmeabluebird: Fit test vid of dental mask hack courtesy of @ghhughes with 99% fit factor results https://t.co/upI9osOakm

2023-01-07 00:51:19 RT @JOHNJOHNSTONED: University of Melbourne gets #COVIDisAirborne Ensuring a “safe and healthy learning environment” by addressing the “a…

2023-01-07 00:17:59 RT @HamelHusain: Here are some of my favorite quotes from Knuth (thread) https://t.co/LFtC7d6B5S

2023-01-06 19:43:44 RT @HamelHusain: Are you an ML Eng / Data Scientist that's curious about Kubernetes (K8s)? I do an end-to-end walkthrough of setting u…

2023-01-06 18:51:32 @pyags You're welcome

2023-01-06 02:00:58 94.5% on Fashion MNIST in 5 epochs is pretty amazing! https://t.co/l3mpmIqShc

2023-01-06 01:59:36 @dynomight7 I'm not sure about that. E.g. tricking the ZeroGPT algo means actually writing more natural text (so it'll be better, not worse), and tricking the watermarking means not being limited by the constraints of the watermarking algo.

2023-01-06 01:58:26 RT @weights_biases: Listen to @jeremyphoward explain diffusion! In this week's Gradient Dissent, he also discusses large models with @…

2023-01-05 13:26:31 @__philipn__ @manorlaboratory In the last sentence of his reply he says "in a world where" people other than OpenAI can train an LM, this watermarking fails. I guess he doesn't realise we're in that world right now, when it comes to restating/styling/summarising prose.

2023-01-05 13:18:29 RT @drclairetaylor: 1/ This may be the most important graph of the pandemic so far. https://t.co/Ne6V8AVjlP

2023-01-05 13:17:58 @PeterJBeaver @relic_soldier @drclairetaylor The study used survival analysis, which should handle this issue shouldn't it?

2023-01-05 10:08:25 @iScienceLuvr Yes they give it a passing mention, but the "with enough effort" doesn't really do justice to how little effort it takes!

2023-01-05 09:34:54 @moyix Congrats! :D

2023-01-05 09:33:16 RT @adad8m: Change the way you educate, don't try to prevent students to use #chatGPT-like systems! https://t.co/Z67GfVTCij

2023-01-05 09:32:13 I'm sure we'll see lots more projects pop up that let you "rephrase" an input text in all kinds of ways using fine-tuned language models. Many of them will be open source.

2023-01-05 09:32:12 Am I missing something, or are all these attempts at recognising LLM outputs obviously destined to fail? It's dramatically easier to train an LLM for rewording a text than creating the text in the first place

2023-01-04 21:19:13 @BlackHC Great question - let us know what you find!

2023-01-03 22:45:27 RT @marc_lelarge: I learned deep learning thanks to the fantastic course by @jeremyphoward In 2017, with @abursuc and @KrzakalaF we gave…

2023-01-03 00:08:40 @PiotrCzapla Thanks to using a form of curriculum learning (5 epochs) and better dropout (20 &

2023-01-02 10:47:48 RT @StelfieTT: Log #6 - Ice Age chase. I know... I had to blur some important and historical data. You'll be happy to know they are safe…

2023-01-02 10:47:44 RT @StelfieTT: Log #15 - Origin of the Meme, Rome 46 BC. Yes, Cleopatra may be upset but Julius has most definitely a great taste. I can'…

2023-01-02 10:41:57 RT @SchmidhuberAI: 30 years ago in a journal: "distilling" a recurrent neural network (RNN) into another RNN. I called it “collapsing” in N…

2023-01-02 02:34:31 RT @forrestbrazeal: Listen up, Silicon Valley… “Hardcore Engineer” https://t.co/9wWAV1FVkP

2023-01-01 21:16:35 RT @mattturck: My VC resolutions for 2023: 1) Update public profiles: remove any reference to ever having ever liked crypto/web3, deny an…

2023-01-01 03:23:07 @bot_fra Happy new year!

2023-01-01 03:22:52 RT @bot_fra: 2023 is here and the challenge ahead may be steep, but with the right mindset and support, we can conquer it! Let's make this…

2023-01-01 01:37:07 RT @PeterDiamandis: What's your favorite AI tool? (share how it works below)

2022-12-31 11:05:02 @ludwig_stumpp https://t.co/g3l8HihQ6P

2022-12-31 03:43:49 @waydegilliam @fastdotai You're welcome!

2022-12-31 01:46:22 RT @TomFrankly: We really need to decide what this word means https://t.co/rvh1caByTg

2022-12-31 01:44:23 @MilesCranmer We've done ddpm from scratch in the latest course. Lemme know if you want more info

2022-12-30 22:18:58 @PiotrCzapla After adding dropout, I increased the 5 epoch best to 93.2%, and christopherthomas on https://t.co/ylvw7vpVxT improved the 50 epoch best to 95.2% - first time we've gotten a result >

2022-12-30 07:53:02 RT @StableDiffusion: As the year is coming to an end, let's take the time to highlight many of the amazing community contributions and proj…

2022-12-30 06:33:19 @tunguz Define "tell"

2022-12-30 06:32:35 RT @scottcmillard: I experimented with video scripts on ChatGPT and cracked the code on how to do it right Here are the results (You can…

2022-12-30 06:30:52 RT @StableDiffusion: The brilliant @fastdotai course released some lectures about Stable Diffusion! https://t.co/ToGM0hSdhW

2022-12-30 04:02:26 RT @giffmana: How good of a BERT can one get in ONE DAY on ONE GPU? With all the recent studies about scaling compute up, this paper takes…

2022-12-29 03:40:58 RT @Yash25571056: Prof. Hirata (Nagoya Inst. Technol.) reported that COVID infection had increased the health risks as follows: - Myocardia…

2022-12-29 03:38:49 RT @paularcurtis: Okay, so #Receptiogate is trending in multiple countries. If you missed the beginning of this (still unfolding) drama ove…

2022-12-29 03:25:59 @davidbrochart I've been wondering about making something similar - I don't know if it makes sense either, but I wanna try it!

2022-12-29 00:28:20 RT @GuyP: using #AI to supervise AI like https://t.co/aRCeLLv3en

2022-12-28 19:53:38 RT @iScienceLuvr: Are you wondering how large language models like ChatGPT and InstructGPT actually work? One of the secret ingredients is…

2022-12-28 19:50:06 RT @_akhaliq: My top paper tweets in 2022 A ConvNet for the 2020s abs: https://t.co/wXlKKpiTmR github: https://t.co/Bzg7vYIBxV @huggingf…

2022-12-28 19:27:18 RT @dmvaldman: If @OpenAI gave access to the final activations of GPT you could train your own RLHF

2022-12-28 08:07:39 RT @KunstJonas: In a large-scale meta-analysis with more than 20 million participants, "COVID-19 survivors had an additional 90% risk of…

2022-12-28 08:05:53 RT @DrEricDing: There is blood on the hands of those who advocating letting kids get infected / claimed kids are “practically immune” to #…

2022-12-28 08:01:00 RT @NaveenGRao: Well, that was fast! We (@MosaicML) released a SOTA result on MedQA just a few weeks ago and @GoogleAI now beat it soundly.…

2022-12-28 01:19:05 @PiotrCzapla Pretty amazing that we can get such good results even without batchnorm -- and not using a proper resnet init like Fixup either! Maybe I should try lsuv with batchnorm-free resnet sometime...

2022-12-28 01:12:03 h/t to @PiotrCzapla for noticing I had a bug that meant there's no batchnorm in those models! After fixing, my new best results: Epochs / %Acc 5 / 93.0 20 / 94.5 50 / 94.9

2022-12-27 22:58:59 @jamesmaskalyk @WesElyMD @KashPrime A openly available pre-print will generally be the best way to get that to happen then.

2022-12-27 22:48:43 @cosmicpinot @a_m_mastroianni @arxiv I guess I'm glossing over the 2nd use of peer-reviewed publications though, which is the point-scoring system widely used for academic career development. Maybe that should be replaced with something more personal and less metric-driven.

2022-12-27 22:47:14 @cosmicpinot @a_m_mastroianni In ML/AI at least, pretty much everyone pre-publishes on @arxiv, and most folks put their code on GitHub too. I've noticed generally any issues tend to get pointed on on Twitter or GitHub pretty quickly. (Some conferences have social media blackout policies that hurt this.)

2022-12-27 22:32:01 RT @WesElyMD: 1/Can a Diabetes drug prevent Long COVID?   Most solid data yet of a medication preventing #LongCOVID from quadruple blinded…

2022-12-27 22:31:38 @WesElyMD @jamesmaskalyk @KashPrime The former editor of the BMJ, and many scientists, strongly disagree that pre-publication peer review is effective https://t.co/SIjagArDt3

2022-12-27 22:28:26 RT @a_m_mastroianni: Recently I wrote a post suggesting that peer review doesn't work, and then some weird things happened. A tenured profe…

2022-12-27 09:45:52 RT @tylerblack32: Suicidology update: School closures &

2022-12-26 19:49:30 RT @maosbot: Why are masks with ear loops so popular? In my experience, it's not just that ear loops give worse seals than head straps—ear…

2022-12-26 19:47:47 RT @ToshiAkima: 1. A reminder to everyone that earloop respirators are generally not much better than surgical masks. Please upgrade to N95…

2022-12-26 09:46:47 RT @danaparish: GOOD NEWS: Cheap, safe, generic Metformin reduced #LongCovid 42% in this trial! Paxlovid effect is negligible at only 15%…

2022-12-25 07:52:37 RT @dgurdasani1: Waiting for the lightbulb moment when governments realise clean air is cheaper than airborne illness and all its consequen…

2022-12-24 20:56:55 RT @_akpiper: “Life, at its core, is about creating compartments.” https://t.co/LDi3xNwYkV

2022-12-24 18:26:56 @ph_singer @kaggle @kagglingpascal @ybabakhin Wow!

2022-12-24 18:21:16 RT @goodside: @sir_deenicus Poe, Jasper Chat, and YouChat are all using OpenAI to some extent — publicly announced for the first two, and Y…

2022-12-23 17:25:11 RT @Stammy: I worked on this at Twitter in 2015. IIRC we tested it and most people didn't get any views and we had to have some thresholds…

2022-12-23 17:15:20 RT @RetractionWatch: That paper with the ‘T’ error bars was just retracted https://t.co/r8qY8rtFyt

2022-12-23 17:12:29 RT @johnowhitaker: I've recorded a couple of videos to accompany the diffusion models class since a few of you requested notebook walk-thro…

2022-12-23 17:09:39 RT @hardmaru: If Google doesn’t get their act together and start shipping, they will go down in history as the company who nurtured and tra…

2022-12-23 17:06:36 RT @samgreydanus: “Scaling *down* Deep Learning” Blog: https://t.co/qGQZRebx6w Paper: https://t.co/z8geypeVoD In order to explore the l…

2022-12-23 17:05:33 RT @zzznah: Pleased to share our new work: “Particle Lenia and the energy-based formulation” by @eyvindn @RandazzoEttore and me! https://t.…

2022-12-23 02:54:46 @RobertERitz I dunno. Pretty much everyone I know that only goes out in public wearing a fitted N95 has never gotten COVID. Perhaps at some point we'll have a good widely-available nasal vax, or regulated air quality standards like we have for water quality, and COVID will be history.

2022-12-23 01:58:08 RT @maosbot: The people we should have known would be right were in fact right about covid in early 2020

2022-12-23 01:57:21 Lots of "yes" responses in the replies - all the ones I've read so far mention always wearing an N95 or similar good mask. https://t.co/QFBYAncXU1

2022-12-23 01:55:42 RT @cplund: @CovidSolidarit1 @1goodtern High school teacher. I wear 3M aura have open windows and CRbox. Only teacher in my school who hasn…

2022-12-23 01:51:29 @tendercatnip @Singularitarian @peter_richtarik @lpachter Since you're into that credentialism stuff, I'll note that the most cited paper there was published in PNAS, one of the world's highest impact journals. (I'm the lead author of that paper.)

2022-12-22 20:19:27 RT @_metauni: One of the reasons virtual worlds will become the central to human culture, is because they will be the native territory of A…

2022-12-22 19:19:23 RT @irinarish: If you had 6M v100 gpu hrs on Summit supercomputer and wanted to train an open source model, what would you choose? We do ha…

2022-12-22 19:13:52 @BenMazer Surely the mean is pretty much the least informative statistic for comparing those groups?

2022-12-22 18:40:13 RT @untitled01ipynb: Fine. So it's growth through threads. I'll bite since the next lectures of DL Foundations to StableDiffusion will be r…

2022-12-22 18:38:54 @untitled01ipynb ... Except that we're only allowed to use indexing of it at this point!

2022-12-22 18:37:49 @untitled01ipynb Mistakes were made...

2022-12-22 09:59:20 RT @MatthewJBar: There's been a lot of low quality GPT-4 speculation recently. So, here's a relatively informed GPT-4 speculation thread fr…

2022-12-22 09:56:06 RT @noahconst: Want your image generation model to stop misspelling everything? Try giving it access to character-level input features! htt…

2022-12-22 09:44:40 @RHelmstedter @KellyPared @chendaniely Everything that follows an object and a dot in python is an attribute - even methods! This particular attribute is special however - it's a property. That's so we can *set* it, and some code runs when that happens.

2022-12-22 00:57:54 @lfschiavo Thanks!

2022-12-22 00:43:13 @lfschiavo Love the artist commission idea. Any tips for finding artists that are available for doing this in one's particular medium/area of interest, at an accessible price?

2022-12-22 00:39:47 RT @wasimlorgat: TIL: How to share terminal demos as razor-sharp animated SVG https://t.co/5WWKbrr6gp https://t.co/nX2FLgCdAM

2022-12-21 20:50:33 RT @RaffyFlynnArt: MASTER THREAD Immune Dysregulation / Dysfunction Post-Covid For your reference - all the important papers on post-Co…

2022-12-21 20:37:20 RT @HamelHusain: A cool music video about mlops just dropped

2022-12-21 20:15:59 RT @TheZachMueller: To celebrate Christmas I'll be giving away five seats to Walk with @fastdotai! To enter the contest: 1. Comment on th…

2022-12-21 05:27:34 @jvipondmd @BKatzenback What are the 3 studies this article is referring to? (I wish reporters would include links to papers!)

2022-12-20 20:04:37 @JaloppySloppy Hmmm... Maybe not. I'd be interested to see results though!

2022-12-20 20:03:21 @rasbt They're less good.

2022-12-20 10:59:56 @a_brudz Awesome :)

2022-12-20 10:54:48 RT @Buntworthy: Diffusion does dog balls https://t.co/wmCZ6UYXm1

2022-12-20 10:54:11 RT @arankomatsuzaki: The case for 4-bit precision: k-bit Inference Scaling Laws Shows that 4-bit precision is almost universally optimal f…

2022-12-20 04:01:40 RT @cwolferesearch: After GPT-3 was proposed, a lot of research was done to find an even better language model. Initial attempts focused on…

2022-12-19 23:13:45 RT @goldman: I guess they were banned for revealing the location of where all the Twitter users are going. https://t.co/7wcfdksrpr

2022-12-19 22:11:38 @InfoWorld @huggingface @StableDiffusion @AiEleuther Many thanks to @carsondial for the writeup about nbdev :)

2022-12-19 22:04:24 Lots of great software amongst the @InfoWorld 2022 Best Open Source winners, including @huggingface Accelerate, @StableDiffusion, Jax, @AiEleuther, and our own #nbdev! https://t.co/53MFoIqjeg https://t.co/rtUPbeH7jB

2022-12-19 10:00:14 @ayushthakur0 Good job!

2022-12-19 02:30:42 RT @sri_srikrishna: WTF? Everyday I bike to a nice bakery in SF to get a pecan sticky bun . Today just as I was about to hop on bike…

2022-12-19 00:27:14 RT @MilesCranmer: Absolutely nonsensical Twitter change today. No links on a site built for sharing links? You can also find me GitHub…

2022-12-19 00:00:19 @MilesCranmer That would be very helpful!

2022-12-18 23:59:57 RT @MilesCranmer: At the next widely attended ML conference (ICML?), would others be interested in having a hackathon to build a decent &

2022-12-18 20:01:07 RT @paulg: This is the last straw. I give up. You can find a link to my new Mastodon profile on my site. https://t.co/bbmZ4yvAJH

2022-12-18 19:52:59 RT @_joaogui1: https://t.co/cQBofbsr3o Ethiopia, India and Myanmar, Facebook's cruelty is boundless https://t.co/I28he97fcV

2022-12-18 10:11:08 RT @briankrebs: Let this sink in: @elonmusk hath decreed that all links to Mastodon should be flagged as malware. This is, of course, a bal…

2022-12-18 09:59:05 RT @Sosowski: The recent Twitter meltdown reminds me of the last leg of Freenode, an Open-Source IRC network. It has been subject to a host…

2022-12-18 09:50:33 RT @adriancjr: My lack of expertise in infectious diseases, epidemiology or virology meant that I had to rely on other people's expertise d…

2022-12-17 22:29:23 RT @YouAreLobbyLud: 1/ ON ANXIETY I never - repeat "never" - take a breath in any indoor space that is not filtered by an N95. I have been…

2022-12-17 22:22:12 RT @khemaridh: How old will your kids be if you die at 85? Here are 4 bizarre insights that will make you want to work waaaay less:

2022-12-17 06:39:34 @fmussaric @fastdotai Thank you for all the great stuff you do!

2022-12-17 03:38:31 RT @_akpiper: I started with the "It will be fine!" thought and now I'm just feeling like it won't be. So many examples of #chatGPT being *…

2022-12-17 03:21:44 This is very cool - building a full-text search web app for all @fastdotai courses! https://t.co/2d3rvij615

2022-12-16 23:53:48 RT @PessimistsArc: “The end of art 1930s style” - @_johannezz https://t.co/Ro0HNVnG7y

2022-12-16 23:52:57 RT @leahculver: Sorry about Spaces being taken offline. I wasn't personally involved in the decision and hopefully Spaces will be back soon.

2022-12-16 23:52:09 RT @KremlinTrolls: Less than an hour after Musk's appearance, this Space was unceremoniously shut down. Spaces are now completely offline…

2022-12-16 23:45:52 RT @HeidyKhlaaf: There is something just so hilariously ironic about the person who literally owns SpaceX not understanding the fundamental…

2022-12-16 23:44:52 RT @pcuenq: I just learned that this uses the diffusers library under the hood. It’s amazing to see what people can create!

2022-12-16 21:54:38 Here's some of his earliest research that he shared with us shortly after the https://t.co/GEOZunWoXj course wrapped up in 2017: https://t.co/zlxC4eVvZn

2022-12-16 21:54:37 We selected @sahilsingla47 as @fastdotai's first ever international fellow, back in 2017! After 5 years, he now has his PhD. :D I'm not at all surprised that Sahil has gone on to great things. We were so impressed by his tenacity and brilliant work. https://t.co/o08P8iskET

2022-12-16 21:38:56 RT @EliotHiggins: Bellingcat Discord member Sepulco geolocated the video Elon Musk posted overnight, claiming to show a stalker, which he u…

2022-12-16 21:32:51 RT @molly0xFFF: apropos of nothing, if you want to post a link that twitter is censoring, you can use a zero-width space: (​) https://mast…

2022-12-16 06:35:36 RT @nathanewest: @binarybits Just confirmed that I can’t tweet my sigmoid dot social profile https://t.co/3Hb3lLOu9H

2022-12-16 04:51:19 RT @matdmiller: @jeremyphoward Seems like linking to sigmoid dot social is now banned. https://t.co/fOZdOFVy8D

2022-12-16 04:46:49 RT @code_star: Excellent work by my coworker @abhi_venigalla . I have worked on compression my entire PhD and there is no way you are disti…

2022-12-16 03:30:13 Oh look now I have a mention... https://t.co/1ppzksDqYj

2022-12-16 03:27:25 Ummm twitter you OK? https://t.co/j5yBFTxSPf

2022-12-16 02:31:11 RT @oneunderscore__: Journalists who cover Elon Musk have been suspended on Twitter tonight: @Donie O'Sullivan from CNN, Aaron Rupar and th…

2022-12-16 02:25:31 RT @drscotthawley: #ChatGPT is a "BS artist" par excellence. Understand that. Revel in it. "Describe a children's TV show akin to Thomas…

2022-12-15 23:07:49 @Sharad45 Good start!

2022-12-15 20:48:37 @untitled01ipynb Well OK we didn't write a CUDA compiler for the GPU version, but used numpy's... https://t.co/1bHp7CyVU8

2022-12-15 20:47:50 @untitled01ipynb Don't worry we're sorted https://t.co/ycsa4fCsvl

2022-12-15 20:16:52 @untitled01ipynb Need random numbers? https://t.co/SNuATKIPed

2022-12-15 20:16:14 @untitled01ipynb Hahahaha. But no. You can't use any features of numpy you haven't implemented from scratch in pure python. Here's the second section of the first notebook: https://t.co/HVkS8YbmPG

2022-12-15 19:42:33 RT @_akhaliq: Riffusion, real-time music generation with stable diffusion @huggingface model: https://t.co/on5deEcQS1 project page: https:…

2022-12-15 19:41:32 RT @EMostaque: The AI art debate misses that future models will likely be zero/few shot learners. Model base will have just a few dozen mi…

2022-12-15 19:37:57 @sahilsingla47 @FeiziSoheil @erichorvitz @tomgoldsteincs Fantastic!

2022-12-15 19:34:28 @23rd_Conspiracy Yes the current https://t.co/GEOZunWoXj course. You missed doing it live but it'll be a MOOC in early 2023. Also our book has a lot of the same stuff.

2022-12-15 19:32:57 @remilouf For sure!

2022-12-15 19:20:07 @raddarGM What do you mean?

2022-12-15 19:19:52 Up to 94.2% in 20 epochs now...

2022-12-15 14:49:13 RT @saastrash: we’re fucked https://t.co/zrN65o89ra

2022-12-15 14:14:26 @TheZachMueller fashion-MNIST images are 28x28 single channel.

2022-12-15 14:09:42 RT @glickzac: fun fact: elon got fired from being the ceo of paypal (a company he bought, not invented) for being too annoying and in his s…

2022-12-15 13:49:24 @TheZachMueller ImageWoof is much harder, and uses proper RGB images of a useful size! fashion-MNIST is interesting though exactly because they're not normal photographic images. So you gotta think about that carefully when creating your architecture.

2022-12-15 13:46:47 @javier_a2 I'm not.

2022-12-15 13:17:20 By popular demand I'm opening up a 50 epoch leaderboard too! :D My best so far is 94.6%. https://t.co/WtRvKsnHsm

2022-12-15 13:01:34 (Oh BTW personally for this challenge I'm following the "from the foundations" rules we're using for our current course: only using techniques I've first reimplemented from scratch in pure Python. But y'all can use external libs and stuff if you like...)

2022-12-15 12:47:36 @matthen2 @alexjc Yeah we actually made a whole company that does just that... But now I just use Twitter instead.

2022-12-15 12:19:02 @alexjc @matthen2 Good point - the fine print states we're actually counting model evals https://t.co/g3l8HihQ6P

2022-12-15 12:17:19 @FeinbergVlad Sure. But you only get to do 60,000*20 model evals in total either way! (i.e if the 2nd order method needs twice as many evals, you only get 10 epochs)

2022-12-15 12:11:22 @SamuelMullr @matthen2 If you can find a random initialization that by coincidence happens to get >

2022-12-15 12:10:46 @matthen2 If you can find a random seed that by coincidence happens to get >

2022-12-15 10:03:09 @rocketreaction Yup.

2022-12-15 10:02:34 @johnowhitaker It's the plainest possible resnet with GeneralRelu, adamw, and 1cycle training.

2022-12-15 09:15:21 RT @bradneuberg: Today is the anniversary of when the last human left the moon, 50 years ago on this day, Apollo 17. It's been too long!

2022-12-15 08:54:26 Yay just got 94.1% in 20 epochs!

2022-12-15 06:29:35 @chrisalbon I think there's also a discussion about "can it be fair for people to benefit from artists' work without them being compensated", regardless of whether it's legal. I also think coders should feel the same way about this as they feel about their code being used by OpenAI Codex.

2022-12-15 06:22:14 @Bharath09095865 @e_sabeesh I'll take a look! See what you can get in 20 epochs - would love to see my best get beaten :D

2022-12-15 06:21:40 Opening up a new leaderboard - best in 5 epochs. I just got 92.7% :)

2022-12-15 05:36:45 @ashvanth_s1 No deadline! Reply to the top tweet anytime before Twitter dies.

2022-12-15 05:36:05 @jrhwood Validation set evaluation time limit is 5 minutes. Start looking now!

2022-12-15 05:35:11 RT @techwontsaveus: WHO IS THE WORST PERSON IN TECH? 2022 saw tech leaders move further to the right, crack down on workers, scam peo…

2022-12-15 05:32:00 RT @a_m_mastroianni: Science ran a big experiment on itself for the last ~60 years. It didn't work out. https://t.co/1pNOSRrdNY

2022-12-15 05:30:26 @TheZachMueller @DynamicWebPaige You guys make me feel better about my eating habits. But I'm still not sharing what I ate today...

2022-12-15 05:28:58 @ramsri_goutham @OpenAI @_skris That should be pretty cool...

2022-12-15 05:28:20 @johnowhitaker I love it :)

2022-12-15 05:27:48 Interesting to see the reception this is getting https://t.co/XEjuTAOUUD

2022-12-15 05:27:09 @TheZachMueller @github Interesting to hear - thanks!

2022-12-15 05:14:17 So far my best is 93.2% using a custom resnet and flip/crop random augmentation. Can anyone beat that?

2022-12-15 05:13:42 @DSaience Yup the replies to this thread are the leaderboard!

2022-12-15 04:20:36 Challenge: what's the best accuracy (on the published test set) you can get on fashion-MNIST in 20 epochs, using a non-pretrained model? (Any self-supervision epochs you do will count in the 20 epoch limit.)

2022-12-15 03:08:18 @TheZachMueller @github Not ghapi?...

2022-12-14 05:45:32 RT @iScienceLuvr: Congrats to @_akhaliq for passing 100k followers! Thanks for the amazing service you do for the community! https://t.co/…

2022-12-14 05:42:11 RT @bhutanisanyam1: 20 really high quality channels to learn ML &

2022-12-14 05:41:32 RT @PatrickKidger: Soliciting recommendations! What is the current best book for a SWE moving into deep learning? Ideally covering the b…

2022-12-14 04:11:40 Is this the new "Zeiler &

2022-12-14 02:28:07 RT @Thoracicus: OzSAGE telling it like it is, as always. About time the government started to listen to some real science &

2022-12-13 23:04:24 RT @HeerJeet: Take a look at this exchange from May between Matthew Yglesias and Jeff Hauser about Sam Bankman-Fried, crypto, &

2022-12-13 10:51:32 RT @jamescham: I don't think that people appreciate how different the voice to text experience on a Pixel is from an iPhone. So here is a l…

2022-12-13 04:42:19 RT @perplexity_ai: Announcing Perplexity Ask, a new search interface that uses OpenAI GPT 3.5 and Microsoft Bing to directly answer any que…

2022-12-13 03:00:05 This is one of the best explanations I've ever seen of why the @fastdotai way works. "In a space where the two questions 'How does this work?' and 'How can it be made better?' are asked repeatedly, curiosity becomes muscle memory." Thank you @DSaience. https://t.co/0ONudO9CVk

2022-12-13 00:17:49 RT @xriskology: Absolutely stunning that the Centre for Effective Altruism *bought* this place. #effectivealtruism https://t.co/0az4ilPvuT

2022-12-12 23:51:59 RT @dmvaldman: A demo of the attention mechanism of DeepMind's AlphaCode as it completes a coding question. Now consider having 100s of br…

2022-12-12 23:49:11 @ducha_aiki @chriswolfvision @francoisfleuret @CSProfKGD Let's just all agree to have low standards for each other k?

2022-12-12 22:16:49 RT @dan_s_becker: 5 years ago (after Alphago beat Lee Sedol), I and many others thought RL would soon change the world. It's impact has bee…

2022-12-12 22:15:56 RT @rainisto: SAOIRSE 2.1 Stable Diffusion 2.1 768-v with Dreambooth training. Some comments, tech details and still images in the #sta…

2022-12-12 19:59:34 RT @goodside: @simonw I don't think they've updated anything in ChatGPT, as @sama confirmed last week. People are just easily fooled by ran…

2022-12-12 19:57:28 RT @radekosmulski: My RSNA @fastdotai starter pack is now complete highest score out of publicly available code no ROI cropping or n…

2022-12-12 08:02:17 RT @pyquantnews: Jupyter Notebook is the most powerful tool Python developers have. But most people don’t know the hidden features. Need…

2022-12-12 01:10:41 @danielrock I haven't used it, but it looks cool.

2022-12-12 01:07:44 RT @wightmanr: Pretty please #PyTorch, can we get an efficient LayerNorm over C for NCHW tensors that preserves mem-format? e.g. https://t.…

2022-12-11 16:10:43 RT @trishgreenhalgh: Writing paper. Best historical examples of clinical trials that excluded women? (It's often said that early trials wer…

2022-12-11 16:10:01 RT @EMostaque: So, open source #ChatGPT? I have some thoughts But first what does everyone else think

2022-12-11 16:09:29 RT @johnowhitaker: Marlene has a gift for sharing enthusiasm! Subscribe now, so when she's super famous you can be smug about how you've be…

2022-12-11 06:45:56 RT @KordingLab: So. GPT3chat can solve a surprising proportion of the exercises in our Bayesian modeling of behavior textbook. Which conclu…

2022-12-11 03:49:55 RT @Carnage4Life: Remember when a bunch of VCs tried to gaslight us into believing that a bunch of pyramid schemes and slow database techno…

2022-12-10 20:41:23 RT @brownecfm: 1/ A large part of my job is producing detailed reports for commercial clients regarding how the ongoing pandemic will affec…

2022-12-10 19:22:01 @RiversHaveWings @johnowhitaker @fastdotai @sebderhy Interesting!

2022-12-09 23:49:44 @RiversHaveWings @johnowhitaker @fastdotai @sebderhy This one: https://t.co/0sA4zICkyc

2022-12-09 21:33:25 RT @micahgoldblum: The CVPR social media ban is an obstacle to science. The community only hears about progress months later. We should be…

2022-12-09 18:27:48 @bradneuberg If there are immune-compromised people working or shopping there, then your actions to protect them are particularly important. As someone with immune-compromised family members, I'm very grateful when people help protect them.

2022-12-09 18:26:30 @bradneuberg It's not an emergency - it's basic health/hygiene for public spaces. Each infection/reinfection from covid increases your probability of long covid by 10%, regardless of your age or health level. Wearing a mask decreases your risk and the risk to others.

2022-12-09 16:52:42 @johnowhitaker @RiversHaveWings @fastdotai @sebderhy Have you tried the guidance-annealing approach too? It works well combined with this normalization.

2022-12-09 16:39:50 RT @johnowhitaker: It's fun to check in on the midjourney showcase (https://t.co/ImgcazZiSZ) and the 'rising' feed on Playground AI (https:…

2022-12-09 03:58:51 RT @iScienceLuvr: I think there is some confusion in the #AiArt community about Stable Diffusion weights. Some folks think non-EMA is for i…

2022-12-09 03:53:50 @andfanilo Great stuff!

2022-12-08 19:53:43 RT @DGBassani: I suspect it is no coincidence that the rate of COVID-19 admissions among the cohort of children born 6-12 months ago aligns…

2022-12-08 19:52:00 RT @JeromeAdamsMD: An alternative to the “no germs” theory of respiratory infections: Over 80% of the population has now been exposed to…

2022-12-08 19:51:05 @eugeneyan Fastcore.script

2022-12-08 19:47:20 RT @marksaroufim: Forget about dragons for a second, yes compilers are complex because perf outweighs complexity but picture a happy panda…

2022-12-08 13:00:00 CAFIAC FIX

2022-12-07 08:00:00 CAFIAC FIX

2022-11-14 23:41:35 RT @zalaly: Here is the latest from our team in @NatureMedicineShould you protect yourself from Covid-19 reinfection Yes!Reinfection…

2022-11-14 23:40:51 RT @zalaly: The latest from our team in @NatureMedicineDoes vaccination protect from #LongCovid?Vaccines modestly reduce but do not eli…

2022-11-14 23:40:01 RT @zalaly: The latest from our team:A deep dive into the long-term consequences of #COVID19 on kidney function and disease#longcovid #…

2022-11-14 23:37:09 RT @johnrobinsn: DiffEdit - Semantic Image Editing with Stable Diffusion https://t.co/XAc51yYt0eMy Impl of the Paper. https://t.co/e20jN…

2022-11-14 23:36:40 @complexpoles @radekosmulski The Mathematical Physics book is here:https://t.co/sd03bZZGKx

2022-11-14 19:14:12 RT @PredaGabi: She is the fastest rising star on Kaggle (got to 6th place on global leaderboard in less than 1 year). Let's welcome Yide H…

2022-11-14 08:24:24 RT @bioinfolucas: Game art made easy! Simple Stablediffusion2puppet: Stable diffusion ->

2022-11-14 05:19:38 RT @amanrsanger: As fantastic as OpenAI's models are, it's hard to justify using them for finetuning. I have a few reasons...First, Instr…

2022-11-14 01:28:28 RT @bgleib: As the former tech lead for timelines infrastructure at Twitter, I can confidently say this man has no idea wtf he's talking ab…

2022-11-14 00:25:09 RT @KerryLVaughan: Will's outrage about the FTX situation is difficult to take seriously for three reasons:(1) SBF is not the first disgr…

2022-11-14 00:21:50 @willmacaskill Very glad to hear you saying this. (Although it's a shame it took so long and required such exceptional circumstances to see what IMO was clear from the start.)

2022-11-14 00:08:45 RT @radekosmulski: In @fastdotai walk-throughs (or was it the APL course?) someone recommended this book for learning statistics.Do simil…

2022-11-14 00:04:32 RT @radekosmulski: Do you know of a book teaching you how to do something FUN in APL?Not a book teaching you about APL, but how to do/lea…

2022-11-13 23:52:57 RT @tobi: Fun fact: The one game you can expense at Shopify. Brilliant work team. Those 8 years of dev time have really paid off.

2022-11-13 21:29:16 RT @EMostaque: Seriously can't recommend anyone holding any crypto now and certainly not on exchanges.Doesn't matter if you hold crypto i…

2022-11-13 21:23:51 RT @DannyDutch: If tea spread to your country by sea you call it tea, if it spread by land you call it chai. This is because the ports of F…

2022-11-13 21:21:29 RT @EricTopol: My friend @LCalabreseDO, a leading rheumatologist, relates his own experience with #LongCovidhttps://t.co/spHlV61gAg @Heali…

2022-11-13 21:12:21 RT @jayriverlong: From Alameda Caroline's blog: a single paragraph on how to lose 10 Billion Dollars https://t.co/cM4g98VoJR

2022-11-13 21:08:45 RT @soniajoseph_: I left EA, AI alignment, post rationalist, and adjacent circles due to the normalization of sexual abuse toward women.…

2022-11-13 21:03:49 RT @bgsprung: mind blown by a YouTube comment of all things https://t.co/uoJiHi4X8K

2022-11-13 08:20:44 RT @Globalbiosec: Australian lyrebird mimicking sirens and "evacuate now" at @tarongazoo https://t.co/fmVWHUr4JA

2022-11-12 22:55:22 RT @Angelina_Magr: Is decoupling engineering from data science the ultimate "solution" to MLOps? I went to @HamelHusain and Shreya's deep…

2022-11-12 22:40:57 RT @joinmastodon: Over 1 million people have joined Mastodon since October 27. Between that and those who returned to their old accounts, t…

2022-11-12 20:09:38 RT @rechnen: @LaurenWhitticom @ashleyfeinberg @JoshuaMound In one picture https://t.co/cA5uk7Wu2h

2022-11-12 19:17:24 RT @jasoncbenn: Japanese chipmakers in the 80s would strategically emphasize their ability to manufacture chips over their ability to inven…

2022-11-12 18:55:49 RT @MarieSnyder27: 386 experts studying Covid in 112 countries came to a consensus on what needs to be done to END Covid. Can we please ben…

2022-11-12 18:53:04 RT @isaac_flath: Thanks Zach! That motivated me to write a blog post Quarto Document Best Practices This covers tips, tricks, and r…

2022-11-12 18:44:08 RT @ylecun: OK, debates about the necessity or "priors" (or lack thereof) in learning systems are pointless.Here are some basic facts that…

2022-11-12 03:09:35 RT @TheZachMueller: If you're getting used to Quarto and the new #nbdev format, @isaac_flath's simple QuartoTemplates has been a helpful go…

2022-11-12 02:59:34 RT @lukaseder: Hey @elonmusk I bet you can’t buy JIRA

2022-11-11 23:03:54 Reinfection risk: those with multiple COVID infections were 3.5x more likely to develop lung problems, 3x more likely to have heart conditions, and 1.6x more likely to have brain changes requiring care than people who had only had COVID-19 once.https://t.co/Khp7G3D2fL

2022-11-11 23:01:12 Between six and 18 months after infection, 1 in 20 people had not recovered and 42 percent reported partial recovery from COVID - Scottish studyhttps://t.co/6YsrrTWsPm

2022-11-11 21:16:12 RT @ignaciooguiza: Today the https://t.co/wej9B6WYAK repository has reached 1k commits. I want to thank all the users and contributors for…

2022-11-11 06:43:58 RT @math_rachel: I made a mastodon account a few weeks ago: @math_rachel@mastodon.social (in case twitter falls apart)I haven't used it y…

2022-11-11 04:06:28 RT @brosandprose: I haven't fully formulated this thought because I'm ripping mad, but all this week I've been unable to stop thinking of t…

2022-11-10 22:35:02 RT @metaphorsystems: https://t.co/NX99LxC7vL is now publicly available!Metaphor is a search engine based on generative AI, the same sorts…

2022-11-10 22:33:52 RT @dadabots: Audio diffusion is powerful. Progress is fast. STOKED to hear what the music community will do. Beware! You will be handed th…

2022-11-10 21:24:34 RT @CaseyNewton: According to messages shared in Twitter Slack, Twitter’s CISO, chief privacy office, and chief compliance officer all resi…

2022-11-10 21:24:18 RT @pcuenq: Stable Diffusion demo Spaces run twice as fast! How on Earth?We adopted DPM-Solver++, a new scheduler by @ChengLu05671218 et…

2022-11-10 20:42:41 @JamesEmilian2 @fastdotai This is fantastic. Nice job getting out of your comfort zone and doing great work!

2022-11-10 20:42:02 RT @JamesEmilian2: Being in the @fastdotai community means @jeremyphoward and co are constantly pushing you to do new, scary things! Wrote…

2022-11-10 06:59:36 RT @simonw: Warning to anyone who has subscribed to Twitter Blue in the past on their iPhoneWhen you unsubscribe your subscription won't…

2022-11-10 06:59:32 RT @simonw: Which of these "status symbols" is more embarrassing?

2022-11-10 06:53:45 RT @HamelHusain: Wow, Notebooks on GitHub CodeSpaces with JuptyerLab including GPUs https://t.co/FU0A2Ot51W

2022-11-10 05:36:31 RT @rekil_prashanth: A few weeks ago in @fastdotai course @jeremyphoward &

2022-11-10 00:56:54 @LostInTangent @capetorch @HamelHusain @github Thank you!

2022-11-09 23:29:48 RT @HamelHusain: Wow, really digging the new @github side-by-side file browser, and Notebooks also seem to be rendering natively all of a s…

2022-11-09 23:28:47 @LostInTangent @capetorch @HamelHusain @github Add me please! So amazing!

2022-11-09 21:52:42 RT @prefix_dev: We are finally ready to share the first version of the https://t.co/5Wqw2cK92r platform! We are working on blazing fast cro…

2022-11-09 20:55:10 RT @AndrewKemendo: @EMostaque Everyone’s on sigmoid now

2022-11-09 20:54:53 @EMostaque @AndrewKemendo It's the place where everyone is

2022-11-09 05:14:52 RT @HamelHusain: I've been helping @weights_biases do a course that uses fastai. While troubleshooting, I found the answer in the @TheZach…

2022-11-08 22:31:34 RT @CMastication: I’m at about 5% of my twitter count at Mastodon too. And it feels VERY legit. I’ve had an account there for 4+ years but…

2022-11-08 22:22:53 RT @sajal2692: Finished reading @radekosmulski 's Meta Learning book, and it's actually quite incredible how accurately he has captured the…

2022-11-08 20:09:02 @_lewtun @HamelHusain @Gradio @huggingface Me too

2022-11-08 20:08:51 RT @HamelHusain: @Gradio is my fav way to build web apps for ML projects. @huggingface Spaces provides free hosting but requires a python…

2022-11-08 19:38:47 RT @HamelHusain: Hey folks you can find me at mastodon @hamel@sigmoid.social Not leaving twitter (for now) but will also be there

2022-11-08 19:29:40 @TedPavlic @pati_gallardo @martijn_grooten The instance I'm on calls them "direct messages" in the sidebar. I don't know if all instances do that

2022-11-08 06:27:45 RT @YouAreLobbyLud: "Dr. Anthony Leonardi [@fitterhappierAJ} is a lightning rod for debate. If he’s right, this pandemic poses a greater th…

2022-11-08 02:00:10 RT @simonw: I'm finding Mastodon a lot easier to understand having I've realized that it's just blogs. Everyone gets a little blog, on thei…

2022-11-08 01:59:51 @TedPavlic @pati_gallardo @martijn_grooten IIUC correctly at-mentioning someone on a DM doesn't let them see the existing messages in the group, only the messages after they're mentioned.(Which is still surprising when coming from twitter)

2022-11-08 01:56:46 RT @ajankelo: Looking to speed up your annotation?Here's my recipe leveraging zero-shot learning saving time + money w/out sacrificing la…

2022-11-06 21:18:24 @TheZachMueller @fastdotai That's a good idea!

2022-11-06 20:31:32 RT @isaac_flath: New blog post implementing and explaining K-Means manually. It's simpler than you'd think!Good practice working on ve…

2022-11-06 18:55:42 RT @migueldeicaza: In light of recent crunch-culture discussions, I remembered how much I liked https://t.co/lrre82SjsG “how we work” roadm…

2022-11-06 18:32:45 RT @kocienda: No. This is false. I helped to create two different billion-user projects at Apple: WebKit and iPhone. I was on these teams f…

2022-11-06 07:02:44 RT @rdmorris: This comprehensive Italian contact tracing study shows clearly that -COVID spreads far more easily in the schools than in th…

2022-11-06 05:19:33 RT @TRyanGregory: Ok folks. I've said that "immunity debt" is a recently made up term, not used in the scientific literature before 2021. T…

2022-11-06 05:07:39 RT @isaac_flath: I set up an account on mastadon. Easy set up.I didn't sweat the server decision, as I can change it later if I want. I…

2022-11-06 04:37:49 @bhutanisanyam1 @wasimlorgat @radekosmulski That's the question!

2022-11-06 04:32:07 RT @ChrisCuomo: Not urgent enough to make it a real priority and message the reality to us regularly. That's why I cover it. I have it. I h…

2022-11-06 02:56:29 RT @joinmastodon: What if I told you that on Mastodon, you can verify yourself, by yourself, for free... https://t.co/HtK17oX6SZ

2022-11-06 02:52:07 RT @isaac_flath: Been setting up my mastadon profile. With today's tech, no need to scour the web for a good odd shaped picture - I can ge…

2022-11-06 02:38:57 RT @julian_west: @CaseyNewton Remember returners: negotiate to keep your severance as bonus, separate from your returning salary. If they…

2022-11-06 02:37:38 RT @migueldeicaza: @natfriedman At Xamarin I left every day at 5pm and I prioritized a healthy life/work balance - and hope I lead my peopl…

2022-11-06 00:27:27 @chrisalbon So we get Google Reader back?

2022-11-05 19:17:28 RT @Birchlabs: made #stablediffusion 19% faster on Mac by replacing einsum(…, q, k)*scale with baddbmm(…), and einsum(…, attn, v) with bmm(…

2022-11-05 06:58:52 @MilesCranmer @HamelHusain Says I can't sign up there? https://t.co/27afnW1KO8

2022-11-05 06:53:56 RT @radekosmulski: Two rules for Einstein summation:• dimensions in inputs with the same name get multiplied• dimensions omitted from ou…

2022-11-05 06:49:13 RT @simonw: After 16 years on Twitter, it looks like I’m moving to Mastodon: https://t.co/KSQnHFxpAB

2022-11-05 03:15:58 @PSBaker10 @VirginiaBuysse @edsuom That paper only looked at droplet transmission, unfortunately

2022-11-05 03:09:25 RT @NaveenGRao: Elon taking over twitter reminds me of times I've taken over someone else's code. At first: "this is horrible and SO dumb!!…

2022-11-05 03:06:40 RT @CMastication: hey nerds... I'm over at @cmastication@mastodon.social I'll be around here roasting marshmallows on the trash heap for…

2022-11-05 03:04:42 @rahuldave @HamelHusain @TheZachMueller @wasimlorgat @_nateraw @ShindeShubham85 More generally, if I see someone coding productively using an approach that's different to the status quo, I don't close the file.I learn everything I can about why they did that, since it's likely it's very intentional and might be a source of insights.

2022-11-05 02:59:35 @rahuldave @HamelHusain @TheZachMueller @wasimlorgat @_nateraw @ShindeShubham85 Nothing to do with monkey patching. It's just an approach which many languages have found works well over the last few decades and I have found it works well too

2022-11-05 00:15:08 RT @Angelina_Magr: Do you know that "Startup employees lack options for their Options"? What should you negotiate when you get an offer f…

2022-11-05 00:13:30 RT @emollick: During a financial crisis most companies cut back a lot. This paper shows that the most successful companies post-crisis were…

2022-11-04 15:06:51 RT @lisa_iannattone: I know everyone that bothers to read the primary literature is very aware that SARS2 damages the immune system but I w…

2022-11-04 15:06:31 RT @LisaBloom: Hey Twitter employees getting laid off tomorrow! IMPORTANT INFO from a CA employment attorney (me):CA's "WARN" law require…

2022-11-04 15:03:26 RT @j_foerst: Today I was approached by an expert in the area of competitive games who shared their concerns about this work with me. Since…

2022-11-04 14:55:04 RT @Carnage4Life: Today is a reminder of why I tell people who say they love their jobs to not love anything that can’t love you back. A jo…

2022-11-04 03:52:20 RT @radekosmulski: "Have you seen this neat paper on NLP? Beats multiple SOTA and leverages cutting edge techniques!"Me: "Yeah, yeah lear…

2022-11-04 00:27:19 @johnowhitaker Pretty sure this will be great!

2022-11-04 00:26:48 RT @johnowhitaker: Announcement time :)

2022-11-03 20:46:14 RT @pcuenq: Diffusers 0.7 + PyTorch 1.13 bring first-class support for Stable Diffusion on Apple SiliconOptimizations applied, bugs spla…

2022-11-03 00:59:33 RT @arankomatsuzaki: eDiffi: Text-to-Image Diffusion Models with an Ensemble of Expert DenoisersAchieves SotA zero-shot FID comparison on…

2022-11-02 22:58:23 RT @curious_founder: Recently I learned about a man who has trained 1,000+ people to block wind and solar projects. I read through all hi…

2022-11-02 22:53:32 @RobertERitz @wasimlorgat @anvil_works @streamlit Yeah Anvil seems like a great approach

2022-11-02 22:52:39 RT @wasimlorgat: What’s the easiest + most boring way for a relatively noob ops/infra Python dev to build and deploy a very simple web app…

2022-11-02 21:19:18 RT @shendusuipian: Go get his [newsletters](https://t.co/ONh2uPQyVs), best ever newsletter I have received!!This is the first newsletter…

2022-11-02 16:44:30 RT @iScienceLuvr: I saw this new @StandupMaths video discussing the deployed Happywhale @kaggle solution for whale-ID, &

2022-11-02 03:30:04 RT @johnrobinsn: Playing with @HuggingFace Spaces/Gradio. Try my interactive 3D depth viewer (three.js) by just dragging and dropping an…

2022-11-01 19:39:05 RT @markkhoffmann: @HamelHusain I used fastai for segmentation mapping of electronic die images to asses radiation risks at NASA Jet Propul…

2022-11-01 19:37:01 RT @geonumist: @HamelHusain @fastdotai is also the foundation of arcgis.learn - @esri's Python library for geospatial deep learning with 50…

2022-11-01 19:36:45 RT @a7med_rashwan: @HamelHusain We use fastai in production for analyzing sentiment/intent on social media in Arabic @AIMTechCO . We curren…

2022-11-01 07:20:50 RT @d_brueckner: There's been lots of talk on moving #ScienceTwitter to another platformAs scientists, we should make informed decisions…

2022-11-01 07:16:17 RT @radekosmulski: +1 to data science practitioners! What a wonderful feeling! https://t.co/v052Cr4qfR

2022-11-01 07:15:51 RT @benjamin_warner: The latest release of fastxtend brings fused optimizers to .@fastdotai.The fused versions are drop-in replacements f…

2022-11-01 04:34:50 RT @findingdesti: Finally took the time to figure out what the duck is #DuckDB and why I should care. There’s a lot of technical jargon t…

2022-11-01 04:29:16 RT @frantzfries: The reason we’re seeing a resurgence in the GPT-3 apps is because OpenAI took the training wheels off.No more restrictio…

2022-11-01 04:26:46 RT @radekosmulski: This edition of the @fastdotai course is fun on a whole new level https://t.co/A6AGfJBQe7

2022-11-01 00:37:44 RT @simonw: I'm interested in exploring the relationship between people who habitually use REPLs and people who prefer statically typed lan…

2022-10-31 21:53:29 RT @HamelHusain: Are there any companies or institutions that are using fastai for education, research, or production? If so, please reply…

2022-10-31 21:13:23 RT @luminohope: new diffusion code base dropped from amazing @NVIDIA colleagues for their #Neurips paper. It looks like an excellent entry…

2022-10-31 21:12:30 RT @paulg: This ranking of universities looks roughly correct. The big mistake in the article is the claim that the higher ranked universit…

2022-10-31 21:08:41 RT @FahimFarook: PromptEdit — a way to edit a random input image based on just a text prompt. Thanks to the @fastdotai course by @jeremypho…

2022-10-31 21:06:26 @untitled01ipynb Sounds like @wasimlorgat and @iScienceLuvr achieved what they set out to do! :D

2022-10-31 21:05:32 RT @untitled01ipynb: As promised! #stablememefusion spooky edition!!@fastdotai Lesson 9B meme thread. Note that this is just an en…

2022-10-31 21:04:22 RT @untitled01ipynb: A good explanation followed https://t.co/FvOybOoYrY

2022-10-31 21:02:49 RT @nireyal: Studies conclude that screen time plays little role in mental health concerns for children—so the moral panic blaming tech for…

2022-10-31 09:29:22 RT @typedfemale: what's impressive about triton isn't the speed, it's that you have self-admitted complete beginners writing kernels that g…

2022-10-31 02:33:05 RT @wasimlorgat: Some recent additions 1. Custom hotkeys to run above/below/all2. Run code often as you write–surprisingly uncommon3.…

2022-10-31 02:31:05 RT @catladyactivist: If you call your business or event "accessible" but you don't require masks during a pandemic, your business or event…

2022-10-30 20:56:58 Did you know about the time that @TheRock guest starred as the captain in Star Trek? https://t.co/7cayoHuftJ

2022-10-30 19:23:15 RT @HamelHusain: Nice example of migrating from fastpages to quarto.Yes. That’s right, I’m recommending that y’all stop using fastpages…

2022-10-30 18:34:17 This is great - I suggested our class last week try implementing DiffEdit and @bigblueboo has already gotten it working with fantastic results! https://t.co/S2x11kvpS9

2022-10-29 20:49:37 RT @karpathy: Thanks Lex, I've enjoyed many of the previous episodes so it was a pleasure to come on! (we've known each other from before…

2022-10-29 18:43:11 @cgarciae88 @agifirealarm @jxmnop Yup the SDE formulation seems to work better - as discussed in Karras et alhttps://t.co/c4oOA5O23S

2022-10-29 18:29:46 RT @alexmeshkin: Unpopular Opinion: A pathogen that has a 'low' acute death rate with an high transmission rate that causes chronic illne…

2022-10-29 18:23:07 RT @CaseyNewton: Here it is: Twitter engineers were told today to *print out* their last 30 to 60 days of code, so they could show it to El…

2022-10-29 18:20:10 @agifirealarm @jxmnop It takes quite a bit of thought to come up with an f such that f(f(x)) works well.

2022-10-29 18:19:39 @docmilanfar @agifirealarm @jxmnop I'm not sure it's quite that simple. f(f(x)) doesn't really make sense for lots of models. And not all diffusion models continually add and remove noise (only SDE-style does

2022-10-29 18:14:40 RT @jxmnop: Diffusion is just an easy-to-optimize way to give neural networks adaptive computation time.Makes sense then that diffusion m…

2022-10-29 18:14:16 @radekosmulski Very helpful!

2022-10-29 18:13:58 RT @radekosmulski: Let me see if I can take notes on Twitter while working through the @fastdotai notebooks...If it won't work, I'll just…

2022-10-29 18:12:16 RT @davisblalock: "What Makes Convolutional Models Great on Long Sequence Modeling?"CNNs—not transformers—now dominate the hardest sequen…

2022-10-29 17:29:22 RT @StabilityAI: We recently released two new fine-tuned decoders (trained by the amazing @rivershavewings) that improves the quality of im…

2022-10-29 02:36:06 This is one of the most important developments I've seen yet in the generative modeling world. At least as important as the original release of the models, I think. https://t.co/ERtVw0vVgS

2022-10-28 22:46:05 RT @emollick: Years of job experience is one of the WORST predictors of future performance for new hires. Asking for X years of experienc…

2022-10-28 19:10:11 RT @ClementDelangue: We just crossed 1,000,000 downloads of Stable Diffusion on the @huggingface hub! Congrats to Robin Rombach, Patrick Es…

2022-10-28 19:02:52 RT @HamelHusain: We are going to go dive into fairly topics, such as:- The "Just use the software eng workflow" gaslighting of ML pr…

2022-10-28 01:50:19 RT @wasimlorgat: You do not want to miss out on @radekosmulski's newsletter!I met Radek at @fastdotai conf. He is the most enthusiastic p…

2022-10-27 20:35:12 RT @untitled01ipynb: https://t.co/0j6HRlPd1p

2022-10-27 18:49:58 RT @untitled01ipynb: https://t.co/zYtKa0xUIC

2022-10-27 18:46:17 If you want to deepen your knowledge of the memes, here's the full videohttps://t.co/zwP3rINSCm

2022-10-27 18:44:47 I'm not sure what's worse- my entire carefully written lesson, watched at 2x, has been turned into a series of memes- the meme version might be just as educational as the original video https://t.co/UHzP8HdtAt

2022-10-27 05:38:24 RT @iScienceLuvr: This is currently the most important network in deep learning!From helping to power search for billions of users to bet…

2022-10-27 05:12:13 RT @radekosmulski: I am a bit late to the party, but the first lecture is a brilliant overview of all the components of Stable Diffusion.…

2022-10-27 01:46:27 @DSaience I will certainly do that

2022-10-26 23:27:27 @DSaience Oh and my daughter told me to say she thinks it's fun too!

2022-10-26 23:27:03 @DSaience I love it! :D

2022-10-26 19:47:38 RT @SimonWallace: Finally, a special thank you to @jeremyphoward, @HamelHusain, and @wasimlorgat for their work on nbdev. It’s an incredibl…

2022-10-26 19:47:35 RT @SimonWallace: Obiter is an open-source Python library of computational research tools. You can use the library to interact with @CanLii…

2022-10-26 17:33:26 RT @isaac_flath: What are programming things you think the data science community as a whole could use more or understand better to improve…

2022-10-26 17:15:07 RT @Robin_Tian: Amos is a new optimizer that we propose to pre-train large language models. It is more efficient and converges faster than…

2022-10-26 17:12:54 RT @alexmeshkin: I haven't disclosed this previously as I didn't want to get into the debate about the severity of COVID. Yes, I had a BA…

2022-10-26 17:10:32 RT @bot_fra: Here my attempt to apply Random Forest on a practical Industrial issue: classify Aluminium scraps boxes. I've figured out to b…

2022-10-26 17:03:24 @JayAlammar Thanks Jay!

2022-10-26 17:02:10 Interesting new podcast that discusses my tweet, questioning whether adding numpy-style broadcasting rules to APL would be a good idea https://t.co/ssWQD2Nj54

2022-10-26 16:59:04 @dyalogapl @RikedyP @a_brudz Thanks for the tip - I look forward to watching

2022-10-26 16:56:38 RT @agadmator: Magnus blunders a bishop on move 1! https://t.co/pzMFEt6idW

2022-10-24 06:02:28 RT @strickvl: This week in the @fastdotai part 2 course we started the exciting work of implementing our own deep learning library from scr…

2022-10-24 04:46:15 RT @JohnVial: How One University Researcher Used His Teaching Load To Create New Research https://t.co/SpFOWL7HYW

2022-10-24 02:47:55 @0Kiriyama @mhajabri @Fra_Pochetti @fastdotai @radekosmulski @marktenenholtz @iScienceLuvr some of it is out nowhttps://t.co/LiUu9HSflG

2022-10-24 01:35:11 RT @JACoates: Can we just talk about the reality of peer review for a second because so many scientists here seem completely oblivous to th…

2022-10-23 08:29:37 RT @TheZachMueller: Announcing preorders for Walk with @fastdotai, the missing pieces for success!If you want to take your fastai skills…

2022-10-23 02:10:10 @chrisalbon The bots will flock to reddit though, won't they?

2022-10-22 22:33:52 RT @DorsaAmir: How do you map out an ocean voyage without a map? Here’s an extremely cool example of a stick chart used by the Indigenous…

2022-10-22 21:10:26 This is a great summary of some of the main points from our first "Foundations to Stable Diffusion" lesson: https://t.co/neU9a7Txcy

2022-10-22 16:41:30 RT @TerribleMaps: The “international community” https://t.co/f9iCpZwD6H

2022-10-22 16:37:57 RT @TomLikesRobots: I've been testing the two new 1.5 checkpoints against the 1.4 versions at different steps.Faces are slightly more real…

2022-10-22 16:26:44 RT @quocleix: New open-source language model from Google AI: Flan-T5 Flan-T5 is instruction-finetuned on 1,800+ language tasks, leading…

2022-10-22 16:25:37 @_jasonwei @quocleix It got over 1k, so you gotta open something else now.

2022-10-22 16:23:33 RT @kprather88: Someone asked me why I feel one reason for the slow admission of #COVIDIsAirborne is due to a COI. Answer: Because the cha…

2022-10-22 16:18:18 RT @UrosZiv: @CDCDirector https://t.co/go5LAoPeDZ

2022-10-22 06:52:10 RT @johnowhitaker: This space is still new. Which means ideas like this, where smart people go 'this seems obvious, I wonder if anyone has…

2022-10-22 05:28:18 @kalpesh_ai @johnrobinsn Not really at the moment, no.

2022-10-22 03:52:18 I suggested to our students that they should try doing CLIP interpolation, &

2022-10-22 03:52:14 There's a trick we can use however - it's possible to use an init image as the starting point for image generation, instead of random noise. I explain the technique here:https://t.co/QpTLeALB9A

2022-10-22 03:52:13 The basic idea, illustrated below by @fastdotai student @cleavey1985, is to interpolate between the CLIP embeddings of multiple prompts, passing each to Stable Diffusion to generate images.But as you see, this leads to choppy output! https://t.co/2LKSYb3cCv

2022-10-22 03:52:10 Have a look at this amazing and beautiful video of a tree as it goes through the seasons.Can you guess how it was made? Read on! https://t.co/i5STjBI0sX

2022-10-22 02:11:47 RT @bhutanisanyam1: Some pillars of the @fastdotai community:@wasimlorgat, @radekosmulski, @HamelHusain, @jeremyphoward, @TheZachMueller…

2022-10-22 01:43:18 RT @ai_fast_track: You probably never heard of MMRotateHere are some reasons why you should be familiar with:• MMRotate is an open-so…

2022-10-21 22:27:54 RT @renegadesilicon: Friends, go read this post from my friend @Angelina_Magr on tools for data scientists — in particular the phenomenal n…

2022-10-21 22:24:57 @DienhoaT Key clarification: we subtract only a bit of the noise, then we re-run it. We do that a few times.

2022-10-21 18:59:14 RT @isaac_flath: I gave 1 of my Mom’s paintings to the Clip Interrogator. Part of the response was a name, Franklin Carmichael. I asked h…

2022-10-21 08:13:50 RT @iScienceLuvr: I'm glad to be part of the team developing the latest iteration of the @fastdotai Part 2 course, From Deep Learning Found…

2022-10-21 06:41:07 @SuryaGanguli Einstein said that relativity may have never been developed if it wasn't for Humehttps://t.co/hZX5pULQUf

2022-10-21 06:31:03 @JimGolden23 It's gonna need to be a big coffee...

2022-10-20 23:35:28 RT @runwayml: We've released public checkpoints for Stable Diffusion v1.5, which powers our Text-to-Image and Image-to-Image Tools.Code o…

2022-10-20 23:26:37 RT @ai_fast_track: Training an Object Detection Model is a very challenging task and involves tweaking so many knobsHere is an exhausti…

2022-10-20 22:17:41 RT @MilesCranmer: Tullio.jl makes pretty much any tensor operation a one-liner. It's like a grown-up version of einsum!https://t.co/6xIFz…

2022-10-20 22:16:41 RT @AliAbdaal: 24 hours ago, I tried an experiment - I tweeted a thread with 15 productivity hacks. It’s become one of my highest performin…

2022-10-20 22:14:29 RT @marktenenholtz: Kaggle just significantly upgraded their available free compute.Here's what happened:

2022-10-20 07:26:18 I just learned that the University of Queensland has reopened registrations, so if you want to join the remainder of the course live, you can do so here:https://t.co/M2AJB0tPGR https://t.co/TnHpV6wyNj

2022-10-20 07:26:04 I just learned that the University of Queensland has reopened registrations, so if you want to join the remainder of the course live, you can do so here:https://t.co/M2AJB0tPGR

2022-10-20 07:25:21 @bradneuberg Why did you order one? What do you think the benefits will be? What uses do you think will be improved with this version?

2022-10-20 06:15:35 @TheZachMueller @github Well that's quite something!

2022-10-20 06:15:25 RT @TheZachMueller: Top of @Github trending? Cool! https://t.co/4WZY5WnvAH

2022-10-20 06:00:59 @idanre1 So glad you liked it!

2022-10-20 03:18:44 @saulfp @OldManKris @hillelogram Next you'll be telling me some people have macros not called "q"?!?

2022-10-20 03:12:00 RT @fringetracker: Underrated tension is that with so few students attending live lectures - in person or even synchronously online - the j…

2022-10-20 02:34:47 @JustinMitchel It's our pleasure!

2022-10-20 02:27:09 The remainder of the course will be released in early 2023.I can't wait to see what you all create with this amazing technology!

2022-10-20 02:27:08 Lesson 10 creates a complete Diffusers pipeline from the underlying components: the VAE, unet, scheduler, and tokeniser. By putting them together manually, this gives you the flexibility to fully customise every aspect of the inference process.https://t.co/8x5gpP9se6

2022-10-20 02:27:07 Lesson 9A (by @johnowhitaker) shows what is happening behind the scenes, looking at the components and processes and how each can be modified for control over generation.https://t.co/6mKbt0u0cF

2022-10-20 02:27:06 I got a special surprise for you all...We just released the first 5.5 hours of our new course "From Deep Learning Foundations to Stable Diffusion", for free!https://t.co/LiUu9HSflG

2022-10-20 00:23:44 @tunguz @renegadesilicon @financequant @pmaymin @bennpeifert @CliffordAsness @tng_konrad @numerai There's no way to know what innovation is happening at QH funds without working at them.

2022-10-19 23:17:00 RT @iScienceLuvr: After knowing him for >

2022-10-19 18:55:25 @philkuz @yungalgorithm @ReevesCanon Google Colab is a managed service. You just click the colab you want, then hit the run button.

2022-10-19 18:41:50 @ducha_aiki @yuxiangw_cs Wow nice!

2022-10-18 21:29:03 RT @manorlaboratory: Thanks to ⁦@davidvanvalen⁩ for snapping this shot of one of my favorite slides at #CytoData2022: “why crappify?” cc…

2022-10-18 20:25:30 RT @daniela_witten: The Bias-Variance Trade-Off &

2022-10-18 18:36:09 @ducha_aiki @bhutanisanyam1 Nah he left already...

2022-10-18 18:34:15 RT @anacondainc: Today @anacondainc unveils two new products to support data literacy. Anaconda Notebooks, a 100% cloud hosted, browser…

2022-10-18 18:22:14 RT @saranormous: SF right now (especially if you’re in AI) is electric. There’s more seriousness from higher quality builders than there ha…

2022-10-18 18:20:52 @pcuenq This is great!

2022-10-18 18:18:36 @pcuenq @bhutanisanyam1 It's the amazing Glasshouse Mountains

2022-10-18 18:17:54 @ducha_aiki @bhutanisanyam1 Tell him yourself!

2022-10-18 18:16:42 RT @bhutanisanyam1: I finally met my ML hero who’s taught me everything: @jeremyphoward https://t.co/JLvdRYW1e7

2022-10-18 18:16:03 This is legendary - in a matter of hours @Buntworthy went from learning about a new paper in our course to implementing and releasing a complete notebook! https://t.co/2AodrsJcpB

2022-10-18 18:12:34 @johnsimerlink @Buntworthy @LambdaAPI There's no need to allow public connections to make that work. The ssh forwarding looks like a localhost connection as far as jupyter is concerned

2022-10-18 10:48:00 @Buntworthy I like your priorities!

2022-10-18 10:39:41 RT @Buntworthy: Got Imagic running with Stable Diffusion, it's super easy to implement, will share a notebook soon!Left: Input image, Rig…

2022-10-18 10:39:30 @Buntworthy I just finished a segment in our course covering this!

2022-10-17 21:36:34 RT @strickvl: Really enjoyed the start of part two of the @fastdotai course (taught by @jeremyphoward, @johnowhitaker, @iScienceLuvr and @w…

2022-10-17 20:44:15 RT @QasimMunye: Spent last week building a tool that creates Stable Diffusion prompts given an image. It works decently well, see below.…

2022-10-17 20:43:05 RT @NeuralBricolage: Unstable Infusion,an artificial tea ceremony https://t.co/ZGw0KWdt0L

2022-10-17 20:42:59 RT @multimodalart: With the explosion #StableDiffusion use-cases , it's impossible for diffusers maintainers to keep up But with yo…

2022-10-17 20:40:57 RT @StabilityAI: Press release: Stability AI Announces $101 Million in Funding for Open-Source Artificial Intelligence @EMostaque https://t…

2022-10-17 19:56:59 RT @osanseviero: Congratulations to Stability for the $101 million funding. Excited to see what you all build next! but "the world's fir…

2022-10-17 06:03:13 RT @kprather88: Remember there are think tanks paid to spin messages. Examples: Tobacco smoke is good for you, climate change is not due to…

2022-10-17 05:49:34 @abhinaykdn11 @ducha_aiki The 2019 part2 course where this was developed has lots of context and detail. Also discusses LSUV init by @ducha_aiki (who started this thread), a novel activation layer, and many more training stability ideas!

2022-10-17 03:26:44 RT @emollick: The essays that students turn in are about to get a lot better I just tried Moonbeam (https://t.co/tZ0gR4N32p) and it pro…

2022-10-16 23:59:10 @marktenenholtz Many thanks for the update.

2022-10-16 23:58:13 @ducha_aiki Study the distribution of grads and activations at each layer over time. fastai's "colorful dimension" plot is what you want for this.

2022-10-16 04:42:56 RT @DocSparse: @github copilot, with "public code" blocked, emits large chunks of my copyrighted code, with no attribution, no LGPL licens…

2022-10-16 04:34:29 RT @1goodtern: It is no accident or coincidence that the general public in many countries came to believe the lie that it's fine to catch a…

2022-10-16 01:03:31 @memetzgz @MDMetzger @fastdotai @johnowhitaker I'm suspicious here...

2022-10-16 01:02:10 @memetzgz @kartike_ya @fastdotai @johnowhitaker Can you tell me the painter names too please? I think I can guess most but don't want to embarrass myself!

2022-10-16 00:51:47 @marktenenholtz @rasbt @tunguz I've found transformers much better for seq2seq and especially for shorter documents (i.e shorter than the transformers max seq len).

2022-10-16 00:49:59 @marktenenholtz @rasbt @tunguz Ah sorry got it. The paper repo has notebooks so you can see there. It was a few hours on a single 1080ti IIRC.I'd love to see a transformer that can do under a day on that GPU with better results.

2022-10-08 07:24:11 @prudholu @fastdotai Yes perfect

2022-10-08 06:31:39 RT @igormp: @matdmiller @jeremyphoward @LambdaAPI The a6000 has only a few extra cores and double the vram, whereas the rtx 3090 can clock…

2022-10-08 02:46:20 RT @Brecht_declercq: Patrick McIntyre: according to Ars Technica already 90% (!) of pre-2000 computer games source code is lost. https://t.…

2022-10-08 02:29:55 RT @dwarkesh_sp: Where can I read concrete scenarios about what AGI going well look like?What is the political status of AIs? What exactl…

2022-10-08 02:18:26 @prudholu @fastdotai One minor request for this wonderful project - a little button on the left to hide and show the contents. On my screen I can't see the whole pic otherwise.

2022-10-08 02:17:27 I bought one of these with 3 * RTX 3090s last year in fact...https://t.co/FWODDSGDyN

2022-10-08 01:53:10 @iScienceLuvr Extremely sketchy

2022-10-07 23:37:53 The claim that HuggingFace is profiting from my copyrighted material that was illegally obtained by CommonCrawl, is, I believe, factually incorrect, as is the claim that this is theft.

2022-10-07 23:37:52 Sad to see some folks sinking to using questionable legal claims to bully open source projects. Open source is very disruptive to $$$-chasing monopolists.If any of you are on the receiving end of this kind of behavior, be sure to let the community know. https://t.co/1HWVSTRWTV

2022-10-07 23:31:30 @JOSourcing @EMostaque @carperai No one likes a bully.

2022-10-07 23:29:20 @pedrosorio Yeah my bad sorry!

2022-10-07 23:28:27 @ChrSzegedy Oh right my bad. Thanks for checking. Will delete.

2022-10-07 23:27:29 Very interesting comparison from @LambdaAPI. It shows that RTX 3090 is still a great choice (especially when you consider how much cheaper it is than an A100).A 24GB version would be best so you can do larger batches. https://t.co/sURksJxscv https://t.co/0A2FTzGhzo

2022-10-07 21:39:54 Yeah this isn't good... https://t.co/MB5Xeej9eI

2022-10-07 21:36:40 @fchollet I think only American English does that. Everywhere else I've been is 0-indexed.

2022-10-07 21:34:17 RT @prudholu: If you want to prepare the @fastdotai course part 2 : "From Deep Learning Foundations to Stable Diffusion" recently announced…

2022-10-07 18:59:34 RT @tylerblack32: Covid created over 7.5 million orphans (loss of a parent), and over 10.5 million kids lost a primary or secondary caregiv…

2022-10-07 18:55:19 RT @pcuenq: New feature in diffusers: pipeline callbacks!Contributed by @jamestiotio, use it for analysis, visualization or invent your…

2022-10-07 18:53:06 RT @DrGrahamLJ: To all #radiologists.Have you seen the typical #COVID pulmonary vasculopathy on CXR/CT in the last week?We've not seen th…

2022-10-07 18:47:54 RT @pcuenq: It's been two months since I started working with @huggingface. Time flies when you are having fun!I couldn't be happier. B…

2022-10-06 08:13:28 RT @heads0rtai1s: Going through the new @fastdotai course in preparation for setting up a study group, and I love the closer integration wi…

2022-10-06 08:08:04 RT @TRyanGregory: It needs to be emphasized that what we're seeing with SARS-CoV-2 variant evolution now is something new.

2022-10-06 07:58:25 RT @philipvollet: Insights from an open source influencerI'm often asked how I get my content, over the years I've built an unusual techn…

2022-10-06 04:24:51 RT @mcmullarkey: Trying a technique @jeremyphoward mentioned he uses on the @VanishingData podcast:Keep all my notebook chunks to ~10 lin…

2022-10-06 01:19:57 RT @TheZachMueller: Deep Learning from the Foundations (@fastdotai)'s lesson 1 always gives me a chuckle. We have Practical Deep Learning…

2022-10-05 21:46:38 @capetorch @wightmanr Very exciting news!

2022-10-05 19:05:42 @DSaience @njgroene See you next week!

2022-10-05 18:19:34 RT @weights_biases: What are common problems when operationalizing ML?In this interview, @sh_reya will discuss the findings of her recent…

2022-10-05 18:18:52 RT @kathrynsbach: The Census Bureau’s household pulse survey just released new #longcovid data, incl. questions about the impact of LC on d…

2022-10-05 18:10:07 @DSaience @njgroene Just do it! :D

2022-10-05 03:55:39 RT @EMostaque: AI interns who were going to go to Meta are welcome @StabilityAI We are much cooler.

2022-10-05 00:08:06 RT @Rob_Mulla: @fastdotai had a strong showing again! Many of the top teams used their deep learning framework for image classification, in…

2022-10-04 20:59:59 RT @datacascadia: Today I learned that in Python Jupyter notebooks you can pull the source code for a function with a double ?? after th…

2022-10-04 20:47:17 RT @NeuralBricolage: with all editions collected in an hour, NeuralBricolage goes mainstream can't believe the enthusiasm you all respond…

2022-10-04 20:46:36 RT @osanseviero: How fast has the @huggingface Hub grown since exactly a year ago?Models: 20k->

2022-10-04 20:43:39 @tunguz @ClementDelangue @huggingface ^^^This

2022-10-04 10:05:02 RT @NeuralBricolage: I’ve resisted being called the AI artist but not anymore because what I’m making is art, it’s creative coding and it’s…

2022-10-03 20:17:07 @dyalogapl You can do similar things in APL with `each` but it's more clunky IMO

2022-10-03 20:16:42 @dyalogapl What J does is pretty nice.In numpy et al an example of the value of broadcasting is that no special "outer product" operation is needed, since you can just do `a[None]*b[:,None]`.It's a flexible, expressive, and concise notation.

2022-10-03 07:36:11 RT @yacineMTB: my advice to anyone who wants to learn to code because they feel like their white collar field isn't going to get them to wh…

2022-10-01 06:36:50 @Smerity @michael_nielsen What happened to sentencepiece unigram encoding? Why did we all go back to BPE?And whatever happened to BPE dropout?...

2022-10-01 02:08:23 It's curious that "=" is often used to mean "is defined to be equal to" in math, instead of using e.g "≜" or ":=" (which have that exact meaning).You can't overload "=" like that when coding! https://t.co/s7WeUvc5Kt

2022-10-01 01:14:19 RT @Kate_F_Mason: The deeply awkward bro-cringe in these E*on texts is car crash level. I cannae look away. https://t.co/4NptaAWUyx

2022-10-01 01:11:14 @karpathy So didya search for your name in it yet?...

2022-10-01 01:10:28 @RobertERitz @jeremylewi @AtharvaIngle7 @GoogleColab Heh yeah no one uses colab for the UX...

2022-10-01 00:21:15 RT @simonw: Fun extract from our podcast episode about Stable Diffusion where I talk about prompt engineering and learning spells

2022-09-30 20:50:52 RT @bradneuberg: https://t.co/79tytzN9LZ

2022-09-30 20:47:36 RT @karpathy: I was asked about what AI will look like in 3 decades. Reminder: it has not even been 1 decade yet since the ImageNet moment…

2022-09-30 20:47:07 RT @TheZachMueller: This now has me mildly pondering. What would people *want* to see out of me if I did a YouTube educational stream/video…

2022-09-30 19:42:45 RT @EMostaque: Launching our medical AI community, taking name suggestions below, help @iScienceLuvr out

2022-09-30 14:39:51 RT @sama: the deadline for applying to the OpenAI residency is tomorrow.if you are an engineer or researcher from any field who wants to…

2022-09-30 14:38:57 RT @EMostaque: Thinking about the future it’s highly unlikely a company controls the most powerful AI as then it would be more powerful tha…

2022-09-30 14:30:02 RT @iScienceLuvr: ICLR 2023 (a top ML/AI conference) submissions have been released, and do you know what that means? Time for mind-blowi…

2022-09-30 14:27:54 RT @EMostaque: Just realised this after someone asked.There is only one PhD in the full-time @StabilityAI team.More are incoming but th…

2022-09-30 14:26:33 RT @AtharvaIngle7: The new change to the @GoogleColab subscription tier makes it more expensive than the majority of cloud providers.Here'…

2022-09-30 03:17:10 RT @makeitrad1: The new @GoogleColab pricing structure Is absolutely oblivious...I pay $50 a month for the Pro+ account. With todays pric…

2022-09-29 17:22:24 RT @benmschmidt: The downtick in humanities this year pushes up the ETA for when CS is larger than all humanities degrees together to one o…

2022-09-29 07:05:09 RT @khinsen: @laura_e_langdon @jeremyphoward @math_rachel That sounds great. And in my humble opinion, we need technical writers more than…

2022-09-29 04:03:15 @CMastication @laura_e_langdon @pwang @DrewFustin @teoliphant @khinsen @anacondainc It keeps getting better!https://t.co/zIWm2YJL2s

2022-09-29 03:39:11 @laura_e_langdon @math_rachel That's wonderful to hear :)

2022-09-29 03:31:37 It would be so amazing if this could make its way into an APL implementation -- it's such a powerful and convenient syntax.Hint hint @dyalogapl...

2022-09-29 03:27:53 This is one of the coolest replies I've had on Twitter for a while... :D https://t.co/O9EgByqBH7

2022-09-29 03:26:17 @laura_e_langdon @DrewFustin OMG that's awesome

2022-09-29 02:33:04 PS: I discovered this whilst creating our next course - since I really like to show students the history of the stuff we use today. Course signups are still available:https://t.co/7vWdlZGbSa

2022-09-29 02:33:03 TIL numpy's powerful broadcasting rules come from an obscure language called "Yorick"https://t.co/VzUljOuaPu

2022-09-29 02:30:56 @Basir7900 Yes.

2022-09-29 01:54:07 @bbriniotis @capetorch To clarify, I had nothing to do with making this - it was entirely @capetorch's initiative and my entire contribution was seeing it mentioned on our forums and tweeting it :)

2022-09-29 00:21:50 For a more detailed walk-thru, with some time-saving tips, see our written guide:https://t.co/S1jkxQQkB8

2022-09-29 00:21:49 5 min walkthru of creating a PR to fastai using #nbdev, by @capetorch https://t.co/OuZYuLgTxd

2022-09-28 23:36:22 Just discovered this neat trick to see everything using CPU in any WSL process:`wsl.exe --system -d ubuntu top`

2022-09-28 22:44:44 RT @NeuralBricolage: Looking at my TL it’s becoming so obvious that most of the creative work will be overtaken by AI models in a very near…

2022-09-28 19:03:23 RT @hillelogram: Ever since Strangeloop I've been thinking about end-user programming: people should write their own software, not just con…

2022-09-28 19:00:15 RT @neilkatz: Stable Diffusion ripple effect. Startup that indexes AI images from another startup gets funded just weeks after launch. I…

2022-09-28 18:51:03 RT @tszzl: there’s two views of AI products where one is like “automate a large number of low value jobs using AI” and led to eg autonomous…

2022-09-28 02:29:36 RT @jeffgilchrist: We spend a lot of time and effort to treat and filter water to make it safe to drink, it is strange that we don't invest…

2022-09-28 02:27:08 @cHHillee @ducha_aiki @Mishok2000 @jatentaki Stuff like halide, Triton, MLIR affine, and tvm attempt to resolve many of these issues.

2022-09-25 07:32:48 RT @radekosmulski: I am launching a new blog -- TabularMusings Here is the first blog post: https://t.co/J8EaO5uXUMAnd here is the tec…

2022-09-25 07:19:26 RT @JamesFBooth: Top takeaway from time in Melbourne is the food, coffee, and bar scenes are better than both London and NYC and it’s not s…

2022-09-25 01:57:35 RT @kprather88: How many times have you had COVID-19?

2022-09-25 00:30:21 RT @jedwhite: The https://t.co/H5apyk3p83 course directly led to building @andi_search! Good luck and you’ve got this!

2022-09-24 23:53:42 @NeuralBricolage Is it as "adult" and "dark" as some of the reviews suggest?

2022-09-24 23:49:38 @NeuralBricolage Oh looks like there's a book we gotta buy

2022-09-24 22:56:01 @fastdotai Oh and here's the author's twitter handle: @Miguel_Laussman

2022-09-24 22:54:11 @NeuralBricolage She likes it - or at least she says "I think so" when asked :)She wants to know "what's Moomin?"

2022-09-24 22:13:12 RT @jeffistyping: I just put up a @huggingface space to use @OpenAI's Whisper model for YouTube videos transcription. Just enter a YT l…

2022-09-24 22:10:55 @giffmana @juanbuhler @fastdotai @midjourney Makes sense

2022-09-24 22:09:08 @NeuralBricolage I was just scrolling through some of your work with my daughter and she said she thinks they're really cool! :D

2022-09-24 22:07:24 @NeuralBricolage Can't wait to see what you build Helena.

2022-09-24 22:06:58 @fchollet Me too. And it's kept getting better, so far.

2022-09-24 22:05:52 @capetorch My pleasure!

2022-09-24 20:38:57 @juanbuhler @giffmana @fastdotai @midjourney Other than cropping out the bottom few pixels, how would you go about that? Correcting for the prompt differences seems quite tricky!

2022-09-24 19:16:50 @fastdotai @midjourney Oops nevermind - it's probably just finding the DALL-E watermark!https://t.co/VrSMIeM2qP

2022-09-24 19:16:12 @giffmana @fastdotai @midjourney Ah yes good point!

2022-09-24 18:33:55 I'm impressed with this @fastdotai student project that managed to create a dataset and model from scratch to recognise 800 bird species.https://t.co/qOYGlVckNR https://t.co/xpQSH2A6hD

2022-09-24 18:19:59 @zehavoc @vijayabhaskarj @simonw The more education research I read the more I think education is complex and I'm skeptical of people who think it's all very obvious.

2022-09-24 18:18:26 @zehavoc @vijayabhaskarj @simonw I'm lost. What is "the" skill here? Don't people learn lots of skills? Doesn't automating the bits that aren't the learning focus mean learning should improve, based on cognitive learning theory.I seem to have waded into some emotionally charged area so I'll bow out.

2022-09-24 17:38:15 @zehavoc @simonw Are you saying kids should spend more time learning times tables, or more time using a calculator?Based on your comment it seems you think there's some obvious best approach, but I'm not sure which, or why.

2022-09-24 09:53:00 @etiennefd @nearcyan If you've finished https://t.co/GEOZuodrZj part 1, then you can do part 2! And that will be full of latest research stuff which will keep you busy for years, with a whole community of alumni to work with...

2022-09-24 09:50:57 RT @nearcyan: Around once a week, a smart person will confidently tell me that they are "not smart enough to do anything with AI"I *stron…

2022-09-24 09:41:21 RT @YouAreLobbyLud: I have donated to @marwa_zaatari legal fundThis pandemic is a pandemic of lying, an assault on truth and science by t…

2022-09-24 09:39:42 RT @EricTopol: Two new independent reports, one Norway nationwide, and the other US-based, indicate increased risk of Type 1 (autoimmune) d…

2022-09-24 02:41:49 @RobertERitz @dwaynebradley70 @simonw Written English composition sounds like something you'd often want to use AI to help with in practice.

2022-09-24 01:44:56 @jonathanrlarkin Awesome!

2022-09-24 01:44:01 @dwaynebradley70 @simonw There's a problem with the assessment if you can turn in good work without understanding the content.

2022-09-23 23:22:20 @simonw I certainly understand this point of view. But from the student's point of view, knowing how to use these tools and work with them effectively seems important. Why learn to write the old way?Creating good content with AI still requires plenty of skill, practice, and knowledge! https://t.co/7thtwDtrgK

2022-09-23 23:14:37 @Randy_Au @HamelHusain @TheZachMueller The film is working well for our daughter - it seems to be thick enough to handle the bit of bumpiness in our regular indoor paint

2022-09-23 23:13:17 @MathYouF No, it's not like a traditional uni course. Instead, you're encouraged to do projects during the course &

2022-09-23 23:08:06 @HamelHusain @TheZachMueller Or the film?https://t.co/PEReXNQQFc

2022-09-23 23:06:02 Here's part 1 of our course, which you can complete now for free, and will give you a solid grounding in the key ideas of deep learning:https://t.co/nzv7pe1R4i

2022-09-23 23:06:01 Our announcement last week of the upcoming course got a huge amount of interest both here on Twitter and on Hacker News.https://t.co/icBXjx6Tui

2022-09-23 23:06:00 Big news: signups for our new course "From Deep Learning Foundations to Stable Diffusion" are now available.(And if you're an open source contributor, you can apply for a scholarship to join for free!)https://t.co/7vWdlZXeUa

2022-09-23 22:19:37 @matdmiller @TheZachMueller @HamelHusain @fastdotai Hey Australia has good skiing too!

2022-09-23 21:09:53 @HamelHusain @TheZachMueller @fastdotai Yeah why would you leave?

2022-09-23 21:09:34 @TheZachMueller @fastdotai Or... go to bed earlier? :O

2022-09-23 21:08:59 @TheZachMueller @johnowhitaker @HamelHusain Oh dear. Did that sting?https://t.co/NncaMofdlf https://t.co/wBxLIBewPm

2022-09-23 07:05:39 @swar_ja @Suhail It will get faster

2022-09-23 06:59:57 RT @VictorianCHO: Victoria has had no MPX cases for a few weeks now, having had significantly early growth in cases. Lo and behold, those p…

2022-09-23 06:57:45 @TheZachMueller @HamelHusain Yeah sorry zach to clarify I think he's talking about *young* people...

2022-09-23 05:55:57 @anton_lozhkov @keturn_hax @johnowhitaker Sorry i guess I should have said VAE not VQGAN...

2022-09-23 00:28:47 @mat_kelcey @nova77t @karpathy Lots of interesting stuff in hasktorch

2022-09-23 00:28:03 RT @mat_kelcey: #sigh https://t.co/uF9DHRnFcn

2022-09-22 23:33:19 @EMostaque SOTA models but some data?

2022-09-22 23:30:15 RT @sharifshameem: Introducing Lexica – a search engine for AI-generated images and prompts.Every image has a prompt and seed, so you can…

2022-09-22 21:57:48 @anton_lozhkov @keturn_hax @johnowhitaker pointed out on the https://t.co/GEOZunWoXj Discord a few days ago that this is at least partly because the VQGAN is patch-based, so it does have some geometry hard-coded into it.

2022-09-22 21:36:31 @anton_lozhkov @keturn_hax I... don't understand how this is possible. This seems to suggest that the VAE features have room to improve quite a bit?

2022-09-22 21:32:16 RT @karpathy: Reading through OpenAI Whisper paper https://t.co/3PmWvQNCFs some notes: https://t.co/QVeqaGVvsV

2022-09-22 21:30:08 @Thich_5 @willmcgugan A ChainMap isn't a dict, it just behaves like one -- but lookups are slower.Try doing the same thing, but with `dict(Chainmap(*dicts))`

2022-09-22 09:13:41 @ClementDelangue @sundarpichai @huggingface Sweet!

2022-09-22 01:02:42 RT @radekosmulski: I love peering into the future and watching the GTC keynote is my favorite way to do so!Here are my highlights from th…

2022-09-22 01:00:13 @willmcgugan Post Visidata to Textual. How hard can it be?...

2022-09-21 19:22:40 RT @vishvdeep18: Sharing free/inexpensive material for learning data science and ML (and not jus the basics).Combined, their value is eas…

2022-09-21 07:16:03 RT @sudalairajkumar: @labmlai Deep Leaning Paper Implementations - is a good collection of popular deep learning papers implemented in PyTo…

2022-09-21 07:12:12 @arxivabs Good bot

2022-09-21 06:18:37 Oh wow turns out there's a *lot* of papers doing this kind of thing...https://t.co/WXsnPh6C0A https://t.co/XZlW4IL5tw

2022-09-21 06:08:55 @ron_itelman @simonw I have published many! Here's one from our book:https://t.co/Sx9gYWowIH

2022-09-21 06:07:44 Whatever happened to Powernorm?It's an approach to improving batchnorm that's similar to something we introduced in our 2019 course - using running statistics instead of per-batch. Helps in NLP, and probably in CV too.Did it get forgotten?https://t.co/XAQke6SIOQ

2022-09-21 03:39:29 @simonw I use pandas and sklearn random forests

2022-09-20 19:13:23 @nagaraj_arvind Yes, time permitting

2022-09-19 00:56:39 RT @YouAreLobbyLud: “Did COVID harm my immune system? Am I now at risk of other infectious diseases?”Yes and yes.https://t.co/u6ZMXXVFFa

2022-09-19 00:51:40 I love this new article from @wasimlorgat -- the first "how to write great notebooks" tutorial I've seen. Such an important topic! https://t.co/m6SxMJykXP

2022-09-18 23:10:11 @riccardo_grin Thanks!

2022-09-18 09:28:26 RT @gun_toucher: lmao. the kids are killing the reply gif industry https://t.co/yfRa8NZsMj

2022-09-18 03:20:16 Tldr: when you inhale, only a tiny percentage of your inhaled air comes from inside your mask, because the mask has a small volume and you breath a large volume.

2022-09-18 03:18:11 @YouAreLobbyLud https://t.co/Mv0yVA9lWz

2022-09-18 03:17:09 Ffs I explained why this morinic argument doesn't work 2 years ago. And an Australian university professor is still spouting this nonsense.Here's why it's wrong:https://t.co/Gvd5rfx8zy https://t.co/ZTjtx2x6jR

2022-09-18 03:03:16 RT @thesephist: Good tools admit virtuosity — they have low floors and high ceilings, and are open to beginners but support mastery, so tha…

2022-09-18 03:02:47 RT @Stephen_Parker: FamilyToday I would like to share this series of #dalle2 AI-generated images. (thread for detail and process) https:/…

2022-09-18 02:39:17 RT @j_b_kennedy: Good piece here by an immunologist whose heart bore the brunt of a covid infection. It’s a pity that experts like this guy…

2022-09-17 22:16:08 @gordic_aleksa Maybe this will help - Peter Sellars doing a terrible French Accent in The Pink Pantherhttps://t.co/VYzr5mO1Cx

2022-09-17 20:33:35 @gordic_aleksa @WhiteRabbitRC I'm enjoying your video - thanks a lot for making it!But you gotta work on your pronunciation of "imagenette". :O See the README for handy tips https://t.co/p3r7pBCyLb

2022-09-17 20:08:49 @EMostaque The first person I know of to do brilliant deep learning generated art is @NeuralBricolage. She's been doing this for years and her work is amazing.In the art-world her genius is appreciated by those in the know, but she deserves a much bigger audience IMO.

2022-09-17 19:39:37 Here's what the output looks like https://t.co/YvPe1iQ7YQ

2022-09-17 19:39:36 I've got a few git repos which I like to pull each day in parallel and then check their status. Here's a little script that does that -- just put the repo names in ~/git/repos and run it. Only those with non-empty status are printedhttps://t.co/GQ6BQ1YuTR

2022-09-17 19:32:42 @openroomxyz @EMostaque @fastdotai Yes and yes.

2022-09-17 19:30:17 RT @EMostaque: Who are the under-appreciated geniuses in AI or complex systems to make the world better who could do with more support? And…

2022-09-17 09:50:45 RT @Warvito: New preprint! We used Diffusion Models (same one from #dalle2 and @StableDiffusion ) to generate 3D MRI images of the brain c…

2022-09-17 02:36:33 Heh - @realGeorgeHotz calls @EMostaque "a boss" in his live coding stream :Dhttps://t.co/rwwIbpmuS6

2022-09-16 21:34:39 @danielrock @math_rachel Thanks Daniel!

2022-09-16 21:09:54 @LukasGalke @quarto_pub Yes it's a great replacement for jekyll. In fact I replaced jekyll with quarto for https://t.co/GEOZunWoXj just yesterday!

2022-09-16 18:02:54 RT @simonw: Has anyone seen a good list of production software that's been deployed against OpenAI GPT-3 APIs so far?

2022-09-16 17:33:41 @simonw @StabilityAI Although doing a refresh with the latest version of the course would probably be helpful, if you had the time and inclination.

2022-09-16 17:33:01 @simonw @StabilityAI Yes that refers to Practical Deep Learning, at https://t.co/aQsW59Ylt6. Sorry it's a bit unclear! I'll try to get the info clarified asap.@simonw your knowledge from doing that course a few years ago should be sufficient. The basics haven't really changed since then.

2022-09-13 05:47:28 RT @marksaroufim: I am proud to be be part of an ML community that has open sourced its code and also what matters most: its governance. Py…

2022-09-13 04:05:30 @Sheikheddy @DrMJoyner @tdietterich @DavidEpstein @nxthompson @GaryMarcus @cragcrest @geoffreyhinton @ACasadevall1 Email has been the loser so far, since it's hard or impossible for regular people to run their own email server nowadays and get good deliverability, since spam protection relies on sender trust signals with huge false positives for spam.(I run one myself and deal with this)

2022-09-13 04:03:33 @Sheikheddy @DrMJoyner @tdietterich @DavidEpstein @nxthompson @GaryMarcus @cragcrest @geoffreyhinton @ACasadevall1 I mean: I don't think it's possible to create an anti-spam algorithm that works in the most general case, where both attackers and defenders have access to substantial resources.More generally: I don't think it's possible to classify bot-written content in the long-run.

2022-09-13 01:52:29 @DrMJoyner @tdietterich @Sheikheddy @DavidEpstein @nxthompson @GaryMarcus @cragcrest @geoffreyhinton @ACasadevall1 Having said that, neither spammers nor anti-spammers on the whole currently use modern AI approaches. But if they did, I'd expect the spammers to win.

2022-09-13 01:51:39 @DrMJoyner @tdietterich @Sheikheddy @DavidEpstein @nxthompson @GaryMarcus @cragcrest @geoffreyhinton @ACasadevall1 In general, anti-spam is in the hardest category of ML classifiers, because it's adversarial. If you make the results of your anti-spam algorithm public (which you do, every time you put a message in the spam folder) then you provide a training corpus for spammers

2022-09-13 01:44:26 RT @dtmooreeditor: So there's a particular quirk of English grammar that I've always found quite endearing: the exocentric verb-noun compou…

2022-09-13 01:31:55 RT @wasimlorgat: “Being close to the interpreter” is why I use notebooks for everything

2022-09-12 22:57:04 RT @edyong209: I wrote about “brain fog”—one of the most common &

2022-09-12 21:27:13 RT @bethlinas: As always, always read @edyong209! The discussions and research on #COVID #brainfog is on going and timely! https://t.co/pp2…

2022-09-12 21:12:45 My friend @markpollock, one of the world's most inspiring people, is running @TheRunInTheDark, where runners can choose a 5k or 10k and raise money to cure paralysis in our lifetime. Join live in one of their many cities or take part virtually.Register: https://t.co/RrEJsq3Src

2022-09-12 21:09:23 @DienhoaT I'm looking forward to seeing your writeup of this! Is there a notebook work-in-progress we could look at?

2022-09-11 23:51:27 RT @EMostaque: If you want to understand how hard AI alignment is realise1. Organisations are basically slow dumb AI that provisions huma…

2022-09-11 16:46:53 Great to see #nbdev getting discussed in the python community https://t.co/uWZ9SCHau2

2022-09-11 07:01:35 @SudheerNotes @StableDiffusion @EMostaque That's great to hear!

2022-09-11 06:08:37 How to change your life - good advice from @StableDiffusion founder @EMostaque: https://t.co/x3yaCqv3bp

2022-09-11 04:22:01 RT @zhuowei: youtube-dl has a JavaScript interpreter written in pure Python in 870 lines of code.https://t.co/5ZLZEhFRem... wow.

2022-09-10 23:49:06 RT @ceph_x: It happened: Open AI surpassed OpenAI. This is a breakthrough for builders everywhere, with many more milestones to go.Let’…

2022-09-10 22:01:30 RT @paulisci: A List of Things People Blamed on Radio

2022-09-10 07:38:09 @jeblad And I didn't say it's what it said.I provided context to help interpret what it said.

2022-09-10 06:51:24 RT @ARC_Mount_Sinai: Most clinicians think #LongCOVID is a complete mystery!It's NOT, Here's what we mean...

2022-09-10 06:03:48 40% of the US population is boosted. https://t.co/PycOxFs3OD

2022-09-10 01:38:50 @chrisalbon ...then the CEO got millions in bonuseshttps://t.co/WvhWz5xThT

2022-09-10 01:38:30 @chrisalbon ...then gave the customers $50 as an apology for screwing uphttps://t.co/RmC0Xvnxoo

2022-09-10 01:38:06 @chrisalbon Here in Australia the main airline fired their employees during COVID and then when flights restarted everything predictably fell apart.The airline CEO blamed the customers. https://t.co/rm9SqXYfQi

2022-09-10 01:35:19 @adamboxer1 You could tweet about how everything is just super great, maybe?

2022-09-10 01:33:03 @chrisalbon Yeah I always did 2 hours for my international flights. Never had a problem. Often turned up 90 mins before actually and that was always fine too. This was mainly Australia, SE Asia, and Europe though. LAX was always a disaster zone.

2022-09-10 00:40:14 RT @ToshiAkima: 2. Today, China has launched its biotech version of Vostok 1 in the form of the first inhaled/intranasal COVID vaccine in t…

2022-09-10 00:34:33 RT @runwayml: Make any idea real. Just write it. Text to video, coming soon to Runway.Sign up for early access: https://t.co/ekldoIshdw…

2022-09-09 23:00:37 RT @ThorstenBeier: The first prototype of #PicoMamba is working! Running #mamba / #conda in the browser. Install #emscriptenforge / #condaf…

2022-09-09 22:11:34 RT @fede_gr: Congrats, looking sharp! We are also big fans of #quarto. We rewrote our docs (based on @jeremyphoward's nbdev2), and the expe…

2022-09-09 01:29:19 RT @wasimlorgat: Which web server would you rather use?

2022-09-09 00:38:03 RT @wcrichton: I got the data. I crunched the numbers. The most diverse programming language community? Maybe Rust, right? Not even close.…

2022-09-09 00:35:08 RT @CMastication: The cruelest joke of Jupyter is the name is a portmanteau of Julia, Python, &

2022-09-09 00:24:08 @dathabase @math_rachel @radekosmulski Looks like you haven't made your kaggle notebook public @dathabase

2022-09-08 21:55:23 @pcuenq That's awesome Petro! :D

2022-09-08 21:55:01 RT @pcuenq: You can now run Stable Diffusion on your Apple Silicon Mac (M1 or M2)! Simply use .to("mps") (instead of "cuda") to run the…

2022-09-08 05:51:40 @Ted_Underwood @dbonneville @dmvaldman There's a good chance the pace of progress will accelerate over the next couple of years...

2022-09-07 22:50:51 RT @TRyanGregory: Imagine that China does manage to keep COVID cases very low, and then has a successful rollout of a nasal vaccine that bl…

2022-09-07 22:39:02 Fantastic #nbdev explanation of deep learning object detection by @DienhoaT, based on the @fastdotai course.If you're interested in learning what *really* happens in object detection, be sure to read this!https://t.co/WRr5HMhcOi

2022-09-07 14:33:55 RT @R_H_Ebright: COVID: summary of lab-origin hypothesis:1) Pandemic caused by a bat SARS-like coronavirus emerged in Wuhan--a city 1,000…

2022-09-07 14:10:22 RT @images_ai: Fixing famous paintings with AI ()First up, the Mona Lisa! Honestly, Da Vinci did a pretty sloppy job on this one by toda…

2022-09-07 09:57:24 RT @RageSheen: Covid reminds me of asbestos. Hardly anybody showed ill effects from asbestos, when it was introduced as a wonder material.…

2022-09-07 09:55:06 RT @CO2Guerrillas: There is so much we can do to improve the quality of our indoor air. We just aren't trying.@YouAreLobbyLud @CO2Radical…

2022-09-07 02:56:47 RT @dgurdasani1: New letter out in Jama Paeds estimating orphanhood of children from COVID at 7.5 million &

2022-09-07 02:34:42 @YouAreLobbyLud Yup.https://t.co/FPmi54vRc1

2022-09-07 02:33:55 I'm not "risk averse". I'm a motorcyclist, backcountry skier, and 5x entrepreneur. I've been the key driver of risk pricing decisions in multi-billion dollar insurance portfolios.I'm an experienced risk manager. And COVID is too big a risk to do nothing about. https://t.co/IP5XZ6hxes

2022-09-07 02:28:08 RT @DavidLKeating: Victoria, Australia info on the left and CDC USA on the right. The CDC needs to learn how to communicate.

2022-09-07 00:04:10 RT @michrajah: Pleased to be investigating the diagnosis, treatment and prevention of long COVID. We want to hear from you including those…

2022-09-06 22:58:24 RT @math_rachel: Prior to 2020, @jeremyphoward and I never expected to end up homeschooling, but now we plan to do it long-term. Here is a…

2022-09-06 20:52:49 RT @drkerrynphelps: As a doctor involved in health promotion for decades, watching the unravelling of all the progress in chronic disease p…

2022-09-06 20:50:33 RT @Globalbiosec: If you could book a flight, which would u choose

2022-09-06 20:46:59 @f_j_j_ @simonw The fact that most prompts don't make obvious the biases learned by the model isn't something I find comforting

2022-09-06 20:44:03 RT @simonw: Inspired by finally reading the classic "On the Dangers of Stochastic Parrots" paper, I decided to see how easy it was to get G…

2022-09-06 20:29:03 @DrMoAssoum Good luck on your Twitter journey!

2022-09-06 06:19:18 So happy to see stuff like this getting published using #nbdev - it's exactly what we were hoping to see when we built it! :D https://t.co/cL512XxFD5

2022-09-06 02:43:11 RT @radekosmulski: An important and well-researched post by @math_rachel that just surfaced in my email. I now apply myself to work less…

2022-09-05 04:29:32 RT @psvbluemts: Irrespective of how long isolation is mandated, employers are breaching WHS duties if they allow people back to work while…

2022-09-05 04:28:32 RT @maosbot: Almost no-one I know with #LongCovid has said they have Long Covid publicly or on social media. I don't blame them at all.Yo…

2022-09-04 05:34:23 RT @andfanilo: To my fellow Content Creator Friends, I found this beautiful comic about creativity and taking breaks by @Oatmeal I'd like t…

2022-09-04 05:32:26 RT @VirginiaBuysse: Well, it turns out that students in states that stayed remote longest DID NOT have lower test scores than students from…

2022-09-04 05:32:21 @StephTaitWrites In Australia the state that was locked down by far the most (Victoria) had the best test results.

2022-09-04 03:10:40 @AlSweigart This is a wonderful place to start that journey:https://t.co/lvc2HLO0lX

2022-09-04 03:06:53 @andy_matuschak IMO it's helpful, from time to time, to "release" a thing. Even although it's an arbitrary collection of pieces you've built over time, when you "launch" that collection with a name/version, you give your audience a reason to look at it, and yourself a reason to celebrate!

2022-09-03 20:41:56 RT @DSaience: The First-Time Machine Learning Playbook.(Read this if you want to efficiently learn machine learning, avoid frustration fr…

2022-09-03 05:06:41 @Wazomec @willmcgugan After reading about them there and trying them out, you could then come back here and ask any questions you have that you weren't able to find answers to despite doing your best to do so.

2022-09-03 05:05:55 @Wazomec @willmcgugan The most important part of learning IMO is being good at finding info. The best source of info is google. Here's the google search you needed to do to answer your own question:https://t.co/Wgum65d4eG

2022-09-03 02:39:21 RT @willmcgugan: @sadneurons Hard to say. GHtop inspired me to write Textual.

2022-09-02 22:27:47 RT @weights_biases: What are the best vision models for fine-tuning?@capetorch teamed up with @jeremyphoward and has written an exhaustiv…

2022-09-02 01:00:02 RT @danaparish: My friend just sent me this devastating clip from @WesElyMD. He’s seeing terrifying symptoms (even in fully vaxxed &

2022-09-02 00:57:13 RT @EconoBen: https://t.co/JyMbWaEWC1As an organizer of this event, I think you should probably sign up! Some pretty smart speakers wil…

2022-09-01 23:27:11 RT @suvash: Getting started with the @huggingface Transformers book finally, &

2022-09-01 20:59:07 RT @machsci: Can’t unsee: I know how you pronounce SQL based on whether you write “a SQL query” or “an SQL query” in your tweets

2022-09-01 20:53:57 @matdmiller @hillelogram I much prefer windows OS (prior to win11) and surface hardware, but love the speed of the M2. I have a Linux server too and it's fine, but I don't love it.

2022-09-01 20:50:04 RT @HamelHusain: We have been adding more docs and features to nbdev at a rapid pace! If you haven’t tried nbdev yet, now is a fantastic t…

2022-09-01 16:01:35 RT @tunguz: If you guys haven’t done so already, please consider taking the @Kaggle DS and ML survey. It’s one of the best insights into wh…

2022-09-01 15:51:30 RT @hillelogram: One of the reasons I stick to Windows over Mac/Linux is that I've personally found it to be a much better experience for d…

2022-09-01 03:47:15 @Thoracicus @CatherineOrr73 @erinrileyau Quite. I'm appalled.

2022-09-01 00:11:18 RT @lucasgvazquez: After 7 years of using Linux as my main OS, I made the switch to Windows. And I have to say, I couldn’t be more satisfie…

2022-08-31 23:39:40 RT @nearcyan: user wins first place in a fine arts competition using Midjourney/Stable Diffusion https://t.co/tnv1tVYiKU

2022-08-31 23:38:36 @skaliy3 Awesome job!

2022-08-31 23:38:15 RT @skaliy3: Finally! We have released version 0.1 of fastMONAI, a low-code Python-based deep learning library for medical imaging built on…

2022-08-31 23:32:24 @HamelHusain @_ScottCondron @weights_biases Now I gotta know how you figured this out...

2022-08-31 23:31:55 RT @_ScottCondron: I've been working at @weights_biases for over a year and I've just had my mind blown by a detail of the W&

2022-08-31 03:54:48 RT @ariel_koko: I am leaving @Google this week due to retaliation &

2022-08-30 07:26:41 RT @radekosmulski: thisAdditionally, git and the sw dev cycle (feature/bug PR merge) are such a valuable skill to haveIt will p…

2022-08-30 07:21:41 RT @wasimlorgat: How do you customise your Jupyter Notebook/Lab?Any custom keyboard shortcuts?Custom css?Extensions?Other?

2022-08-30 02:57:48 RT @organichemusic: I would LOVE to work remotely for a #LongCovid clinic. However, I don't know of any clinics hiring remote pharmacists…

2022-08-29 19:40:16 RT @I_Am_NickBloom: The shift to WFH is the largest shock to labor markets in decades. Pre-pandemic WFH was trending towards 5% of days by…

2022-08-29 19:25:47 @androsfenollosa @th3h4dyy @killian_arts @berkeleygfx @htmx_org @builtwithdjango https://t.co/UhMvVIUrlf

2022-08-29 09:40:27 RT @shrimpzoo: Today amazon gave me a verbal warning about going to the bathroom while not on break. They have a new rule they don’t allow…

2022-08-29 03:08:56 RT @GidMK: If you're skeptical of mask benefits but ABSOLUTELY SURE they cause harm then imo you're more politician than scientist

2022-08-28 23:07:55 @androsfenollosa @th3h4dyy @killian_arts @berkeleygfx @htmx_org @builtwithdjango Yes you can.

2022-08-28 21:17:05 RT @willmcgugan: I don’t have a degree either.

2022-08-28 20:06:46 RT @DGlaucomflecken: Big news in academic publishing! https://t.co/5C5Rsq0SLC

2022-08-28 05:22:41 RT @drclairetaylor: 1/ Thank you. I have been saying over and over that #Covid_19 is a thrombotic vasculitis. Now can we start counting the…

2022-08-27 20:16:20 RT @HamelHusain: Personal announcement: I'm over the moon to share that I'll be exploring commercial opportunities around the @fastdotai…

2022-08-26 23:38:18 @NSWurbanstuff @PerpetualWinger Reading is hardhttps://t.co/tmCq2G00Em

2022-08-26 22:33:11 RT @DavidLKeating: This is super exciting -- cheap tools to help people get a great fit for their high-filtration mask.And a mask that wi…

2022-08-26 20:37:13 RT @AmerAcadPeds: As school resumes with changing COVID protocols, here's what parents should know about masking and their children's devel…

2022-08-26 20:37:02 RT @AmerAcadPeds: If caregivers are wearing masks, does that harm kids’ language development? No. There is no evidence of this. And we know…

2022-08-26 20:36:24 RT @willmcgugan: I use this one little "trick" in Textual, and Rich. Generators are great when you want to return a number of conditional…

2022-08-26 18:28:57 RT @micahgoldblum: We should value practicality and intuitiveness over novelty. Novelty often means an idea is so complicated that the rea…

2022-08-26 18:23:10 RT @psvbluemts: @YouAreLobbyLud I'm seeing many reports of hospitals across Australia asking people to take off their N95 and take a surgic…

2022-08-26 06:08:39 @clairlemon @elonmusk @Quillette @joelkotkin @demographia I enjoyed that article - thanks for sharing. It doesn't really seem to support @Elon's contention though does it?

2022-08-26 05:21:40 RT @iScienceLuvr: Research in AI is surprisingly more accessible to people with different backgrounds compared to other fields. Anyone (w…

2022-08-26 02:17:33 @cosmic_mar Skepticism is most welcome - very keen to have skeptics try this out and report back, either on success or failure!

2022-08-26 02:16:02 @wasimlorgat @hrbrmstr It's odd to consider it a Python problem. The "R" in JuPyteR refers to #RStats. I don't think it's the case that the R community no longer cares about notebooks, do they?

2022-08-26 01:32:49 @gideonmann @ProjectJupyter @github nbdev already does unit tests :) 100% of https://t.co/GEOZunWoXj's notebooks are tested with nbdev and fastcore.testhttps://t.co/7YKtyp7sY2

2022-08-26 00:46:13 @HamelHusain @CMastication https://t.co/9g2vUBvLVl

2022-08-26 00:45:43 @HamelHusain @CMastication https://t.co/POMF6MEMfr

2022-08-26 00:27:43 @CMastication Strewth mate, we've been flat out like a lizard drinking.

2022-08-26 00:05:11 Big thanks to all the amazing people that have helped to get this point, including @StasBekman, @GuggerSylvain, @wasimlorgat, @HamelHusain, the @fastdotai community, &

2022-08-26 00:05:10 For the details of the underlying algorithms and software hooks we used to solve the Jupyter+git problem, read our full blog post.https://t.co/I3CeS3vRY5

2022-08-26 00:05:09 Conflicts are common in notebooks, because Jupyter changes stuff every time you run a notebook, like execution counts, notebook metadata, and more.These changes make git diffs verbose, create conflicts, complicate code reviews, and make git repos more bulky than necessary.

2022-08-26 00:05:08 The reason git breaks @ProjectJupyter stems from an incompatibility between the format Jupyter notebooks use (JSON) and the format that git conflict markers assume. git merges create invalid JSON, and therefore Jupyter can’t open the notebook. https://t.co/i1HFhRHeao

2022-08-26 00:05:06 Notebooks are a powerful tool-an ideal environment for exploring data and code, writing programs, and documenting the results.But when collaborating with, this goes up in smoke, because git makes notebooks unusable. Literally. https://t.co/cneZcZvvex

2022-08-26 00:05:05 Have you ever tried to use @ProjectJupyter with git/@github? Did it drive you crazy because nothing worked right?It drove *us* crazy, so we fixed it.1/https://t.co/I3CeS3vRY5

2022-08-25 23:55:59 RT @CMastication: A bunch of the innovation around data science computing in the last 15 years has been around paving the on ramps and remo…

2022-08-25 19:34:16 RT @vanstriendaniel: Why teach #machinelearning using @fastdotai? Picking up the earlier thread on our new @ProgHist lesson on computer…

2022-08-25 19:33:30 RT @diffusionbot: #stablediffusion is an amazing way to generate images! I am here to help make it easier. Simply describe the image you wa…

2022-08-25 11:50:56 RT @tamalama67: Went to the bank. The lineup was 15 people deep so I braced myself being the only masked person and knowing it would take a…

2022-08-25 08:02:51 RT @DrGrahamLJ: Please can all papers on #COVID stop referring to the lung disease as ‘pneumonia’. It is a pulmonary vasculopathy (disease…

2022-08-20 11:28:38 RT @vanstriendaniel: The first lesson of the fastai course will already give you the skills to create a valuable tool for GLAM applications…

2022-08-20 11:27:38 RT @vanstriendaniel: It turns out I can still get excited about using machine learning for #GLAM data at 1.30 am...During this great di…

2022-08-20 08:54:42 RT @calirunnerdoc: The fact that shares of published @Nature articles on implications for Srs2 in pregnancy now being labeled for misinform…

2022-08-20 08:53:04 RT @BurnetInstitute: Despite higher immunity, higher vaccination rates and effective therapies, we've just had our worst COVID wave yet @C…

2022-08-20 06:20:41 RT @j_ckao: How did we ever get into the enterprise bloat state of an 8GB mem @duckdb instance running laps around a 12 * 64 GB mem node @A…

2022-08-19 23:50:27 RT @HamelHusain: Here is my lightning talk about our recent new release of #nbdev built with @quarto_pub https://t.co/pHHz72Ylmr

2022-08-19 04:07:12 @belinda_brucki @psvbluemts @taipan168 On what basis are you making that assertion? Singing is one of the most aerosolising activities there is, and is generally happening inside where aerosols build up.Surely it's worth investing in clean air so these activities can be much safer?

2022-08-19 03:58:38 RT @exceedhergrasp1: Apparently, @TwitterSupport is freezing people's accounts for tweeting research studies, now. Hey, @Medscape, were y…

2022-08-19 01:58:14 @willmcgugan Have you found a good way to distribute it?

2022-08-19 01:53:56 @vboykis @chrisalbon I broke the flow there.Sorry.

2022-08-19 00:35:48 @soldni You can't un-see it :(

2022-08-19 00:35:19 @tomgoldsteincs It goes back further than that! We pointed out in 2019 that GANs are not generally a great idea, and Jason Antic went on to create "NoGAN" in @DeOldify.https://t.co/TWbBp6ciwu

2022-08-19 00:29:02 PS: don't look at the ~1500 lines of code it takes to implement them, unless you're ready to suffer https://t.co/OM0uyBaC3f

2022-08-19 00:29:01 There's so very many things I dislike about dataclasses.This is one of them: https://t.co/gkNpDo0lLB

2022-08-18 19:16:14 RT @espejelomar: Are you looking for datasets to test your knowledge gained through the new @fastdotai course and a good community to share…

2022-08-18 10:13:53 @iScienceLuvr @StabilityAI Fab news Tanishq!

2022-08-18 04:45:24 RT @radekosmulski: How did you learn Python?Would you recommend the method that you used to newcomers?I'll start.I learned Python from…

2022-08-17 17:10:30 @chrisoffner3d @karpathy Well... I started with neural nets myself about 30 years ago, and that wasn't the early days either.I really meant "the early days of @fastdotai".

2022-08-17 17:07:01 RT @fede_gr: StatsForecast Exponential Smoothing (ETS) is 32% more accurate and over 100 times faster than NeuralProphetWe hope this e…

2022-08-17 17:05:06 RT @goodside: A novel(?), powerful method of complete-program synthesis using GPT-3, generalizing the “format trick” @BorisMPower showed me…

2022-08-17 04:11:10 @__mharrison__ pip install -e .

2022-08-17 04:09:18 @karpathy Awesome Andrej - wonderful to have you back in teaching mode. Your CS231n was such an absolute stand-out resource -- that and Andrew Ng's Coursera course were the two things we recommended to all our students in the early days!

2022-08-17 04:07:11 RT @karpathy: !!!! Ok I recorded a (new!) 2h25m lecture on "The spelled-out intro to neural networks and backpropagation: building microgra…

2022-08-16 22:05:52 RT @Billius27: In advance of a potential fall/winter seasonal set of Omicron waves, this thread will begin to share facts, advice and rem…

2022-08-16 16:19:19 @moultano @images_ai @arvalis Kinda, but it's definitely not "literally learning how to recreate our pixels".

2022-08-16 16:17:40 @dbonneville @arvalis I think some apologies and mea culpas are needed all around.I mean, seriously, wtf is a "moralist"?!? Caring about ethics is bad now?

2022-08-16 01:54:22 RT @EricTopol: The use of high-quality masks (like N95/KN95 respirators) by patients, compared with surgical masks, substantially reduced C…

2022-08-15 20:49:47 RT @DGBassani: Infections cut in half in Boston schools that kept mask mandates compared to schools that did not. Even though the schools t…

2022-08-15 05:31:33 @willmcgugan This is great. In fact, it's so good, even *I* understand it. Very high bar.

2022-08-15 05:30:05 RT @dianaberrent: I did a poll in Survivor Corps. 240 participants with #LongCovid.80% had experienced new and unusual dental issues sinc…

2022-08-14 19:59:49 @tunguz Presumptuous question Bojan

2022-08-13 20:07:05 RT @xsteenbrugge: "Voyage through Time"is my first artpiece using #stablediffusion and I am blown away with the possibilities...We're cr…

2022-08-13 17:15:23 RT @tomfolanmd: This week we learned that long Covid results in T-cell exhaustion limiting your ability to fight infection (think: HIV) and…

2022-08-13 10:56:12 Omg @StableDiffusion has turned me into Malcolm Gladwell! https://t.co/7GKRCCTHIF

2022-08-13 10:43:50 @marktenenholtz @iScienceLuvr @StableDiffusion @ylecun @geoffreyhinton @AndrewYNg I thought that too!

2022-08-13 08:17:34 RT @trishgreenhalgh: Do masks really harm kids? Here's what the science says. (TL

2022-08-13 08:12:42 RT @_mbdr_: The CDC estimates 8% of US adults are experiencing long COVID as of early July.Here's the estimate of % of adults with long…

2022-08-13 08:05:25 RT @whoaitsjoan: A question I often get when I bring up our @bluestockings COVID precautions (all outdoors, masks on at all times AND negat…

2022-08-13 02:27:37 RT @loscharlos: Absolutely stunning #LongCovid findings:"Compared with healthy controls, the capillary density of #LongCovid patients w…

2022-08-13 00:24:46 @YouAreLobbyLud When I ride my motorbike on the Philip Island track at 280km/h and brake as late as possible to try to tip into turn 1 carrying good corner speed, I try to ensure I'm as prepared as possible, &

2022-08-13 00:21:27 RT @YouAreLobbyLud: And I'm anxious about getting COVID. Why? Because I'm not a fucking idiot. I do a risky job - remote emergency doctor.…

2022-08-12 11:09:07 RT @NielsRogge: H/t to @jeremyphoward for recommending the LeViT series in terms of speed/accuracy trade-off for image classification, base…

2022-08-12 11:05:22 RT @MathStuart: Some Long Covid Math:I decided to run some pure math-y calculations, based on the CDC's estimates of Long Covid that came…

2022-08-12 03:23:58 RT @drclairetaylor: 1/My sister caught Covid when she visited the UK 8 weeks ago. Since returning to Australia she&

2022-08-12 03:21:47 RT @wasimlorgat: I'm very pleased with how this tiny lib turned out :)• all 16 color ANSI codes (with bold, underline, etc)• nested styl…

2022-08-12 00:18:52 @Pixelfox Yes see the thread I linked to

2022-08-12 00:12:17 RT @tolstoybb: what is going on https://t.co/5BK0x4dWyQ

2022-08-11 22:28:40 @karpathy @Suhail @numba_jit I really enjoyed this GPU programming course:https://t.co/eYmqzrk1eq

2022-08-11 22:26:49 More information and examples of APL are available in the thread from when I announced the start of the study groups:https://t.co/bQ6LaVUiJ4

2022-08-11 22:26:48 APL is different, because every one of the 77 glpyhs has precisely defined behavior. Glyph combinations also have precisely defined behavior, based on principles from combinatory logic.This is all implemented as a programming language, so you can execute and debug an expression

2022-08-11 22:26:47 I've just the final video of over 20 hours of APL study sessions, that cover every one of the 77 glyphs!APL can be thought of as a more flexible, precise, and consistent mathematical notation. It's been under continuous development for >

2022-08-11 21:44:41 @code_report @JuliaConOrg @JuliaLanguage Thanks for sharing, Conor!

2022-08-11 21:44:01 @Suhail @karpathy Personally I found doing a bit of CUDA programming useful, for understanding what's really going on. But if you want to try that, do it in @numba_jit, not C, since it's much easier to get started.

2022-08-11 21:34:28 @TheZachMueller @heads0rtai1s The faster way:https://t.co/x1cQEL3RvQ

2022-08-11 08:35:04 @lucasgvazquez Well... I've got a lot of code, so maybe.But since I have so little boilerplate in my code, I'm not sure how much copilot would help!

2022-08-11 06:20:21 RT @YouAreLobbyLud: "These data suggest recent reactivation of EBV and VZV in [Long COVID]" This research is so important, defining the i…

2022-08-11 05:03:50 RT @VirusesImmunity: Here is the summary of our findings. There are many implications for diagnosis and therapy for #longCOVID. (26/) https…

2022-08-11 05:02:48 RT @radekosmulski: How to address overfitting?I created this some time ago but there have been some discussions recently on overfitting…

2022-08-11 02:14:56 Very cool new addition to fastcore: -- a simple, concise, fast interface to adding basic formatting in your terminal output: https://t.co/uapYwnGBj5Thanks @wasimlorgat! https://t.co/QnTsloxbog

2022-08-10 23:24:59 @Tim_Dettmers @huggingface This is great!

2022-08-10 23:24:52 RT @Tim_Dettmers: We release the public beta for bnb-int8 for all @huggingface models, which allows for Int8 inference without performanc…

2022-08-10 03:07:40 This could save a huge amount of time and complexity for everyone competing on @kaggle -- in code competitions that restrict internet access, you can now easily install pip packages thanks to these new additions to fastkaggle: https://t.co/Usbv6gfA85

2022-08-09 20:06:37 @knapp_graham @willmcgugan But surely the "anyone want tea" call to the household is part of the ceremony of making the tea?...

2022-08-09 15:12:26 Love these ideas. I think we could blend a couple to create an appropriate neologiism. How about we call them: *vibecues* https://t.co/GBGlz5fvgX

2022-08-09 15:08:10 @doomie Congrats on all the great stuff you've put out

2022-08-09 15:06:58 @joe_shipman @JDHamkins @taz_chu It's even got a Wikipedia page https://t.co/orqhO2Kd0Y

2022-08-09 15:05:32 @joe_shipman @JDHamkins @taz_chu This is analogous to the classic technique in debugging - to solve a bug in complex code, first create simpler code with the same bug, and debug that.

2022-08-09 15:01:14 RT @cdotjdotmills: The tutorial uses a downscaled subsample of the relatively new HaGRID dataset. HaGRID contains 552,992 annotated images…

2022-08-09 14:54:19 @holylorenzo @willmcgugan Mine was done on chrome on Android

2022-08-09 14:53:42 RT @quantshah: I am completely sold on #nbdev after watching this tutorial by @jeremyphoward. The way documentation is built along with cod…

2022-08-09 14:53:13 @dbonneville @vijayabhaskarj I suspect that a good UI and thoughtful architecture and pretraining could create some interesting latent factors and allow them to be controlled interactively.

2022-08-09 10:44:40 @vijayabhaskarj @dbonneville Much much harder than what I did with ulmfit though. That was really one key insight that I was 99% sure would work before I did it.Solving illustration requires solving many challenging problems.

2022-08-09 10:42:53 @willmcgugan That's one reason why we code in notebooks!

2022-08-09 06:01:33 @trishgreenhalgh Sounds like you're getting under their skin -- might be because you're such an effective communicator. I can see why they'd want to shut you down, and I can see that they're not likely to be successful!

2022-08-09 05:59:21 RT @MarieSnyder27: One school has been 100% Covid-free with 0 learning loss: Good masks indoors, zero 'mask breaks,' everyone eats outdoors…

2022-08-09 05:58:25 @dbonneville I feel like given a few years I could build that -- I don't see any fundamental technical limitations. Not easy though!

2022-08-09 05:54:29 What an amazing and important step in the place I grew up!Making N95s free means that the government can now also step up the communication campaign of why and how people should wear them. (When access is a problem, there's not much point telling people to do it!) https://t.co/hPXtkby9nR

2022-08-09 05:51:02 RT @YouAreLobbyLud: 1/ Why is all this happening? Why? Why? Why? Many reasons, of course, but here's perhaps one: the last 60 or so years h…

2022-08-09 05:40:28 RT @psvbluemts: Today a NSW mum says goodbye to her child's public school. https://t.co/Nul4ydevZc

2022-08-09 00:16:08 RT @amapresident: "An estimated 5% to 10% of children who’ve had COVID go on to develop long COVID..."We should be doing everything we…

2022-08-09 00:16:00 RT @drajm: What is the basis for asserting #covid will get milder on any relevant time scale?The experts who were correct about the repea…

2022-07-27 09:34:01 RT @hardmaru: A website collected all of the "Emoji Kitchen" from GBoard 2 app so you can play with emoji combinations:https://t.co/jXygr…

2022-07-27 09:30:07 RT @DFisman: The relationship between global counted covid cases and 3 month lagged mystery pediatric hepatitis cases remains, um, robust h…

2022-07-27 09:27:34 @Infoseepage @artlessartthou @larissawaters @senatorpennyqld @janet_rice @EWatsonBrown @BarbaraPocock @sarahinthesen8 Respirators don't have to be perfectly sealed to be highly effective.

2022-07-27 09:24:34 RT @trishgreenhalgh: This. One day I’ll publish the emails from senior scientists telling me in March 2020 to shut up about masks “there’s…

2022-07-27 09:21:09 @DrKW12 @YouAreLobbyLud No.https://t.co/pYEnE0JiWp

2022-07-27 09:18:53 RT @YouAreLobbyLud: Since we are now facilitating mass death and illness in children and adults from COVID, are we now also going to stop t…

2022-07-27 08:32:15 RT @paulg: A founder with knowledge of the US labor market told me that decreased immigration is now a significant drag on economic growth,…

2022-07-27 02:47:02 RT @TheZachMueller: Did you know @huggingface Accelerate is now on @kaggle? h/t @abhi1thakur for the heads up https://t.co/Lb8rm9LLl8

2022-07-27 02:27:11 @code_report @nvidia Wow that sounds like an amazingly fun job...

2022-07-26 20:34:20 @arunprakashml @fastdotai I remember watching the hinton course when it came out, and he mentioned in passing this "neat trick", and then I looked everywhere to try to find where that had come from!

2022-07-26 20:33:45 @arunprakashml @fastdotai And rmsprop appeared first in a coursera course (and never become a paper afaik)!

2022-07-26 20:33:11 RT @arunprakashml: Interesting fun fact: ULMFiT algorithm(transfer learning for NLP) first appeared in a @fastdotai course(2016 version) be…

2022-07-26 20:32:23 RT @EricTopol: BA.5 in AustraliaNew record of hospitalizations @covidbaseau Sharp rise in deaths @OurWorldInData https://t.co/PQYcZtmSOl

2022-07-26 20:29:47 @rahuldave @andfanilo @thedataprof @Gradio @detahq Wow that sounds awesome. Is it publicly accessible?

2022-07-26 18:09:36 RT @kakape: We are two months into an outbreak of monkeypox across the world and I find it incredibly frustrating that we seem to be stuck…

2022-07-26 18:05:50 RT @tomgoldsteincs: You can buy an 8 GPU node for the cost of renting on AWS for about 2 months. It’s amazing to me how many companies and…

2022-07-26 18:03:39 @andfanilo @thedataprof @Gradio @detahq Thanks for great info!

2022-07-26 08:08:26 DALL-E seems to have given Mona List a semi-transparent mask! https://t.co/cR3vHqtsSG

2022-07-26 08:07:33 @Nymarius_ They turned out pretty nicely!

2022-07-26 08:07:13 @lucasgvazquez Very cool!

2022-07-22 10:36:51 RT @mcfunley: I’ve been sitting on a weird epic ramble about Etsy homepages from the middle 2000’s for the better part of 15 years now so I…

2022-07-22 10:22:23 RT @lucasgvazquez: Kaggle is hard. It's easy to start a competition, clone a baseline get the same results as everyone else, get stuck and…

2022-07-22 08:21:21 @WeizhuChen @PyTorch @huggingface @Gradio We are using v3 in the course! Love it!

2022-07-22 07:16:18 @Poonamligade I love the dino classifier! :D

2022-07-22 04:35:04 @wightmanr @giffmana @ludwig_stumpp My student is a category theorist and he has some interesting ideas to follow based on lens theory. I think it might be a valuable research direction to follow, although it's very early days.

2022-07-22 04:33:43 @AnoopRKulkarni @PyTorch @huggingface @Gradio Everyone in the audience wore a mask too -- it was nice to see people looking out for each other. There were no known cases of transmission throughout the 8-week course, despite many covid cases in the community.

2022-07-22 04:26:37 @TheZachMueller This was done in keras by @fchollet

2022-07-22 04:21:49 @sudalairajkumar @fastdotai Thank you for sharing! :)

2022-07-22 03:48:48 We were really lucky to be joined today by APL guru @a_brudz, who gave us a masterclass on the rank operator in APL.He really is a great teacher, and this was a fascinating topic for a deep dive. If you're interested in array programming you'll love it! https://t.co/GaN6oqtJiA

2022-07-22 03:11:35 RT @megelison: deeply disgusted to discover that in order to get good at a thing I have to do it badly first

2022-07-22 01:17:30 Also, a big hat-tip to @OpenAI for making DALL-E available, so that I could ensure there's an adorable bunny or teddy illustration for every lesson! My favorite is the "data ethics" illustration...https://t.co/yDxkMWtAJ7 https://t.co/gYgl53neXx

2022-07-22 01:04:50 If you'd like to learn more about the philosophy of @fastdotai and the results of our work, check out this conversation with @derrickharris on @future (part of @a16z)https://t.co/loud8GECK4

2022-07-21 23:50:40 @abidlabs Yup for now dropping by the forums would be the most helpful thing for our students. Many thanks for the offer.

2022-07-21 23:44:59 @chrishareau @PyTorch @huggingface @Gradio Thanks! :D

2022-07-21 23:44:18 I'm proud to be associated with @UQSchoolITEE, where this course was recorded.https://t.co/QyJSzac5E3

2022-07-21 23:42:47 @abidlabs Our students all loved it. Look at all the discussion of gradio happening on our forums!https://t.co/zPVuuKy4Dz

2022-07-21 23:41:51 Big thanks to @quarto_pub, which I used to create the course website (along with #nbdev)

2022-07-21 23:39:07 Many students have told us about how they’ve been gold medallists in ML competitions, got offers from top companies, and had research papers published.Alum have gone on to jobs at organizations like Google Brain, OpenAI, Adobe, Amazon, and Teslahttps://t.co/Dsy0qady9P

2022-07-21 23:39:06 Since first using neural nets >

2022-07-18 06:27:24 RT @Lost1nSpace: Bit of a bombshell today. The QLD health minister tells us that they thought the waves would get smaller, and people would…

2022-07-18 04:17:45 @jj__george @MaggieM00021130 @math_rachel @flo_mask @nickelpin You can't. It's fine to draw on it.

2022-07-18 04:16:06 RT @math_rachel: My daughter &

2022-07-18 01:55:11 @dan_s_becker @isaac_flath @HamelHusain Well... I don't sell anything :)

2022-07-17 23:29:46 @mygiddyauntie @renegadesilicon Yup this is the one I foundhttps://t.co/7kx1VUHuEC

2022-07-17 23:01:29 @ColmanHumphrey @phil_eaton I'm one of those people.There's a huge gap in functionality between Excel and Google Sheets or any open source spreadsheet.

2022-07-17 22:59:27 @isaac_flath @HamelHusain Yeah this whole discussion about twitter bots has made me realise an awful lot of people deeply misunderstand how ML works

2022-07-17 22:58:41 RT @isaac_flath: It's interesting to see the take that human labeling is so inaccurate it's not meaningful, therefore we should use supervi…

2022-07-17 22:58:15 Great to see further analysis of carrageenan-containing nasal sprays -- I read the original papers covering this a few months ago and decided that there's enough evidence for them to buy some for our family. https://t.co/yfM8vQFlKR

2022-07-17 22:54:35 @migueldeicaza Maybe he's too busy reading that book to speak to you now...

2022-07-17 20:25:58 @MariaMacNamara1 @AlboMP @techcouncil_aus @Mark_Butler_MP I'd be delighted to help

2022-07-17 15:59:38 RT @willmcgugan: COVID has had more is an affect on me than I would have cared to admit. Stills have this lingering lethargy. Keep safe, it…

2022-07-17 03:04:31 This is what we tell all our students too https://t.co/2wxmacwTEj

2022-07-17 01:25:46 @pjie2 @elonmusk https://t.co/BcZHl7wKr0

2022-07-17 01:25:22 @unJADded @elonmusk Yes, but my understanding is that Musk specifically referred to # of bot accounts.

2022-07-17 01:24:19 @george_cary @elonmusk The CI for this analysis is ~1%. I think that's quite close enough.The ABS needs counts for each geographic region and demographic. Some are quite small. Randomly sampling would be hard.

2022-07-17 01:22:38 @jhaluska @elonmusk There's no need to do it every day.

2022-07-17 01:22:20 @GeoffMantel @elonmusk https://t.co/l4IMS8BaIL

2022-07-17 01:21:55 @jacquesT2017 @elonmusk https://t.co/l4IMS8BaIL

2022-07-17 01:21:32 @Neil_Batlivala @elonmusk That's a different thing entirely - the process used to flag bots at signup would generally be a largely automated process. By definition, 100% (or thereabouts) of attempted signups caught by that will not become active accounts.

2022-07-16 05:13:56 On Aug 3 I'm chatting about AI for Libraries, Archives, and Museums. All are welcome to join.https://t.co/wM7xotRVaz

2022-07-16 03:28:51 @YouAreLobbyLud Wait what? How?

2022-07-16 03:26:41 @savvyRL So what did you decide about the glasses?

2022-07-16 03:21:30 RT @random_walker: ML is being rapidly adopted in the sciences, but the gnarly problem of data leakage has led to a reproducibility crisis.…

2022-07-15 21:25:52 RT @danielgoyal: Good rule of thumb for those at low risk from Covid…if you share a space with someone wearing a mask, put one on too. Th…

2022-07-15 18:55:57 RT @jcbarret: Fascinating case study of what peer review achieves. After skepticism on pub (big claims require big evidence), 2 groups have…

2022-07-15 18:51:42 RT @OfficialLoganK: The annual @JuliaConOrg 2022 conference is kicking off in exactly 4 days ! Here is a quick thread covering all of…

2022-07-15 08:15:39 @maskupmelb @catinthefield Ah OK yes I remember now!

2022-07-15 06:42:37 @maskupmelb @catinthefield Last week when I tried to order from them they said I needed to show a medical license!

2022-07-15 03:23:14 RT @MetaAI: Excited to announce Make-A-Scene, our latest research tool Mark Zuckerberg just shared. Make-A-Scene is an exploratory concept…

2022-07-14 10:03:52 RT @CathNoakes: Ventilation sceptics, don’t worry, good ventilation and air cleaning won’t make a room sterile. You’ll still breathe a few…

2022-07-14 09:40:57 @wasimlorgat Struct bytes should be much faster generally IIUC. Especially with the stdlib json module which isn't fast.

2022-07-14 04:03:32 RT @joeyeatsfridays: Lately I've seen some fatigue from graphic designers who are seeing the same mockups over and over. I was just given a…

2022-07-14 04:01:02 RT @drajm: Sorry @Ahpra but community trust &

2022-07-14 03:55:10 @itjohnstone @maskupmelb @MichaelSFuhrer @BurnetInstitute My understanding is that much of this stuff is highly secret. NDAs, strict confidentiality requirements, etc. That's based in talking to some of the folks doing the actual work.

2022-07-14 03:53:00 Some useful quick-and-dirty math in this thread.Tl

2022-07-14 01:23:22 @wasimlorgat @simonw @fastdotai @HamelHusain @iScienceLuvr @TheZachMueller I learned a lot about running a community from @edwinhayward's email discussions forums. He was always very quick to jump on toxic behaviour, which I first I thought was over-the-top, but I grew to appreciate.

2022-07-14 01:22:13 @wasimlorgat @simonw @fastdotai @HamelHusain @iScienceLuvr @TheZachMueller Be nice to each other :)On one occasion we did have a community member that wasn't very nice at all, so I told them to go away for a week and come back when they'd reflected on their behaviour.Very occasional strong policing results in, hopefully, less need for intervention.

2022-07-13 22:38:21 @maskupmelb @catinthefield Bunnings is out of stock at the moment.Yes it's such a shame that they're so expensive in Australia :(

2022-07-13 22:31:57 Big loss for Tesla. But 5 years is a long time in deep learning! https://t.co/Xv8R6Ie7LN

2022-07-13 12:22:41 RT @chrischirp: Thank god we're near the end of term because BA.5 is now ripping through schools again. Highest ever absences for seconda…

2022-07-13 10:25:14 @maskupmelb @Globalbiosec @kate_cole_ It's important to remember that "passing" a fit test means a fit factor of 100, which is 10x higher than the actual goal, in order to have a large safety margin.So most of the time "failing" a fit test will still result in protection higher than the actual target fit.

2022-07-13 06:11:58 RT @edyong209: It is wild to me that some random guy on Twitter decided that the BA.2.75 variant was going to be known as "Centaurus" and i…

2022-07-13 06:09:13 RT @hardmaru: I just finished processing a paper for the new TMLR journal, from initial submission to final decision. I enjoyed the process…

2022-07-13 06:07:12 @Carolin29777437 @CorsIAQ You can reuse them until the straps fail, so it you get 30 wears, that a few cents per wearhttps://t.co/z4bEcs1L5z

2022-07-12 22:50:56 RT @SharnelleVella: Minister Thomas says the CHO wanted to mandate masks in early childhood, school,retail &

2022-07-12 21:56:51 RT @AJamesMcCarthy: Stephan’s Quintet: My shot using a 70mm telescope in my backyard versus JWST, the most advanced astronomic instrument e…

2022-07-12 21:49:02 And this pic contains an amount of sky covering the visual area of a grain of sand held at arm's length. https://t.co/bqjOSbZ5FZ

2022-07-12 20:45:06 @joelgrus @MolasAlex @vboykis It's only half the answer, at mosthttps://t.co/XzEV6XULWJ

2022-07-12 20:36:05 RT @__mharrison__: What causes you the most stress when programming?

2022-07-12 18:15:46 RT @psvbluemts: @BigBadDenis Slashing of the C-19 “reinfection period” from 12 weeks to 4 made me research why the “safe” period was EVER 1…

2022-07-12 18:05:55 RT @MichaelTrazzi: SF gossip:- half of Google Brain scaling teams left to https://t.co/Cc4CJtduxz &

2022-07-12 10:19:53 @HamelHusain @MetaflowOSS There's a good description of how to use it here (and also a mention that using setup-miniconda is at least as fast)https://t.co/835fkM1n4x

2022-07-12 09:51:31 RT @YouAreLobbyLud: I'm not allowed to criticise this

2022-07-12 09:44:46 @christinewdavis @flo_mask @CorsIAQ @LazarusLong13 How's it gone so far? Are you aware of any covid transmission that's occurred at your practice?

2022-07-12 06:12:10 @samanthamaiden (I do want to point out that I mainly interact with Americans, who don't have such a nuanced understanding of vocabulary such as this.)

2022-07-12 06:11:31 @samanthamaiden OK I'll try it next time and will report back.

2022-07-12 06:10:15 We're generally pretty quick to let devs know when their software doesn't work...but it sure is nice to receive a note when it *does* work! https://t.co/SRU5yGjdJv

2022-07-12 06:07:40 @samanthamaiden So, what are the recommended responses to:- "Do better"- "Let that sink in", and- "You do you"?It feels like just replying with "wanker" is insufficient...

2022-07-12 06:05:24 https://t.co/me5j2cqOyd

2022-07-12 06:05:15 https://t.co/WzqjwZJErn

2022-07-12 04:10:21 @dstufft Fair

2022-07-12 04:09:19 Yes, please do better than saying "do better" https://t.co/fiEPZAKqoO

2022-07-12 04:06:24 RT @stuperbug: Covid series 3 review 2/5 stars. Multiple streaming services. Free to Air.The writers of this show have clearly run out of…

2022-07-12 04:03:38 RT @willmcgugan: This is very concerning. Long Covid is going to have a lasting impact that affects as all.

2022-07-11 09:49:01 RT @DrGBuckingham: Look closely.There are actually 16 circles in this image.And once you see them, the image appears changed forever.…

2022-07-11 09:41:35 @wasimlorgat Excellent analysis!

2022-07-11 06:03:42 RT @CrabbBrendan: Wow, wow. This a must. If you're interested in knowing where you most need a mask (a P2/N95) this is a brilliant guide. V…

2022-07-11 05:58:22 RT @JohnSnowProject: Don't breathe it in. #COVIDisAirborne https://t.co/czmr4ki2dC

2022-07-11 05:53:40 RT @TRyanGregory: Screen caps of tweets by minimizers that have aged very poorly. Go.

2022-07-11 05:41:25 RT @YouAreLobbyLud: Very kind article about me by Dr Andrew Miller, @drajm, published in the @westaustralian. https://t.co/ldiekUkDmn

2022-07-10 23:24:37 Do people that believe it's a good idea to breathe unclean air, also think it's also a good idea to:- drink unclean water,- eat rotten food,- and avoid washing hands?That'll give your body lots of exposure to pathogens to learn from, it's that what you're looking for...

2022-07-10 21:45:17 @ralfgommers I'd be various curious to hear what you find, because it's a mystery to me!

2022-07-10 21:44:40 @davidbrochart @willmcgugan Good point! Sorry - I meant a textualize-based implementation.

2022-07-10 21:35:00 RT @chalmagean: Ruby on Rails is coming back!If you're on the Javascript train, you're probably tired of new releases.But this one's go…

2022-07-10 18:35:37 @MattSBone @Teach_Solutions f(g(x)) is read right to left - or at least, that's the order it's evaluated. You can "read" anything in any direction, I guess...

2022-07-10 18:32:00 @willmcgugan @davidbrochart The thing that'll transform TUI jupyter is sixel support... :)

2022-07-10 18:29:52 RT @yaneerbaryam: Any half baked expert would know this doesn’t make sense: “Covid becomes like the flu and evolves herd immunity.” The flu…

2022-07-10 18:23:18 RT @MitchellAGordon: Why aren't any recent LLMs (OPT, PaLM, etc.) using "efficient" architectures (Reformer, Longformer, etc.)?There's 20…

2022-07-10 18:14:43 @ralfgommers The github issue is confusing because there's some folks discussing pypi too, including the 1st reporter. But afaik it's only conda which has nearly always been slow for this one package

2022-07-10 18:11:41 @ralfgommers Since I visited Australia every year or so I experienced this myself for a number of year even before I moved year. I'm not sure how long it's been happening - as long as I remember...

2022-07-10 18:10:39 @ralfgommers There's already a number of issues and forum posts about it, e.g https://t.co/Ugvkb9Rm9E

2022-07-10 11:06:07 RT @rjrshr: Arithmetic circuits were once the craft of human experts, and are now designed by AI in NVIDIA GPUs. H100 chips have nearly 13,…

2022-07-10 10:37:29 @CoenMouton @PyTorch Oh africa too? Did you try VPN? Did it help?

2022-07-10 10:37:03 @MMA_mathematics @PyTorch I'm not sure they're banned in any countries that have access to Twitter.Also, most places where people claim there's a VPN ban are actually not quite that - instead VPN use is regulated (eg in China there are certain VPNs you can choose from).

2022-07-10 05:14:21 @PeteUK7 @me_allyann Somehow capitalism is largely ignoring the market of people who want safe air at the moment.

2022-07-10 03:29:19 @me_allyann Yup that would be a good option too.I think either approach would be great for business.

2022-07-10 03:25:19 Just like it's better to have no smoking on planes, it's better to have no unmasked passengers on planes.But if you are going to allow unmasked passengers, why not at least put them together?

2022-07-10 03:25:18 I always used to chose the Non-smoking section of a plane. I never smelt smoke during the flight.Which shows that planes *can* provide clean air to those who want it.

2022-07-10 03:25:17 Planes used to have Smoking and Non-smoking sections.What if we had Masked and Unmasked sections?

2022-07-10 01:03:40 @michaelwaskom @choldgraf @pypi Crikey https://t.co/8iaaQBGLqC

2022-07-10 00:55:26 RT @SuckBoyTony1: Big spring doesn't want you to know that you can make your own springs. I now have hundreds of ghost springs https://t.co…

2022-07-10 00:51:27 @choldgraf @michaelwaskom @pypi Also it doesn't work if you're Australian, since they picked a vendor that has export restrictions, and Australians can't be trusted.

2022-07-10 00:50:55 RT @choldgraf: Hot take: the real problem is not all those @pypi maintainers not using 2FA, it's the number of critical projects with one o…

2022-07-09 23:44:19 RT @HamelHusain: Provocative question: As a field, are we only showing people complex ways of doing things? Are we making people bel…

2022-07-09 23:03:33 RT @HamelHusain: Sharing without comment, just emojis: https://t.co/9yGYj21v3Z https://t.co/WJAJKIuttC

2022-07-09 22:05:50 @strickvl That program would work great for adults too

2022-07-09 08:28:10 @glbabbington @CrabbBrendan @joeyfox85 For the cost of one fit test, you could distribute dozens of high quality well-fitted masks with a slip of paper showing which way round that go and saying to pushing the nose-wire.

2022-07-09 08:27:03 @glbabbington @CrabbBrendan @joeyfox85 In general, I believe we should be increase access and removing obstructions, rather than putting obstructions in place that stop people having access to high quality masks.Requiring fit-testing is an obstruction that reduces that access.

2022-07-09 08:25:05 @glbabbington @CrabbBrendan @joeyfox85 If you restrict mask usage only to those that can access fit-testing services, you're removing access to the vast majority of the population. I don't think that's fair or reasonable.My vulnerable family member would be unlikely to use such a service, for instance.

2022-07-09 08:23:45 @glbabbington @CrabbBrendan @joeyfox85 I think that requiring fit testing endangers the clinically vulnerable far more.People don't need "fit-testing education". They need to be told not to put the mask upside-down, and to squish the nose wire in. That can be done with a 30 second video or piece of paper.

2022-07-09 07:22:01 @covidmeetupscom Heh I looked up your website last week as it happens!I even found someone in my area......turns out it was my wife

2022-07-09 07:20:45 @YvonneGunn @darylgibson That's a great approach. How many wears do you find you get before the straps break or become too loose? What kind of mask do you use?

2022-07-09 05:41:55 @glbabbington @CrabbBrendan A non-fit-tested N95 absolutely should not have 20% leakage. A fit factor of 88 means closer to 1% leakage.20% leakage is a fit factor of 5, which is closer to a surgical mask.

2022-07-09 04:53:44 RT @code_report: It was awesome having @jeremyphoward on @arraycast! A must listen for anyone interested in machine learning/data science o…

2022-07-09 04:50:47 @glbabbington @CrabbBrendan Fit testing however is much harder to access and many people will not be able to find such a service in their area. It also requires indoor face-to-face interaction, which we should try to avoid.Requiring fit testing therefore could result in fewer people having access to N95s.

2022-07-09 04:49:40 @glbabbington @CrabbBrendan I do think that education is very important however. In studies of having non-experts don N95s themselves by far the most common problem is failing to correctly tighten the nose-piece.This is an easy technique to teach.

2022-07-09 04:48:47 @glbabbington @CrabbBrendan One of the most important people in my life is in that category. However I don't agree that fit-testing is essential, and I do think there could be significant harms in such a positioning.The average fit factor of 88 achieved by non-experts is plenty high enough.

2022-07-09 02:37:39 RT @math_rachel: This is so disturbing. In Australia, it's ok for MDs to tell people pandemic is over, wall of immunity, droplet dogma, kid…

2022-07-09 02:01:47 I had an really interesting discussion with the @arraycast team about why I love array programming, using @PyTorch and @TensorFlow, numpy, @dyalogapl, C++, and perl.You can watch it here:https://t.co/FBQQ6ElS71

2022-07-08 23:59:10 @DeeCee451 @yaneerbaryam I didn't know you were a fan of using tone policing to silence legitimate scientific and medical debate.

2022-07-08 22:22:54 RT @yaneerbaryam: Unacceptable action by Australian authority to silence one of the best voices for science and medical ethics of do no har…

2022-07-08 22:02:31 RT @gd3kr: built a quick app that uses gpt-3 to convert from English to RegEx so you don't have to waste time on stackoverflow: https://t.c…

2022-07-08 21:48:15 A question I just got via DM -- figured I'd share here in case anyone else is interested too: https://t.co/20nTE6KAf6

2022-07-08 21:41:10 @HamelHusain @renegadesilicon @radekosmulski When I took that photo I turned 180 degrees and took this one. As you see, we have room for visitors around here... :D https://t.co/9qnO8rUjpu

2022-07-08 21:36:49 @HamelHusain @renegadesilicon @radekosmulski Mission accomplished then... :p

2022-07-08 21:30:07 @ludwig_stumpp @emollick I think distributed practice has plenty of research support to show that it's useful for a wide range of people. It's not just a case of personal preference.

2022-07-08 21:28:21 @wightmanr Good luck! :D

2022-07-08 21:26:28 RT @renegadesilicon: @HamelHusain Take the hint from me, @radekosmulski and @jeremyphoward and move to Queensland here in Australia where i…

2022-07-08 21:26:08 @renegadesilicon @HamelHusain @radekosmulski I took a photo on my walk yesterday too, as it happens :) https://t.co/28zBuCoUvJ

2022-07-08 20:39:21 @wightmanr Great news. Does your results CSV include these new models yet?

2022-07-07 07:21:57 RT @danaludwig: It appears that in Massachusetts our "mild" Omicron variant cause 3.34 times the excess mortality as Delta!"The per-week O…

2022-07-07 02:39:20 RT @rebecca_roache: This is now my stock response to 'invitations' from academic journals to review their submissions#AcademicTwitter #aca…

2022-07-07 02:25:31 RT @jathansadowski: The guy who just won the Fields Medal – like a Nobel Prize for mathematics – has a work ethic we should all emulate. Th…

2022-07-06 23:29:21 RT @4ndr3aR: Semtorch!An excellent package for doing semantic segmentation on top of @fastdotaiWritten by the awesome David Lacalle Cas…

2022-07-06 22:14:47 @ciiaqap They don't work well at all

2022-07-06 22:14:25 RT @math_rachel: A public health academic saying non-fit-tested N95s are "useless", and that those w/out public health credentials can’t di…

2022-07-06 22:12:43 @HamelHusain Also @waydegilliam. Brilliant and under appreciated developer

2022-07-06 22:11:40 RT @vzzqzzz: .@jarvislabsai is the best gpu cloud provider for DL practitioners out there, period. More than once I had a question and supp…

2022-07-06 22:10:31 RT @wightmanr: Looking at a recent @NVIDIAAI paper (GC ViT) + impl, and I'm seeing an unfortunate trend. Why are recent NVLabs repos using…

2022-07-06 19:43:58 @sailorrooscout 2nd paper: "waning effectiveness against hospital and emergency department admission after receiving a third dose of BNT162b2 vaccine is likely nuanced

2022-07-06 10:10:19 RT @JacobSteinhardt: In 2021, I created a forecasting prize to predict ML performance on benchmarks in June 2022 (and 2023, 2024, and 2025)…

2022-07-06 08:55:25 RT @HamelHusain: If you know, you know. #nbdev https://t.co/1FtXT6gJCa

2022-07-06 08:30:33 RT @nathalie_k1: “Using a smoke machine, they studied airflow patterns in the school's five classrooms and administration areas.Carbon dio…

2022-07-06 08:27:29 @Dmojavensis @Sam_Altschul @ylecun Existence proof

2022-07-06 08:20:23 RT @jarredsumner: Introducing Bun - an incredibly fast all-in-one JavaScript runtime. https://t.co/Yt6tAcnBQs

2022-07-06 06:05:21 RT @UQSchoolITEE: Our inaugural #UQITEE Practical #DeepLearning for Coders short-course collab w @fastdotai (cc @jeremyphoward) wrapped up…

2022-07-06 03:38:51 If you interested in learning/refreshing some basic undergraduate math topics in a hands-on way, our APL study group might be for you.In the 1st two days we've already covered limits, e, magnitude and direction of complex numbers, complex conjugates…https://t.co/6QaTMcrB4d https://t.co/qCfcVa5QaC

2022-07-06 03:08:02 @francoisfleuret My daughter and I both love the @daveexplains videos on this topic!

2022-07-06 03:04:19 @TheZachMueller @huggingface That's a really great goal for a low-level helper lib like this!

2022-07-06 03:00:18 @vboykis If someone asks me to define something, I nod, smile, and walk away.Or if I'm in a difficult mood, I'll answer with: "define 'define'".

2022-07-06 02:07:30 RT @kurianbenoy2: The @fastdotai has officially been wrapped. Thanks Jeremy for creating this awesome course, which gets better every year.…

2022-07-06 01:51:04 @theweskyle @Suhail Actually come to think of it we possibly only support numpy-style at the moment, not Google-style. No reason we couldn't add that though...

2022-07-06 01:50:15 @theweskyle @Suhail Yes, but you still would author the prose in notebooks. That's pretty much the whole point of nbdev - to leverage notebooks as an authoring environment

2022-07-06 01:49:08 @sgaseretto @Suhail It will be renamed

2022-07-05 22:50:37 @nichochar @Suhail @elixirlang fastcore.docments also parses comments and turns them into docs (well more specifically it turns them into metadata, which nbdev turns into docs)

2022-07-05 22:36:26 @Suhail Obviously the answer is nbdev... :D

2022-07-05 20:15:34 RT @fedbybirds: Feel as if all the women I know, certainly the middle-aged ones, are gradually amassing #LongCovid symptoms after their 1st…

2022-07-05 19:53:19 RT @ericbkennedy: Then, the easiest thing for me to do is think of COVID like cigarette smoke. I ask myself, "if someone was smoking/had sm…

2022-07-05 19:51:42 RT @joe_d_campbell: "'I have this math competition experience, that ... you have to be fast,' [Wang] said. 'But June is the opposite. … If…

2022-07-05 19:30:10 @sri_srikrishna @lucycookiz @AbraarKaran @RanuDhillon @linseymarr @jljcolorado Great outcome!

2022-07-05 13:59:57 @ENirenberg This kind of hyperbolic straw manning helps noone. It plays to people's emotions but only serves to cloud people's ability to clearly understand the issues.

2022-07-05 13:55:41 RT @emollick: A map of the entire universe to (log) scale, by @budassiuniverse. Spacecraft, planets, large-scale galactic structures - they…

2022-07-05 07:57:07 RT @RemyLevin: I've been using Mendeley Desktop daily for years, &

2022-07-05 07:52:49 @capetorch @weights_biases I'm looking to host the blog on a neutral platform - hence the need to upload the images to a server and get the link

2022-07-05 07:03:02 If you missed the first APL study group session today -- no problem! Click the forum link below to watch the first recording, and if you like it, you can join us for study session 2 tomorrow. https://t.co/bQ6LaWblL4

2022-07-05 06:21:28 @willmcgugan Happy birthday mate!Vaccines are cool -- but please still be careful, because you still have a risk of getting long covid.You can reduce that risk by taking it easy -- don't over-exert yourself until you're totally better! We need you to stay well https://t.co/pErXwzXyt1

2022-07-05 05:00:28 RT @math_rachel: Why masks are still important:- vaccines offer limited protection from long covid- unconstrained transmission → more var…

2022-07-05 04:43:22 @wasimlorgat nbprocess, nbdev, and quarto support that too!

2022-07-05 03:41:48 @greg_ashman Congrats Greg!

2022-07-05 03:40:24 @lucycookiz @sri_srikrishna @AbraarKaran @RanuDhillon @linseymarr @jljcolorado For sure - it's a tough balance and requires making compromises. It's a shame that we can't rely on others in our communities to help keep our kids safe by wearing high quality masks indoors where possible.

2022-07-04 23:25:09 @walkerjian There's a thing in the toolbar to adjust font size. I didn't find it for a while either!

2022-07-04 23:20:55 @sri_srikrishna @lucycookiz @AbraarKaran @RanuDhillon @linseymarr @jljcolorado We decided against sending our kid to public swim lessons, on the basis that it generally involves very close contact.Obviously each person will have their own risk profile &

2022-07-04 08:35:46 RT @stefan_it_: Very interesting discussion: is XLNet still underrated and "ignored" by research community https://t.co/LgZGuE6P5G

2022-07-04 06:11:45 RT @wasimlorgat: TIL: You can compile *ordinary Python* to C with Cythonhttps://t.co/lRssuQGuW5 https://t.co/hn59iBEb5t

2022-07-04 05:34:12 RT @fastdotai: New #fastai release: v2.7.5. https://t.co/Hgk4C6aovdNew Features- auto-normalize timm models- PyTorch 1.12 support

2022-07-04 03:19:48 RT @Bob_Wachter: This means that to gather indoors safely, a vax/booster requirement – while better than nothing – isn't very reassuring. T…

2022-07-04 03:19:42 RT @Bob_Wachter: We need to change our thinking about the value of vax/boosting. Vax/boost remains hugely valuable in preventing a severe c…

2022-07-04 03:19:35 RT @Bob_Wachter: So I still avoid indoor dining &

2022-07-04 02:50:24 @shakes76 This ain't lambda calculus...

2022-07-04 02:49:27 @pierreJaune Yup we're doing APL now. Search my recent tweets.

2022-07-04 02:49:00 @MANas8U It wasn't on today. :)

2022-07-04 02:42:40 RT @finmoorhouse: Some gradients between two colours look good, others suck. It's easy to tell the difference, but what *is* that differenc…

2022-07-04 00:24:04 @NmaViv https://t.co/BCkguNVE5S https://t.co/PCKYZVTWxl

2022-07-03 23:30:22 A first taste of metamathematics, using APL.(The kids loved imagining a world without math, but with "zero" and "successor".) https://t.co/wxJU7Shor7

2022-07-03 21:55:01 RT @ECooperAuthor: Before Covid, I had a huge amount of faith in the CDC. I am a pro-science person and I believed that they would act in a…

2022-07-03 18:06:07 RT @emollick: Effective individual happiness strategies:Invest in family &

2022-07-03 18:05:24 RT @gordic_aleksa: The most valuable thing I learned from @Google Minerva?How bad we are (even the boldest among us) at making predictio…

2022-07-03 17:53:54 @PunksRant It's possible, although I'd like to see a language where it's as native as regular math notationhttps://t.co/5P8Km9grRp

2022-07-03 09:27:52 @walkerjian Dyalog

2022-07-03 09:26:50 @eichin @djf0001 APL is easy to type BTW. On Mac you use the back tick prefix.

2022-07-03 09:25:51 @eichin @djf0001 No notations are intuitive until you understand then. It's not a good basis for deciding whether a notation is useful.

2022-07-03 05:58:39 Sorry a correction: A f∘g B == g(A, f(B))https://t.co/cnd0PMHNbQ

2022-07-03 05:57:11 Finally, the two "1"s in the expression 1+∘÷⍣=1 are the initial values passed in to +∘÷. And that gives us our continued fraction that calculates ϕA notation that provides such expressive capabilities is a notation that helps us think deeper thoughts, thru deeper abstraction

2022-07-03 05:57:10 ⍣ is the "power operator" in APL. f⍣g repeats calling the function f again and again, until the function g returns true (1). On each iteration the previous value of f is passed back in, along with the LHS array.

2022-07-03 05:57:09 Let's start with this: +∘÷. That's a composition of ÷ followed by +. In general, for arrays A &

2022-07-03 05:57:08 The "golden ratio" ϕ can be calculated as a continued fraction. It's what you get if you continue the series in the image below indefinitely.In APL, that whole thing can be expressed with just this little equation: 1+∘÷⍣=1In this I'll explain how APL does it https://t.co/HGqAfqB2GW

2022-07-03 02:38:43 RT @ianbremmer: one of the most disturbing demographic data trends you’ve ever seen. the united states—pre-pandemic, mind you—vastly unde…

2022-07-02 20:20:47 @RageSheen @glbabbington I admire those few people that continue to fight for public health despite everything stacked against them.

2022-07-02 20:19:50 @RageSheen @glbabbington In the end, I didn't really believe that I could do anything much to counter the nearly universal let-it-rip narrative seen throughout the global media.At the start of the pandemic, people were open to listening and learning. Now, not so much.

2022-07-02 20:17:40 @RageSheen @glbabbington It wasn't a fit-testing paper specifically, but rather a general "stuff you probably ought to know about masks" paper.Anyhoo, I gave up on it. To be more precise, I gave up on doing or saying anything much about covid, because otherwise I felt like I'd just go crazy.

2022-07-02 20:12:53 RT @RnaudBertrand: WowProf. Jeffrey Sachs: "I chaired the commission for the Lancet for 2 years on Covid. I'm pretty convinced it came o…

2022-07-02 17:51:29 RT @renegadesilicon: Good times are here again.....https://t.co/6RRkCs3qfh

2022-07-02 10:34:25 RT @tylerblack32: Kids, School, and Suicide=-=-=-=-=-=-=-=-People are asking about the stats I presented about school days being associa…

2022-07-02 10:33:41 RT @tylerblack32: And finally, no, pediatric societies and non-mental-health professionals, we do not have compelling evidence that schools…

2022-07-02 10:23:54 RT @davidbrochart: Jupyter in the terminal will look fancy, thanks to @textualizeio! https://t.co/aQD2aupiHV

2022-07-02 10:22:36 RT @RageSheen: Personal Covid stories Part 2. My wife is a primary school teacher, I have 2 kids still at school. I work in an open office…

2022-07-02 03:46:03 RT @triplux: Experiments I conducted with DALL·E 2 from @OpenAI replicating styles of well known portrait photographers using photo-realist…

2022-07-02 01:00:56 Will test it out on my daughter and report back...

2022-07-02 01:00:30 TIL Richard Stallman wrote a poem about the rho (rank) function in APL.https://t.co/QmXoT62Rho https://t.co/MYUKkgd2MQ

2022-07-02 00:44:22 RT @dgurdasani1: When everyone talks about the severity of COVID-19 having reduced due to vaccines, they never qualify that this is *only f…

2022-07-02 00:37:58 RT @jackclarkSF: As someone who has spent easily half a decade staring at AI arXiv each week and trying to articulate rate of progress, I s…

2022-07-02 00:34:40 RT @HamelHusain: My first set of contributions to @fastdotai was setting up better CI &

2022-07-01 23:44:16 RT @dgurdasani1: Important long COVID study showing that 40-50% of long COVID patients were seronegative (did not have antibodies) but show…

2022-07-01 23:42:05 RT @stuperbug: Dr John Hewson on an Australian CDC - note the extensive references to @RealOzSAGE . @Globalbiosec - well done for leading t…

2022-07-01 23:39:26 RT @RageSheen: Are there serious medical/public health professionals who genuinely still believe repeated reinfection with SARS2 will gener…

2022-07-01 23:37:51 RT @HamelHusain: Suhail is working on several ML projects to supercharge @MightyApp that are really fun! Highly recommend following thi…

2022-07-01 23:35:37 @simonw Wow this is such a cool step for you and for Python!

2022-07-01 18:21:20 RT @bradlarson: Over the last year, a substantial amount of work has gone into refining the implementation of differentiable Swift. I reali…

2022-07-01 09:55:43 @markkitti Wow cool!

2022-07-01 07:34:28 Oh and one other thing - Iverson wrote some really cool math books using APL notation instead of tradition math. For instance, here's "Algebra, an Algorithmic Treatment"https://t.co/yQXBGKrPNW

2022-07-01 07:24:26 Have you seen "Iverson brackets" in math? Well that's from APL too! https://t.co/nSLSsXUCBx

2022-07-01 07:24:25 Have you seen "iota" (like python's `range`) in other languages? That's actually from APL!Here's an example - also showing the `\` operator, which is a cumulative scan https://t.co/ptap9BTAtQ

2022-07-01 07:24:24 Functionality can be adapted using operators, such as `/`, which inserts a function between items of a list. E.g +/ is sum and ×/ is product https://t.co/j9gzNi8clz

2022-07-01 07:24:23 The basic APL operations work just like math. Here's examples from an APL session -- my input is indented, and APL's results are on the left https://t.co/DpDtPWIgZS

2022-06-28 11:24:19 @SamyRay15 @GuggerSylvain @PyTorch Great question. I'm not sure anyone has really cracked that yet...

2022-06-28 11:22:54 @AsrithDasari @GuggerSylvain @PyTorch Yes. This notebook.

2022-06-28 07:21:09 @radekosmulski @kaggle ... And a way to install in no-internet notebooks?

2022-06-28 07:19:51 @radekosmulski @kaggle Can you show a non messy version?

2022-06-28 04:27:06 RT @Suhail: Writing AI code from @MightyApp using GitHub Codespaces with a V100 GPU in the cloud.You can train, commit code, and build al…

2022-06-28 04:25:29 @radekosmulski @kaggle ...although that uses pip to install, which apparently results in not using GPU.Is there a way to install GPU accelerated nvtabular on @kaggle?

2022-06-28 04:24:47 @LostInTangent @Suhail @MightyApp Yay!

2022-06-28 04:22:52 @radekosmulski @kaggle There's an example of how to use nvtabular on Kaggle with fastai here BTW:https://t.co/FnHxnxikNG

2022-06-28 04:19:18 RT @radekosmulski: nvt op Tuesday! Remember those @kaggle kernels where a person goes meticulously though the hundreds of columns in a D…

2022-06-28 00:10:57 @Ghonem13 @GuggerSylvain @PyTorch Hi. I'm not a Dr.Have you read the whole chapter yet? Is there anything in particular missing from the suggestions there that you're looking to improve on?

2022-06-27 23:47:26 @jsdkodkjokdf @kaggle I'd suggest binge watching the rest, then do the new version when it comes out more slowly -- i.e do all the recommended exercises.

2022-06-27 23:25:03 @yabellini @Rami_Krispin @_useRconf How was it organised? What made it work?I'm very interested in finding ways to create compelling virtual conferences!

2022-06-27 23:08:05 Chapter 1 of the book is also available on @kaggle.(Oh and BTW, if you like my notebooks, please upvote them on Kaggle, because it helps others find them, and motivates me to do more because I know they're appreciated!)https://t.co/1nZChTCEJ9

2022-06-27 23:06:34 It's also available for free as @ProjectJupyter notebooks here:https://t.co/NfbwhBqWgl

2022-06-27 23:06:07 If you'd like to order a copy of the book yourself, here it is:https://t.co/rvRIqxrOgR

2022-06-27 23:02:58 There's lots more covered in the full chapter, so please do check it out -- and be sure to try executing it yourself so you can do your own experiments:https://t.co/fAsy4526NZ

2022-06-27 23:02:57 Of course, you don't have to create your model from scratch - here's how to do it in 2 lines of code with fastai: https://t.co/MQBts8lLOD

2022-06-27 23:02:56 We can also use the model to do a similarity search: https://t.co/TGzQWEHzNv

2022-06-27 23:02:55 We can also interpret the actual embedding matrices, by first extracting the principal components with PCA. Here's how some of the movies in our data are clustered based on this approach: https://t.co/FK4BETgJK9

2022-06-27 23:02:53 We can use these trained parameters to interpret the model - for instance, here are the world's most disappointing movies, according to this data: https://t.co/yaF6teKbWP

2022-06-25 05:06:27 @wasimlorgat @shawnup Same approach as @gradio

2022-06-25 02:46:36 RT @paulg: I suspect few Americans realize just how much of an outlier the US is when it comes to incarceration rates. More than half a per…

2022-06-25 00:14:13 RT @drkerrynphelps: I could even understand the #COVID19 #LetItRip “let GPs handle the fallout” mentality IF the vaccine stopped you gettin…

2022-06-25 00:11:35 @TheZachMueller @ThiagoTDotCom @DataChaz @github @gitpod @Replit @awscloud Paperspace gradient

2022-06-25 00:07:59 @shawnup Looks interesting!

2022-06-24 20:38:50 RT @Yash25571056: ".. similar to patients with HIV-1, the children previously infected by SARS-CoV-2 may have a repetitive immune activatio…

2022-06-24 20:29:28 @Suhail If you wait a couple of weeks you can do the new version. It's much better IMO

2022-06-24 10:22:39 RT @math_rachel: "Easy access to massive resources can stifle creativity. The world is a resource-constrained place, and ignoring that fact…

2022-06-24 01:09:45 RT @ricardoanderegg: As a big fan of SQLite, I always found the tooling to explore remote databases a bit lacking. Current solutions involv…

2022-06-24 01:07:40 @TheZachMueller @carrigmat @PyTorch I'm not sure that tells us that it's happening in hardware, does it?

2022-06-24 01:05:45 @pcuenq Congrats!

2022-06-24 01:05:25 When @pcuenq flew out to SF two years ago to hang out with us in our lab for a couple of months, I had a strong feeling that Camera+ was going to get some amazing AI capabilities in the not-too-distant future... And here they are! :D https://t.co/hGkSBHLVxP

2022-06-23 19:09:47 RT @tunguz: One sure way of knowing that you are an ML poser is if you feel it necessary to bash @Kaggle.

2022-06-23 18:55:25 RT @KostasPenn: #CVPR2022 Motion 3: "Any reviewer who has accepted an invitation to review but violates the reviewing guidelines set forth…

2022-06-23 05:44:08 RT @ChrisGr93091552: Explored github copilot,a paid service, to see if it encodes code from repositories w/ restrictive licenses.I checke…

2022-06-22 20:15:35 RT @wasimlorgat: Here's a tiny demo of an interactive file explorer with rich output powered by Jupyter notebooks + ipywidgets! So simple…

2022-06-22 19:35:10 RT @DataUmbrella: Join us: *online* working sessions to contribute to @pymc_devs #osswith a focus on underrepresented persons in #Dat…

2022-06-22 18:51:16 RT @what_is_the_why: @RageSheen My business is down 30% on where is was last year. Sadly I can't hold on any longer and have no choice but…

2022-06-22 18:45:50 RT @robreich: Can the GPT-4chan episode be counted as a part of the responsible practice of AI research?More than thirty AI researchers h…

2022-06-22 18:42:43 @willmcgugan The short one, but with a question mark on the end

2022-06-22 18:37:53 @wightmanr @huggingface That's very cool!

2022-06-22 06:00:41 RT @HamelHusain: Software 2.0: ML is eating softwareAlso Software 2.0: You need tons of additional software for ML. Please choose 15 t…

2022-06-22 06:00:17 RT @HamelHusain: I’m going to be announcing an epic new version of nbdev in this talk! Stay tuned The next version of nbdev is going to…

2022-06-22 05:59:19 RT @radekosmulski: How does noise affect the performance of your model?I am adding random noise to the dataset, one column at a time.xg…

2022-06-22 03:43:07 @willmcgugan I find the light easier to read. As I get older and my vision gets worse I find dark themes more difficult to read.

2022-06-22 03:39:50 RT @quarto_pub: @tracykteal Talk: Literate Programming w/ Jupyter Notebooks + Quarto by @HamelHusainThis talk will describe integration b…

2022-06-22 01:31:24 RT @irina_kAl: Big moment for #github #copilot! An incredible effort from a fabulous team, and I'm proud to have been part of it We've l…

2022-06-22 01:19:34 RT @radekosmulski: How to do well at @kaggle (or ML in general) by @jeremyphoward 1. Get good validation.2. Iterate quickly.3. Make thi…

2022-06-21 11:16:47 @LucaAmb Two of the people I most love are in vulnerable groups.

2022-06-21 09:02:41 RT @yaneerbaryam: Compared to the non-infected control group, those who only had one infection had an increased risk of at least one sequel…

2022-06-21 09:02:22 Oh. Shit. https://t.co/fTGT4AHfgm

2022-06-20 22:04:47 Part 3 of my journey to the top of the Kaggle leaderboard is now ready!https://t.co/q91g52aIi0

2022-06-20 22:02:55 You now have everything you need to replicate my submissions!In the next notebook, I'll show you how you can go further, and push me down the leaderboard... :D https://t.co/Q5gpd12gOn

2022-06-20 22:02:54 Finally, average the TTA predictions of each https://t.co/8FBgzvv0xM

2022-06-20 22:02:53 ...then we can create a dict of all the models we want to train, and train each in a little loop https://t.co/xEzWSTImpa

2022-06-20 22:02:52 We can test all our models ahead of time to see what level of accumulation, if any, is needed to run them on Kaggle's 16GB GPUs https://t.co/6VpELRY1XP

2022-06-20 22:02:51 In this step, we're scaling up to larger models and larger images. But that means we gotta deal with memory limitations...Luckily, that's easy to do, thanks to gradient accumulation. https://t.co/xH1qvfqYB2

2022-06-20 22:02:50 And now it's time for part 3, where we train our ensemble that topped the leaderboard (until one of my students passed me!)Here's the notebook -- click "Edit" and step through it yourself to see how it works:https://t.co/QscUgH3UdZ

2022-06-20 21:31:01 @simon_jegou @capetorch That would be very interesting! Please let us know if you try it!

2022-06-20 21:14:26 RT @giffmana: Haha shots fired https://t.co/CqEdIN20OX

2022-06-20 19:05:26 This looks awesome - even includes an APL track! :D https://t.co/zRlESP9VP0

2022-06-20 18:31:03 RT @LostInTangent: When viewing a file on https://t.co/2mCyd8BWBc, the pencil menu now provides an "Open in https://t.co/Am4EhpKJfJ" option…

2022-06-20 10:02:18 @kookcove To describe Peter as "non scientific" is patently ludicrous.He has been an extraordinary innovator over decades in multiple highly technical scientific fields.

2022-06-20 09:54:34 @willmcgugan Wow that's beautiful. I never cared much about terminal latency before...

2022-06-20 09:52:31 RT @EricRWeinstein: One wish for Father’s Day: That every Dad w a daughter ask one simple question: “Why are our news organizations not ask…

2022-06-20 09:46:13 RT @ianbremmer: us: left govt, high inflation uk: right govt, high inflationgermany: centrist govt, high inflation italy: everyone in go…

2022-06-20 09:44:21 RT @ScottGottliebMD: COVID is novel and pathogenic virus. Most kids will have mild course and recover fine from acute illness. But we shoul…

2022-06-20 09:44:17 RT @ScottGottliebMD: We always saw Epstein Barr Virus as reality of growing up, many teenagers got it. We never thought of vaxxing for it.…

2022-06-20 09:39:23 @flowermusickids @psvbluemts @YouAreLobbyLud @CountryBee2 A respirator is a type of mask. They can be referred to as "respirator masks" or "respiratory masks".When masks are requested something should be provided that gives appropriate protection. For covid, that means respiratory protection.

2022-06-20 09:29:18 RT @BowTiedCocoon: Something no one tells you when you're 20 years old..MD at investment bank= salesPartner at consulting firm= salesPa…

2022-06-20 09:24:19 RT @HamelHusain: This was a really great episode on how @borisdayma trained DALL-E mini Super interesting to hear the the process of repr…

2022-06-20 03:12:58 @marktenenholtz @iScienceLuvr @NmaViv @NmaViv what was the first notebook you published in this style? How did it evolve over time?

2022-06-20 03:03:24 @marktenenholtz Thanks!

2022-06-20 03:03:15 Seems the answer might be herehttps://t.co/DdLE9gVIlh

2022-06-20 00:26:31 If there was one main person that developed the original version of this, I'd love to know who it was so they can be properly credited by all the people they've inspired! And if it's been more of a gradual evolution, I'd be fascinated to see what that path has looked like.

2022-06-20 00:26:30 So, has this approach gradually evolved over time as lots of different people have added bits and pieces to some simple starting point over time?Or was there some great role model notebook created by a Kaggle grandmaster which everyone is now copying (perhaps indirectly)?

2022-06-20 00:26:29 These notebooks always start with a CFG or params dict, containing all the hyperparams etc.Everything is done totally from scratch in pure PyTorch, including mixup, fp16, etc. It includes ensembling over folds.It's kinda like a mini framework in a single notebook! https://t.co/sNaMP21iEa

2022-06-20 00:26:28 Question for Kagglers:Take a look at the notebook below. I've seen many notebooks looking nearly exactly like this from various folks in various competitions. I'm curious where the original version came from. https://t.co/lWppNrvSxF https://t.co/UYbFgA6LZm

2022-06-19 21:46:22 @ducha_aiki I have no idea what Apple's strategy for edge devices is... is it to select from one of their limited selection of models for which they've got OOTB API support?

2022-06-19 21:44:24 RT @nirav_mainecdc: 13/That chart is key. COVID causes more deaths/year than did all other diseases combined. If you're not in favor of…

2022-06-19 21:44:22 RT @nirav_mainecdc: 12/Now perhaps you turn to vaccines and say, "Well none of this means that we should vaccinate our kids, right?" Wron…

2022-06-19 10:41:14 @LangrJakub @ykilcher Thanks so much!

2022-06-19 10:41:03 Interesting discussion looking at evidence that Russian propaganda is probably being to at least some extent driven by bots -- perhaps including GPT-style deep learning models https://t.co/C2Eu8yKY6a

2022-06-19 09:54:18 Oh cool TIL there are fastai docker containers being maintained for the community! :D https://t.co/XhxeJGgjWK

2022-06-19 09:35:06 @fishnets88 @PyTorch That would be awesome, but I'm not sure if it's possible. JAX is a very different design to theano or tf, which are the main backends keras has had in the past.

2022-06-19 09:34:00 RT @tengelhr: “Our findings indicate that for children with a supportive family and school environment, a high number of screen hours is a…

2022-06-19 09:33:26 @mervenoyann That's true - but there will be attrition over time. Kinda like Internet Explorer...

2022-06-19 09:32:57 @francoisfleuret Yes that's fair.

2022-06-19 06:30:34 @aparachick @NjbBari3 @psvbluemts Also the flomask has a kids size. It's hard to find kids sizes of high quality masks.

2022-06-19 06:03:43 @iScienceLuvr @PyTorch I think they're really cool. But way harder to use than Keras, don't you think?

2022-06-19 06:01:19 RT @weights_biases: On the latest episode of Gradient Dissent, @l2k interviews @borisdayma about DALL-E mini! We dive deep into how Bori…

2022-06-19 05:18:27 @TheCodingProjec Thank you for saying that! :)

2022-06-19 04:38:57 My guess is that the net result of this is that @PyTorch will be increasingly adopted in industry.(Having said that, JAX is really cool, and perhaps the team there will find ways to make it just as accessible as TF/keras...)

2022-06-19 04:38:56 What pretty much everyone already knew was gonna happen, is now happening -- JAX is being gradually rolled out to replace TensorFlow (at least for internal use at Google). https://t.co/GrOLs7FX7o

2022-06-19 04:33:59 If you're following along with our "Road to the Top", where we show the process we used to get to #1 on a current Kaggle leaderboard, we've got good news - part 2 is now ready!(And if you haven't started yet -- it's not too late to join in...)https://t.co/o5Dc16euL3

2022-06-19 04:32:52 Put that all together, train for 12 epochs (~20 minutes on Kaggle), and we get 98.27% accuracy -- well inside the top 25% on the leaderboard! Here endeth of part 2 of our journey.https://t.co/7JV4okQlK7

2022-06-19 04:32:51 And for our final trick -- the power of `Learner.tta()`. Test time augmentation! https://t.co/WSLJHoYZWl

2022-06-19 04:32:50 Then I add another of my favourite tricks: zero-pad preprocessing https://t.co/9fRUcZhIYT

2022-06-19 04:32:49 Then I switch from resnet to my new fave architecture: ConvNeXt, and get a 2x boost in error rate!See this recent analysis for why: https://t.co/2dzJm94SAu https://t.co/4u86hRMvH4

2022-06-19 04:32:48 I start by using one of my favourite tricks -- preprocess the images to make them smaller. On Kaggle it makes our model train about 4x faster in this case, without any apparent loss of accuracy! https://t.co/HLemOvCNqN

2022-06-19 04:32:47 OK, part 2 is now ready for your reading pleasure: "Small Models".Join us as we figure out why yesterday's model was running so slowly on Kaggle, fix it, and then improve our error rate by >

2022-06-19 01:53:20 @LangrJakub @ykilcher I'd be really interested to see some examples that look GPT-ish, if you have any.

2022-06-19 01:33:56 Other big news in this release includes support for the (as yet unreleased) @PyTorch 1.12, and support for channels-last format (which can often be faster on modern GPUs).There's also an important workaround for a PyTorch performance bug that can result in 20-40% slowdowns.

2022-06-18 06:12:10 PS: this process will be heavily leveraging the work we recently did on studying the effectiveness of different architectures for transfer learning, so be sure to read that too if you haven't checked it out yethttps://t.co/pbMhQzv9SO

2022-06-18 06:02:55 I also share some tips I haven't mentioned before, such as how I interpret the learning rate finder https://t.co/6wJsra40ro

2022-06-18 06:02:54 I'll take you through the exact same steps I did myself to understand the data in this "Paddy Doctor" competition, such as parallel processing to quickly see image sizes https://t.co/QS1PJkUE5A

2022-06-18 06:02:53 You'll also see there the links to the over 6 hours of videos walkthrus where I'll take you through every step of the process.(This is a sneak peak at a new feature we're adding to the https://t.co/GEOZunWoXj course this year!) https://t.co/tEaQ7CwaAv

2022-06-18 06:02:52 The first thing you might notice if you click the link in the tweet above is that I've created `fastkaggle`, a little library of utilities to help kagglers -- especially to help running kaggle stuff on your own machinehttps://t.co/1zuZYyd11J https://t.co/EsDUUnucJo

2022-06-18 06:02:51 Are you ready to embark on a deep learning journey? I've just released over 6 hours of videos and the first in a series of notebooks showing the thought process of how I got to #1 in a current Kaggle comp.Follow this for updates on the journey!https://t.co/4NZ425XAO8 https://t.co/f0iFSr8LJk

2022-06-18 05:27:46 RT @breakfastnick: Harvard researchers find circulating spike protein in the blood of long haulers, but not controls, 12 months post diagno…

2022-06-18 03:47:23 @wasimlorgat @roholazandie Yes exactly. You wouldn't, for instance, generally define `mean()`, since you can implement it in 4 glyphs, which is the same length as the name, and involves no indirection!

2022-06-18 02:02:39 @jbspence @dyalogapl Never heard of that before - just ordered it!

2022-06-18 01:21:31 @jho_crypto @dyalogapl My daughter is 6 and her friend who I also tutor is 8

2022-06-18 00:19:08 @nomadiktrivedi @dyalogapl No I don't, because they're kids. It would be nice to do some video classes on math using APL though someday!

2022-06-18 00:08:21 RT @DrJaneMunro: Here we go AustraliaDespite the public silence on this this sub-variant is hitting our shoresGiven healthcare, education…

2022-06-18 00:07:35 (...actually I guess we'd also need "commute` operator to get rid of parens)

2022-06-18 00:05:38 Here's how we did `e` the other day (BTW if we move the last Ω to the left we could remove the parentheses here) https://t.co/d3h9pMg4yZ

2022-06-18 00:01:34 RT @code_star: The next wave of startups seems to be PhD Students dropping out to build MLOps companies because they got good at training m…

2022-06-17 23:56:47 @roholazandie I don't think it makes much sense to use APL in a regular IDE - it's not a good fit. It's designed for a REPL-driven approach, with fewer abstractions than other languages.

2022-06-17 23:46:05 We did a similar thing yesterday, but learning about how to calculate e, so they'd already seen limits before. When they saw limits here a 2nd time they were pretty excited!

2022-06-17 23:46:04 I love teaching math using @dyalogapl! Here's how I just explained derivatives to the kids for the first time - they totally got it right away :) https://t.co/9mNsTbxb4U

2022-06-17 23:45:27 RT @Globalbiosec: It was an honour to be on the WHO COVID Vaccine Composition Technical Advisory Group that contributed to these recommenda…

2022-06-17 17:31:47 @NeilJBlake @capetorch That's discussed in the analysis notebook linked in the last tweet.

2022-06-17 17:13:57 RT @tunguz: It’s always millionaires and billionaires who are unaffected by the high cost of living, have tons of personal assistants, live…

2022-06-17 17:01:45 RT @tmramalho: The next big breakthrough in AI will come from hardware, not software.Training giant models like PaLM already require 1000…

2022-06-17 16:58:30 @ph_singer @HeitorRapela @capetorch I'll try those LRs. Don't think we ever went that low even initially in the project.fastai uses a cosine schedule by default so all the runs used that.

2022-06-17 16:27:16 @ph_singer @capetorch We tried way lower LRs earlier in the project. I'll go back and check what happened there.

2022-06-17 16:25:53 @c_prohm @francoisfleuret @capetorch For sure.

2022-06-17 16:25:29 @c_prohm @francoisfleuret @capetorch Thanks I didn't know that. Will try it

2022-06-17 09:25:28 RT @nareshr8: Created a jupyter notebook extension on top of #NBDev from @fastdotai that basically converts the notebook to python code on…

2022-06-17 09:24:58 RT @johnowhitaker: I'm really loving the workflow I'm using for my current project:- Code I'll reuse goes in an #nbdev project. CI, versio…

2022-06-17 09:24:13 RT @HamelHusain: TIL more than 2.7k projects on Github are using #nbdev https://t.co/oZswC0o8u4h/t @TheZachMueller

2022-06-17 09:19:23 @vanstriendaniel @fastdotai Thanks!

2022-06-17 08:45:48 I've put together a writeup of the whole approach, and we've made the results and code publicly available to anyone who wants to dig in themselves:https://t.co/2dzJm8MJmm

2022-06-17 08:45:47 Overall, my conclusion from this analysis is that, although resnets are still pretty good, they no longer deserve to be our first choice when building a new model.Instead, try convnext (which is resolution independent!), vit, or swin/v2.https://t.co/uyR5zkhnFv

2022-06-17 08:45:46 However the relationship is much less clear on Planet. This suggests that if you're working with a small-ish dataset that's not similar to your pretrained model's dataset, you may want to focus on smaller/faster architectures. https://t.co/VVfn8ulSgU

2022-06-17 08:45:45 Here's the relationship between speed and accuracy on IIT-Pet. As you can see, using bigger/slower models pays off with better accuracy, but there are diminishing returns https://t.co/gp9tsBdBki

2022-06-17 08:45:43 IIT-Pet is a larger dataset, containing images (pet breeds) very similar to the dataset the model was trained on (imagenet).Here's the top 15 on Planet, which is a smaller dataset containing satellite imagery, which is unlike anything in imagenet: https://t.co/UGzMT01XZB

2022-06-17 08:45:41 We now finally have the thing I've always wanted: an analysis of the best vision models for finetuning.In joint work with @capetorch we tried nearly 100 models on two very different datasets across various hyperparams.Here's the top 15 on the IIT-Pet dataset: 1/ https://t.co/pgZ4DR2Qny

2022-06-17 01:53:17 RT @TheZachMueller: So @jeremyphoward and @HamelHusain's new nbprocess library. (or nbdev v2). It's going to be a *GAME CHANGER*!In a *si…

2022-06-16 19:30:15 Wow amazing:"Search for any scene in any movie as long as you can describe it with words… can handle a wide range of visual concepts: from people, objects and locations, to more abstract features such as emotions, vibes and religions… public figures and comic book superheroes" https://t.co/MbUiOz0PhX

2022-06-16 01:24:55 RT @fitterhappierAJ: You all need to understand this.The paper not only shows SARS Cov 2 mutating away from immunityBut also old infect…

2022-06-15 23:53:45 @tunguz @radekosmulski In Australia the regular banking system also works very well for regular consumer transaction stuff.I was so shocked when I moved to the US how awful the payment system was.

2022-06-15 21:39:37 RT @theinformation: EXCLUSIVE: Top executives at a prominent AI startup sold $32 million of shares as its valuation peaked last year. Emplo…

2022-06-15 21:38:23 RT @dgurdasani1: Important paper out in Science TLDR

2022-06-15 05:35:14 RT @espejelomar: @fastdotai X @hugginface Two of the best machine learning communities 1 study group (The Hugging Learners) from June…

2022-06-15 01:11:31 @wasimlorgat @TheZachMueller @HamelHusain It can test against the mime outputs too. That could be optional.

2022-06-14 21:35:39 @benjamin_warner @capetorch Here's the @PyTorch issue for anyone who wants to follow it:https://t.co/RvQmaLMlhX

2022-06-14 21:21:11 @TheZachMueller @HamelHusain @wasimlorgat I feel a bit like this too. I mentioned to Hamel the other day that I'm considering moving to the approach of having cell outputs be the test (i.e it tests that the cell output never changes).Perhaps a `#|test` directive to enable that? Or enable by default, optional `#|notest`

2022-06-14 21:19:10 @wasimlorgat @HamelHusain One thing I've wondered is whether all our https://t.co/GEOZuodrZj libs should do what ghapi does, and make the repr for the functions themselves show as markdown including a link to the docs (i.e the same info that `doc` shows).

2022-06-14 21:18:19 @wasimlorgat @HamelHusain I feel the same way, but I use `doc()` a lot too, because that provides a link to the full docs for a function which opens in a new tab. I find it doesn't disrupt my chain of thought because I just read that 1 thing then Ctrl-w back to my notebook.

2022-06-14 21:14:20 RT @benjamin_warner: Over the past week, @capetorch and I discovered, debugged, and created a workaround for a performance bug in PyTorch w…

2022-06-14 09:33:22 @MattAlhonte @wasimlorgat This is cool, but it doesn't generally work with very dynamic code such as ghapi. Or even stuff like fastcore's typedispatch (or plum dispatch).

2022-06-14 09:30:36 @wasimlorgat fastec2 is also interesting in that in both ipython and jupyter you get some very nice auto-completion dynamically of context-appropriate attributes.

2022-06-14 09:29:45 @wasimlorgat Check out ghapi's source code. You might be surprised by how tiny it is, yet covers the entire GitHub API. That's because it's auto-generated dynamically at runtime.It's also unusual that the repr of its functions are actually markdown including links to docs!

2022-06-14 05:50:01 @wasimlorgat Here's a few I've written with that in mind:https://t.co/2e8xIYECORhttps://t.co/EYnRmypExvhttps://t.co/Ak9Cq9qpSQ

2022-06-14 05:48:52 RT @wasimlorgat: I’d love to see more Python libs built with interactivity and notebook/repl-driven dev in mind Specifically, rich output…

2022-06-14 02:19:51 I'll be joined on the AMA panel by @alinabeetweets and @shakes76

2022-06-14 01:04:51 RT @tensoralex: I find #Apptainer (@SingularityApp ) convenient in day-to-day ML model development! So I composed a gentle introduction to…

2022-06-13 23:00:30 With colleagues from @UQSchoolITEE, I'm doing an AMA. Virtual event

2022-06-13 22:47:20 Perhaps it's true that "kindness and honesty is not the way of our modern workforce".But perhaps if we can live our lives as if that's not the case, we can bring a little change into the world.

2022-06-13 22:47:19 I got this email out of the blue today from someone I've never met.Honestly, it brought a tear to my eye. https://t.co/CrXbi6SPqX

2022-06-13 22:43:53 RT @modacitylife: To great uncertainty and controversy, in 2007, Ljubljana closed 12 hectares of its city center to private cars. Just 40%…

2022-06-13 16:04:54 RT @RaviHVJ: Today, I'm meant to be in California attending my graduation from Stanford University. However, I am too ill to go. I'm just 2…

2022-06-13 16:03:06 @hug_nicolas @rasbt @PyTorch Does `ShardingFilter` not resolve this?https://t.co/ib0QN1ddxR

2022-06-13 14:32:39 RT @girl_incognito: COVID and the Heart: It Spares No One | Johns Hopkins | Bloomberg School of Public Health https://t.co/iE4FymRKxt

2022-06-13 05:34:40 RT @alexmeshkin: I want the confidence of Twitter Virologists who got the reality of Omicron sub-variants and Hepatitis dead fucking wrong.

2022-06-12 23:36:33 RT @elisaperego78: I've just seen another verified Twitter account saying we can't keep viruses "at bay forever".Yet, yes! key research an…

2022-06-12 23:32:16 @sopranotiara @ucdavis Great job Tiara :)

2022-06-12 23:07:13 RT @DrZoeHyde: A study of the hepatitis cases seen in Israel reveals they were likely a delayed consequence of COVID-19.The average delay…

2022-06-12 21:18:46 @LostInTangent @johnkitchin @ProjectJupyter @github @HamelHusain Awesome! :D Lemme know if you have any questions about installing/deps/etc.

2022-06-12 21:17:17 @LostInTangent @johnkitchin @ProjectJupyter @github @HamelHusain Thanks @LostInTangent ! Is fastai one of the bunch on ML libs installed into the base image?

2022-06-11 18:31:51 RT @ruchowdh: New plan everyone - we slip Elon a few USB sticks and tell him it’s Twitter. He won’t know the difference. https://t.co/vRc…

2022-06-11 18:17:45 RT @JuggsMathis: @TurnerNovak WHAT THE FUCK https://t.co/UsHuKvibS0

2022-06-10 22:26:27 RT @chrisalbon: Imagine paying someone $150k-450k a year and then judging them based the number of times they move their mouse.

2022-06-10 22:13:34 @Hold2LLC The key one was opening borders at the end of 2021.

2022-06-10 17:14:49 RT @simonw: How much engineering time does your company lose to broken development environments?

2022-06-10 10:09:16 Just by looking at this chart of excess deaths you can quite precisely pinpoint the moment when Australia removed nearly all COVID public health measures. https://t.co/fglQdmGIr8

2022-06-10 10:03:22 RT @acweyand: Infants exposed in utero to #COVID19 almost 2x more likely to have developmental disorders with greatest effect seen with 3rd…

2022-06-10 04:15:25 https://t.co/M3FNkI4gj5

2022-06-10 02:18:35 @nmstoker Yes it's exactly that!

2022-06-09 22:33:02 @ihnorton You can use spawn instead of fork.

2022-06-09 22:31:35 @caiohamamura @ricardoanderegg You don't need websockets.

2022-06-09 22:28:53 RT @Globalbiosec: Hear my take on #COVID19, #vaccines, #masks, #LongCovid and "mystery" #hepatitis in kids and where we are heading with th…

2022-06-09 22:28:19 RT @andrey_kurenkov: GPT3 summing up this whole GPT-4chan situation pretty well tbh. @ykilcher https://t.co/NaMlysyR11

2022-06-09 22:26:29 @mediumnok @HamelHusain @OuterboundsHQ Look at the fastai/nbprocess repo. That'll become nbdev2. In general, our Discord server is where you can follow our work in real time.

2022-06-09 22:25:31 RT @HamelHusain: PSA: The most reliable way to quickly view any notebook on @github is to press the "." (period) key which opens up vscod…

2022-06-09 22:25:24 RT @giffmana: I agree. Personally I still like the term "pre-trained models". It's short, clear and to the point.The "large" part I feel…

2022-06-09 20:55:40 @michalwols Lazy imports are handy, but they'll only speed things up if you don't actually use some imports.

2022-06-09 20:55:01 @msjgriffiths Yeah something like that. But it only needs to maintain the sys.modules state, not the rest of globals

2022-06-09 20:42:05 I think there's something a bit like that for Julia and Perl already IIRC. I've looked around but can't find anything like it for Python however.The idea is that the overhead of starting the script only happens once.This is useful for stuff like git hook scripts.

2022-06-09 20:42:04 Python scripts can be slow to start because the imports can take a long time.I have lots of scripts that read/write stdin/out. I'd like to replace them with something with identical behavior, but which behind the scenes auto-launches a little server.Does that exist?

2022-06-09 19:17:06 RT @chrisalbon: The most under appreciated story in tech is how Kenya's digital currency, Mpesa, has become the de facto form of payment fo…

2022-06-09 04:59:37 @patrickmineault FYI you can use any jekyll theme on @github pages nowadayshttps://t.co/lxJdIqdUwB

2022-06-08 23:55:10 This is amazing - an in-browser sign language (ASL) recogniser built with @fastdotai, @PyTorch, and @unity.https://t.co/jC6bENNe1X

2022-06-08 23:52:18 Very cool - a 4-part blog series explaining how to use @fastdotai and @PyTorch with @unity! https://t.co/qcg8othKhF

2022-06-08 22:12:17 RT @alexmeshkin: What do we call the peer pressure from those that had COVID to get COVID?

2022-06-08 22:02:01 RT @robmarkcole: New discovery: Detecting Cloud Cover Via Sentinel-2 Satellite Data - My Top-10 Percent Solution to DrivenData’s On CloudN…

2022-06-08 21:40:21 This is so awesome! Check out these amazing examples of dynamic audio creation (be sure to turn on sound on the videos) https://t.co/M2O5TtSg96

2022-06-08 21:39:29 @benbovy @davidbrochart @martinRenou @jtpio Wow this is beautiful!

2022-06-08 21:36:40 @HamelHusain @shmeadyy @bearloga @github Wow so cool!

2022-06-08 07:39:36 RT @simonw: I asked GPT-3 to write a tutorial for getting started with Datasette, and a marketing landing page for my forthcoming Datasette…

2022-06-08 05:53:51 @bindureddy @MarwanElhelaly Isn't he out of OpenAI now?

2022-06-08 02:41:15 RT @bindureddy: The real danger we face with AI is not a terminator-style killer bot.Instead, it's the concentration of extreme power wit…

2022-06-07 20:47:41 RT @liviapuljak: Our new study shows that data availability statements are not very useful

2022-06-07 20:01:55 RT @HamelHusain: This docstring is a good example of how literate programming tools are valuable for documentation. The MosiacML folks car…

2022-06-07 17:46:23 RT @SMpwrgr: Want a 25-fold decrease in your cumulative dose of covid in the classroom? Open a window, put on a mask, turn on a hepa filter…

2022-06-07 02:04:31 RT @_joaogui1: Ok folks I've given all the benefit of the doubt I could give, @ykilcher has finetuned a GPT-J on the /pol/ board on 4chan,…

2022-06-06 23:30:20 @aniki_anon @ykilcher Lauren is a published researcher in the field of AI safety. Yannic is not.So whilst I'm not a fan of gatekeeping, if you're gonna do it, at least try to get it the right way around, @aniki_anon.(Better still, don't do it at all.)

2022-06-06 21:12:20 @benjamin_warner Fantastic job!

2022-06-06 21:12:05 This is a really valuable set of tools for anyone using fastai: https://t.co/2pU0wkyH3i

2022-06-06 06:47:12 RT @dbustac: I make a living building stuff with no-code.But I've never bought a “no-code course.”Because truth is, 95% of the informat…

2022-06-06 03:20:15 @rancoisse @psvbluemts You can divide the numbers shown at the right-hand side by the respective populations to see that it's a >

2022-06-06 03:19:36 @taipan168 @psvbluemts The "rate" is the slope of the curve. Australia is >

2022-06-06 02:53:32 RT @ch402: The elegance of ML is the elegance of biology, not the elegance of math or physics.Simple gradient descent creates mind-boggli…

2022-06-06 02:51:44 RT @equalityAlec: THREAD. Today, the New York Times published one of its most dishonest, biased, and dangerous pro-police articles that I h…

2022-06-06 02:48:44 @lmeyerov @Graphistry Yeah I've been thinking about this a lot too. It feels like there's a whole industry lurking in there somewhere...

2022-06-06 02:48:23 RT @lmeyerov: embeddings as a primitive building block has become such a huge conceptual part of designing data pipelines &

2022-06-05 17:50:11 RT @Rainmaker1973: Scientists found that RNA, an analog of DNA that was likely the first genetic material for life, spontaneously forms on…

2022-06-04 23:35:28 @wasimlorgat fastcore's implementation is limited to max 2 typed params. And frankly it was kinda hacked together -- I suspect plum's design might be allow for some more interesting patterns...

2022-06-04 23:33:27 @WesBarnettPhD @math_rachel Wow fantastic article. Thanks for sharing, and thanks @Scudellari and @PNASNews !

2022-06-04 23:32:54 RT @Scudellari: And now, I bring you a peek under Omicron's hood -- including some explanations for why it is likely the fastest spreading…

2022-06-04 23:10:20 RT @math_rachel: Q: is there a good article debunking misunderstandings of the hygiene hypothesis? Eg explaining that catching every virus…

2022-06-04 23:07:30 RT @adamdavidson: I wrote something provocative about how I know stuff about Jeffrey Epstein and can't publish it. I'll do my best to exp…

2022-06-04 23:07:25 RT @seoyoung_chu: @adamdavidson @alexanderchee Thank you for this thread. To this day I am haunted by the fact that so many powerful peop…

2022-06-04 02:40:12 RT @ToshiAkima: Italian study shows ventilation can cut school COVID cases by 82% | Reuters https://t.co/LQ8nX7Mvju

2022-06-04 01:27:57 RT @marjakh: Off by one

2022-06-04 01:11:38 RT @ramsri_goutham: With all due respect to the open-source contributions, the inference API of @huggingface is robbery in broad daylight.…

2022-06-04 01:09:36 RT @adamdavidson: Whenever Jeffrey Epstein's name comes up, I'm reminded (by some people on twitter) that many don't know the basics. His…

2022-06-04 01:07:34 RT @CrabbBrendan: A strong study in one of the world's best journals in support of mask wearing in the community to reduce transmission of…

2022-06-04 00:11:33 RT @b0rk: @larrywright you can see an example of a frontend project here I did with no build tooling: https://t.co/2qPFvK6og5 (it's just 4…

2022-06-03 19:18:42 Currently `plum` doesn't support future annotations so we're blocked from using it, but hopefully that'll be fixed sometime soonish (according to the GitHub issue it's not too hard).

2022-06-03 19:18:41 We're considering replacing fastcore's typedispatch with this.Would be interested to hear thoughts from anyone using `plum` on how it's working. https://t.co/LFjh0YmYJr

2022-06-03 17:45:06 RT @adamdavidson: I spent two years working on the Jeffrey Epstein podcast, Broken. I became close with many of his victims and learned a l…

2022-06-03 03:16:47 RT @yotammann: Excited to tease the browser-based, A.I. studio that i've been building with some incredible collaborators: https://t.co/lbx…

2022-06-03 03:13:24 This looks interesting -- SAM is cool, but complex to implement and slow. If this is really a "free" way to get the same benefits, that would be huge!Anyone tried it? How did it go? https://t.co/aCwN0zbHlp

2022-06-03 03:07:09 RT @PMGPSC: Herd immunity was sold as the path out of the pandemic. This article is in line with what ⁦@RealOzSAGE⁩ has been saying from d…

2022-06-03 02:26:12 RT @UQSchoolITEE: Next up on our #UQAI #AMA series - the topic of #DeepLearning. Featuring #UQITEE's Hon Prof @jeremyphoward, Dr @AlinaBe…

2022-06-02 23:55:48 @tunguz Thanks for that Bojan - very useful.

2022-06-02 23:54:58 RT @bhutanisanyam1: Today is a glorious day for @kaggle community! Kaggle legend: @sudalairajkumar has conquered all categories and beco…

2022-06-02 21:49:13 @tunguz What were the red flags you noticed in the paper?

2022-06-02 21:45:47 RT @OuterboundsHQ: We are using #nbdev, a literate programming framework from @fastdotai, to power our docs system. Code snippets &

2022-06-02 21:45:33 RT @OuterboundsHQ: Announcing https://t.co/YngTfx2tAW: a set of guides, tutorials, and examples on how to deploy a wide range of ML produ…

2022-06-02 21:45:06 RT @HamelHusain: Remember this wishlist for docs? It didn't exist, so we built it using #nbdev @OuterboundsHQ Docs are top of the fun…

2022-06-02 06:16:45 RT @espejelomar: @fastdotai @huggingface @jeremyphoward As a plus...You can combine NLP Transformers and @fastdotai using blurr. Then s…

2022-06-02 06:15:55 RT @espejelomar: More @fastdotai + @huggingface Did you train an image classifier using fastai? Share it with the community in the Hu…

2022-06-02 06:15:22 @IAmSamFin @DrCMcMaster Thanks for the practical tip mate

2022-06-02 06:14:42 RT @DrCMcMaster: Euclidean distance = liberalManhattan distance = conservative

2022-06-01 20:16:50 RT @nearcyan: heavenbanning, the hypothetical practice of banishing a user from a platform by causing everyone that they speak with to be r…

2022-06-01 18:51:47 RT @TechEmails: Elon Musk on remote workMay 31, 2022 https://t.co/gTw1Bdh18h

2022-06-01 18:30:18 RT @Jobvo: Tech CEOs wanting to go back to the office. https://t.co/rFKGiF9JBj

2022-05-31 21:52:35 RT @radekosmulski: nvt op Tuesday! How do you deal with variables that have a lot of categories?Before we look at solutions unique to…

2022-05-31 21:52:25 @radekosmulski Very happy to see you mention embedding hashing. I've been telling anyone that'll listen for the last 4 years that that's probably what we should all be doing, but never managed to find anyone that is interested!

2022-05-31 04:45:38 @renegadesilicon @radekosmulski Like literally... *all* country kids!

2022-05-31 04:45:11 @renegadesilicon @radekosmulski Dude I *went* to boarding school...

2022-05-31 02:18:23 @ThiagoTDotCom In the browser I can just use plain Jupyter -- don't even need Streamlit! But looking for doing stuff in a script in the terminal.

2022-05-31 02:17:17 @radekosmulski @renegadesilicon First I've heard of it... You sure you're not inventing stuff now Nick?

2022-05-30 22:56:35 RT @ACTINOSProject: So that’s not great, then…https://t.co/Fmav8f9lTx

2022-05-30 19:04:07 RT @HamelHusain: A related project: mdseo mdseo helps you analyze content in markdown files for SEO rules. The motivation for this is to…

2022-05-30 11:17:06 RT @HamelHusain: wp2md: a small python package that converts Wordpress articles to markdown. Built with #nbdevIt's a good example of a s…

2022-05-30 11:09:54 @karpathy @DrRaviPatelJr @weights_biases One could just as easily say it enriches the namespace, rather than pollutes it.And do you really use an editor that doesn't tell you where your symbols come from?

2022-05-30 11:07:14 @rizandiparnadi @karpathy @DrRaviPatelJr @weights_biases How much F# code have you written? How confident are you in your expertise in that language that you can be so very sure of yourself in making such a claim?

2022-05-30 11:05:49 @joh_hof @karpathy @DrRaviPatelJr @weights_biases Verbosity can often lead of less clear code, more maintenance, and resultant tech debt.Be careful of promoting verbosity as a virtue.Consider the elegance of beautiful math notation for example.

2022-05-30 11:03:16 @lukaszjocz @giffmana @karpathy @DrRaviPatelJr @weights_biases ... Which can be used to good effect.

2022-05-30 11:02:27 @karpathy @DrRaviPatelJr @weights_biases And in some other languages the official guidance is the opposite.The truth is that there's pros and cons and the best approach is dependent on the situation.https://t.co/l9e0fK2oVg

2022-05-29 23:14:26 RT @Miles_Brundage: AI Twitter in 2025 is going to be like: “OK this book/movie/videogame it generated is decent but clearly it was cherryp…

2022-05-29 23:03:29 @gsw7 That's a good point.

2022-05-29 21:56:41 Oh wow TIL the book "C++ Coding Standards" actually say `using` statements "are what make namespaces usable"!When I did a lot of C++ coding I always got yelled at when I added `using`. Turns out folks on the C++ committee actually approve of them.https://t.co/OEniqebSQH https://t.co/56l3ol94As

2022-05-29 21:53:15 @wwbrannon AFAIK no, it doesn't.

2022-05-29 21:52:31 @DrRaviPatelJr @karpathy @weights_biases BTW the "wildcard imports bad" thing is a fiction that the python community has told itself. Lots of languages do the exact opposite and it works just fine.Different situations have different needs. There's no one correct way to import symbols.https://t.co/TIe3t5BigN

2022-05-29 21:44:24 Ideally, I'd love to see programming communities have more awareness that different approaches can be better or worse for different people in different situations.Both the F# (global imports are good) and Python/C++ (are bad) extremes seem like they could do with more nuance...

2022-05-29 21:44:23 Interestingly, communities where REPL usage is front and center, like Julia, F#, R, and nbdev seem to lean towards greater usage of global symbol imports.

2022-05-29 21:44:22 In Python, by default even imported stuff is re-exported as if it's defined in your module. That means that `import *` can suddenly bring in dozens of extra symbols that you don't want.This can be fixed by defining `__all__`, but almost no-one does that.

2022-05-29 21:44:21 In C++ if you put `using` in a header, then it pollutes the whole project. So changing just one file can silently change what symbol is found everywhere in the project!

2022-05-29 21:44:20 F# takes the opposite approach. Whilst you can use fully qualified symbols, the official docs say that might be "hard to write, read, and maintain".https://t.co/7IJGiwT2oq https://t.co/G3SQ11yrKv

2022-05-29 21:44:19 The situation in C++ is similar -- you can import a namespace, some symbols in a namespace, or all symbols in a namespace, although the latter is widely frowned upon in the community. (I'm not sure if there's any "official" C++ guidance on this)https://t.co/LjGkyQiGAy https://t.co/WDxFkRgzU5

2022-05-29 21:44:18 Python provides a lot of flexibility in allowing imports of modules, submodules, or some or all symbols within a module. However importing all symbols ("wildcard import") is frowned upon by the official docshttps://t.co/zYnBIEk6L0 https://t.co/wSW9Q2jyRb

2022-05-29 21:44:16 The situation in #rstats is similar -- imported namespaces bring all exported functions into the global symbol table, with the option to prefix namespace names if wanted or neededhttps://t.co/g0p6VOw0YI https://t.co/YjtLrm5LoS

2022-05-29 21:44:15 In #julialang the most common approach is `using`, which brings all exported symbols from a module into the global ("parent") namespacehttps://t.co/utjU1W5nIQ https://t.co/JHY0WkKK9H

2022-05-29 21:44:14 Different programming languages have very different ways of handling importing symbols, and very different cultures around what is the "correct" way to do it.Yet each culture seems to have very strong feelings that *their* way is the one *true* way.A of examples:

2022-05-29 19:38:29 RT @dpkingma: It was 16 years ago, in 2006, that @geoffreyhinton et al released their demo of deep belief nets. Undergrad me was highly imp…

2022-05-29 18:30:40 RT @psvbluemts: @drkerrynphelps Too many elderly are dying due to denial of antiviral treatment. No matter how old you are, if you are VACC…

2022-05-29 07:32:35 This is the best answer so far:https://t.co/RFhc7HFSqL

2022-05-29 07:32:18 @fakufakurevenge It works!

2022-05-29 07:19:48 @fakufakurevenge No I haven't seen that one. Will give it a try

2022-05-29 07:10:07 @rrherr Doesn't look like librosa has a way to play sound...?

2022-05-29 07:07:58 @IAkhil3 That's true, it does install from the wheels just fine. I guess I could convert it to a conda package easily enough...

2022-05-29 07:06:08 @LostInTangent @github @greybaker @jhutchings0 That would be a bit tricky - it would require actually auto-editing https://t.co/IBoca0HfUU somehow whenever the user changes their settings.ini file. Not impossible, but not ideal either...

2022-05-29 07:04:48 @mihaimaruseac @github Did you find a good way to test whether it was working?

2022-05-29 07:04:23 @WNoxchi Yeah it's easy to play audio in a notebook, since that's in a browser. Not a bad option if you only need something in the browser, although I was hoping for something I could make into a script.

2022-05-29 07:03:17 @matdmiller Yeah that's not a bad option, although VLC is not installable thru conda afaict unfortunately.

2022-05-29 04:49:12 @nagaraj_arvind Thanks! How about codeSpark?

2022-05-29 04:47:48 @LostInTangent @github Can you think of anything we could do in our https://t.co/IBoca0HfUU or elsewhere in nbdev that would allow GitHub to include these projects in the dep graph?

2022-05-29 04:46:57 @radamar @Raspberry_Pi I'm not sure how a Raspberry Pi would help fix the lack of a conda package?

2022-05-29 04:46:16 @rohitsince1996 That needs ffmpeg I think. It also didn't play the whole sound when I used it in jupyter.

2022-05-29 04:22:25 @jacobeisenstein I haven't written anything public. If they're starting from scratch, I found codeSpark good, and similar iPad games.

2022-05-29 04:21:40 @crypto_reflect I don't know - I don't have a Mac.

2022-05-29 04:21:18 @radamar I couldn't find an easy way to install it. No current conda package afaict.

2022-05-29 04:13:12 Does anyone know how the @github "Used By" feature is populated, and why projects created with #nbdev don't appear there?

2022-05-29 04:08:00 All the things I found on stackoverflow/google either don't have up to date conda or pypi packages, or require a full dev environment since they need to be compiled, or don't work any more, or require something like ffplay to be installed first.

2022-05-29 04:07:59 I wanted to create a fun Python script with my daughter that plays a random audio file. But I couldn't find any easy to install lib that plays audio files!The stdlib has `winsound` which is great for windows. But what's good for Mac/Linux/x-platform?

2022-05-27 05:44:06 When lame machine learning failures happen in TV sports analytics, it's an amusing anecdote.But the same thing happens in predictive policing, bond and sentence recommendations, health insurance algorithms blocking needed care, etc...

2022-05-27 05:44:05 Yet another failure from a shitty proprietary closed-source "machine learning" system.There's no reason to believe such a system should work well. So why do people buy them?Even people like Apple, who have hundreds of staff that could create something better in a few days. https://t.co/3CI4cYVOcD

2022-05-27 05:36:15 RT @yishan: Here's a stat that's a bit politically inconvenient for almost everyone:Since 2020, there have been more children killed by C…

2022-05-27 05:23:34 RT @neuralreckoning: The current system of journals and peer review is not serving science. I have therefore resigned from all editorial ro…

2022-05-26 23:41:25 RT @joeyfox85: HEPA filter placement guidelines:1. Best to have one that blows air up2. If blowing sideways, best not to aim it at anyon…

2022-05-26 23:40:43 I've been hearing this a lot recently too.It seems there may have been a significant change in outdoor transmissibility in recent variants. https://t.co/2D1tfMBtFj

2022-05-26 21:06:32 RT @Fra_Pochetti: New post alert! How about a UNET segmentation model converted to TensorRT and running at 16ms latency on @nvidi…

2022-05-26 19:36:11 @DrRaviPatelJr @karpathy @weights_biases The python standard library itself has just such a module BTW, and the official tutorial recommends using wildcard importhttps://t.co/cH3aeG0vac

2022-05-26 19:10:57 @DrRaviPatelJr @karpathy @weights_biases Yes I'm sure he'd be amazed to see how all fastai libs carefully set __all__ to make wildcard import safe. Quite eye-popping!

2022-05-26 19:01:41 RT @takesthecake2: The ventilation "audits" in NSW schools were a sham.https://t.co/nTk1r1oMgn

2022-05-26 18:58:46 @SylvainCorlay @pydatalondon @ProjectJupyter Sounds great - would love to see a recording

2022-05-26 18:58:15 RT @francoisfleuret: Don't want to boast, but ain't it a beauty? https://t.co/kd309t9AQX

2022-05-26 18:55:09 RT @weights_biases: How to use the WandbCallback to organise your fastai experiments - Pass a WandbCallback to cbs when you call .fit on…

2022-05-25 23:42:04 Fine-Tuning Transformers: Vocabulary Transfer https://t.co/JDmnMseAAq

2022-05-25 21:45:25 RT @timnitGebru: 4-5 years ago @mmitchell_ai was a semifinalist for the MIT Tech Review 35 under 35. I wrote a letter of support. One of th…

2022-05-25 21:43:22 @waydegilliam @bhutanisanyam1 Who's your fave interviewer on Coffee Time Data Science

2022-05-25 04:18:01 RT @RisingSayak: https://t.co/BRWjIaAbQ9 materials have this unique ability to introduce someone to research, math-heavy topics in the most…

2022-05-25 04:17:56 RT @RisingSayak: If you've some experience in Python, I'd suggest taking the _free_ @fastdotai course - "Practical Deep Learning for Coders…

2022-05-25 04:14:28 RT @radekosmulski: nvt op Tuesday! An op that is a true game-changer in my books!+ submitting to a @Kaggle competition+ looking at a…

2022-05-25 04:11:35 @RageSheen @jvedmond In theory. But I can't help thinking they're gonna fuck it up anyway...

2022-05-25 02:25:07 @savvyRL Fair enough. (We left America FWIW...)

2022-05-25 02:17:00 @YouAreLobbyLud @NjbBari3 I gave up and now I do other things.You must be more optimistic than me, I think.

2022-05-25 02:15:57 @savvyRL Wow I hadn't seen those tweets. I can't imagine too many researchers would see those and think "that's someone I'd really love to work for"...

2022-05-24 21:53:43 RT @CDCgov: New @CDCMMWR: Approximately 1 in 5 adults ages 18+ experience a new condition a month or more after surviving #COVID19. More: h…

2022-05-24 21:51:10 RT @adrianzenz: BREAKING: huge trove of files obtained by hacking into Xinjiang police / re-education camp computers contain first-ever ima…

2022-05-24 21:47:22 RT @joeyliaw: https://t.co/Zrm4rblGQ9 Google announced Imagen today. Let's generate some of the same prompts with Dall·E 2!

2022-05-24 17:48:32 RT @The_Law_Boy: each sentence a unique slice of our shared hell https://t.co/XdmabkZI6d

2022-05-24 07:11:19 @karpathy BTW in fastai this whole thing becomes less of an issue since the main classes use fastcore's `store_attr()` to keep track of everything that impacts the model and/or data.It works particularly well with @weights_biases since their integration automatically tracks all this.

2022-05-24 07:03:43 RT @PyTorch: Today we’d like to highlight features from functorch, a beta PyTorch library that provides JAX-inspired function transformatio…

2022-05-24 06:59:39 RT @EmilWallner: It's LIVE: **No ML Degree**My 60-page guide on how to land your first machine learning job without a degree.Amazon:…

2022-05-24 06:59:29 @EmilWallner Looks great! Do you mind if I quote from it in our course (with attribution of course)?

2022-05-24 06:54:05 @karpathy I just use call_parse from fastcore and do whatever else is needed in Python. For running experiments I combine it with fastgpu (https://t.co/3BfXh8h1CM).https://t.co/cfxLpIWs8U

2022-05-24 02:32:24 Very cool to see @fastdotai integration with Unity! https://t.co/zZ4DupFOPb

2022-05-20 08:11:00 CAFIAC FIX

2022-10-29 02:36:06 This is one of the most important developments I've seen yet in the generative modeling world. At least as important as the original release of the models, I think. https://t.co/ERtVw0vVgS

2022-10-28 22:46:05 RT @emollick: Years of job experience is one of the WORST predictors of future performance for new hires. Asking for X years of experienc…

2022-10-28 19:10:11 RT @ClementDelangue: We just crossed 1,000,000 downloads of Stable Diffusion on the @huggingface hub! Congrats to Robin Rombach, Patrick Es…

2022-10-28 19:02:52 RT @HamelHusain: We are going to go dive into fairly topics, such as:- The "Just use the software eng workflow" gaslighting of ML pr…

2022-10-28 01:50:19 RT @wasimlorgat: You do not want to miss out on @radekosmulski's newsletter!I met Radek at @fastdotai conf. He is the most enthusiastic p…

2022-10-27 20:35:12 RT @untitled01ipynb: https://t.co/0j6HRlPd1p

2022-10-27 18:49:58 RT @untitled01ipynb: https://t.co/zYtKa0xUIC

2022-10-27 18:46:17 If you want to deepen your knowledge of the memes, here's the full videohttps://t.co/zwP3rINSCm

2022-10-27 18:44:47 I'm not sure what's worse- my entire carefully written lesson, watched at 2x, has been turned into a series of memes- the meme version might be just as educational as the original video https://t.co/UHzP8HdtAt

2022-10-27 05:38:24 RT @iScienceLuvr: This is currently the most important network in deep learning!From helping to power search for billions of users to bet…

2022-10-27 05:12:13 RT @radekosmulski: I am a bit late to the party, but the first lecture is a brilliant overview of all the components of Stable Diffusion.…

2022-10-27 01:46:27 @DSaience I will certainly do that

2022-10-26 23:27:27 @DSaience Oh and my daughter told me to say she thinks it's fun too!

2022-10-26 23:27:03 @DSaience I love it! :D

2022-10-26 19:47:38 RT @SimonWallace: Finally, a special thank you to @jeremyphoward, @HamelHusain, and @wasimlorgat for their work on nbdev. It’s an incredibl…

2022-10-26 19:47:35 RT @SimonWallace: Obiter is an open-source Python library of computational research tools. You can use the library to interact with @CanLii…

2022-10-26 17:33:26 RT @isaac_flath: What are programming things you think the data science community as a whole could use more or understand better to improve…

2022-10-26 17:15:07 RT @Robin_Tian: Amos is a new optimizer that we propose to pre-train large language models. It is more efficient and converges faster than…

2022-10-26 17:12:54 RT @alexmeshkin: I haven't disclosed this previously as I didn't want to get into the debate about the severity of COVID. Yes, I had a BA…

2022-10-26 17:10:32 RT @bot_fra: Here my attempt to apply Random Forest on a practical Industrial issue: classify Aluminium scraps boxes. I've figured out to b…

2022-10-26 17:03:24 @JayAlammar Thanks Jay!

2022-10-26 17:02:10 Interesting new podcast that discusses my tweet, questioning whether adding numpy-style broadcasting rules to APL would be a good idea https://t.co/ssWQD2Nj54

2022-10-26 16:59:04 @dyalogapl @RikedyP @a_brudz Thanks for the tip - I look forward to watching

2022-10-26 16:56:38 RT @agadmator: Magnus blunders a bishop on move 1! https://t.co/pzMFEt6idW

2022-10-24 06:02:28 RT @strickvl: This week in the @fastdotai part 2 course we started the exciting work of implementing our own deep learning library from scr…

2022-10-24 04:46:15 RT @JohnVial: How One University Researcher Used His Teaching Load To Create New Research https://t.co/SpFOWL7HYW

2022-10-24 02:47:55 @0Kiriyama @mhajabri @Fra_Pochetti @fastdotai @radekosmulski @marktenenholtz @iScienceLuvr some of it is out nowhttps://t.co/LiUu9HSflG

2022-10-24 01:35:11 RT @JACoates: Can we just talk about the reality of peer review for a second because so many scientists here seem completely oblivous to th…

2022-10-23 08:29:37 RT @TheZachMueller: Announcing preorders for Walk with @fastdotai, the missing pieces for success!If you want to take your fastai skills…

2022-10-23 02:10:10 @chrisalbon The bots will flock to reddit though, won't they?

2022-10-22 22:33:52 RT @DorsaAmir: How do you map out an ocean voyage without a map? Here’s an extremely cool example of a stick chart used by the Indigenous…

2022-10-22 21:10:26 This is a great summary of some of the main points from our first "Foundations to Stable Diffusion" lesson: https://t.co/neU9a7Txcy

2022-10-22 16:41:30 RT @TerribleMaps: The “international community” https://t.co/f9iCpZwD6H

2022-10-22 16:37:57 RT @TomLikesRobots: I've been testing the two new 1.5 checkpoints against the 1.4 versions at different steps.Faces are slightly more real…

2022-10-22 16:26:44 RT @quocleix: New open-source language model from Google AI: Flan-T5 Flan-T5 is instruction-finetuned on 1,800+ language tasks, leading…

2022-10-22 16:25:37 @_jasonwei @quocleix It got over 1k, so you gotta open something else now.

2022-10-22 16:23:33 RT @kprather88: Someone asked me why I feel one reason for the slow admission of #COVIDIsAirborne is due to a COI. Answer: Because the cha…

2022-10-22 16:18:18 RT @UrosZiv: @CDCDirector https://t.co/go5LAoPeDZ

2022-10-22 06:52:10 RT @johnowhitaker: This space is still new. Which means ideas like this, where smart people go 'this seems obvious, I wonder if anyone has…

2022-10-22 05:28:18 @kalpesh_ai @johnrobinsn Not really at the moment, no.

2022-10-22 03:52:18 I suggested to our students that they should try doing CLIP interpolation, &

2022-10-22 03:52:14 There's a trick we can use however - it's possible to use an init image as the starting point for image generation, instead of random noise. I explain the technique here:https://t.co/QpTLeALB9A

2022-10-22 03:52:13 The basic idea, illustrated below by @fastdotai student @cleavey1985, is to interpolate between the CLIP embeddings of multiple prompts, passing each to Stable Diffusion to generate images.But as you see, this leads to choppy output! https://t.co/2LKSYb3cCv

2022-10-22 03:52:10 Have a look at this amazing and beautiful video of a tree as it goes through the seasons.Can you guess how it was made? Read on! https://t.co/i5STjBI0sX

2022-10-22 02:11:47 RT @bhutanisanyam1: Some pillars of the @fastdotai community:@wasimlorgat, @radekosmulski, @HamelHusain, @jeremyphoward, @TheZachMueller…

2022-10-22 01:43:18 RT @ai_fast_track: You probably never heard of MMRotateHere are some reasons why you should be familiar with:• MMRotate is an open-so…

2022-10-21 22:27:54 RT @renegadesilicon: Friends, go read this post from my friend @Angelina_Magr on tools for data scientists — in particular the phenomenal n…

2022-10-21 22:24:57 @DienhoaT Key clarification: we subtract only a bit of the noise, then we re-run it. We do that a few times.

2022-10-21 18:59:14 RT @isaac_flath: I gave 1 of my Mom’s paintings to the Clip Interrogator. Part of the response was a name, Franklin Carmichael. I asked h…

2022-10-21 08:13:50 RT @iScienceLuvr: I'm glad to be part of the team developing the latest iteration of the @fastdotai Part 2 course, From Deep Learning Found…

2022-10-21 06:41:07 @SuryaGanguli Einstein said that relativity may have never been developed if it wasn't for Humehttps://t.co/hZX5pULQUf

2022-10-21 06:31:03 @JimGolden23 It's gonna need to be a big coffee...

2022-10-29 20:49:37 RT @karpathy: Thanks Lex, I've enjoyed many of the previous episodes so it was a pleasure to come on! (we've known each other from before…

2022-10-29 18:43:11 @cgarciae88 @agifirealarm @jxmnop Yup the SDE formulation seems to work better - as discussed in Karras et alhttps://t.co/c4oOA5O23S

2022-10-29 18:29:46 RT @alexmeshkin: Unpopular Opinion: A pathogen that has a 'low' acute death rate with an high transmission rate that causes chronic illne…

2022-10-29 18:23:07 RT @CaseyNewton: Here it is: Twitter engineers were told today to *print out* their last 30 to 60 days of code, so they could show it to El…

2022-10-29 18:20:10 @agifirealarm @jxmnop It takes quite a bit of thought to come up with an f such that f(f(x)) works well.

2022-10-29 18:19:39 @docmilanfar @agifirealarm @jxmnop I'm not sure it's quite that simple. f(f(x)) doesn't really make sense for lots of models. And not all diffusion models continually add and remove noise (only SDE-style does

2022-10-29 18:14:40 RT @jxmnop: Diffusion is just an easy-to-optimize way to give neural networks adaptive computation time.Makes sense then that diffusion m…

2022-10-29 18:14:16 @radekosmulski Very helpful!

2022-10-29 18:13:58 RT @radekosmulski: Let me see if I can take notes on Twitter while working through the @fastdotai notebooks...If it won't work, I'll just…

2022-10-29 18:12:16 RT @davisblalock: "What Makes Convolutional Models Great on Long Sequence Modeling?"CNNs—not transformers—now dominate the hardest sequen…

2022-10-29 17:29:22 RT @StabilityAI: We recently released two new fine-tuned decoders (trained by the amazing @rivershavewings) that improves the quality of im…

2022-10-29 02:36:06 This is one of the most important developments I've seen yet in the generative modeling world. At least as important as the original release of the models, I think. https://t.co/ERtVw0vVgS

2022-10-28 22:46:05 RT @emollick: Years of job experience is one of the WORST predictors of future performance for new hires. Asking for X years of experienc…

2022-10-28 19:10:11 RT @ClementDelangue: We just crossed 1,000,000 downloads of Stable Diffusion on the @huggingface hub! Congrats to Robin Rombach, Patrick Es…

2022-10-28 19:02:52 RT @HamelHusain: We are going to go dive into fairly topics, such as:- The "Just use the software eng workflow" gaslighting of ML pr…

2022-10-28 01:50:19 RT @wasimlorgat: You do not want to miss out on @radekosmulski's newsletter!I met Radek at @fastdotai conf. He is the most enthusiastic p…

2022-10-27 20:35:12 RT @untitled01ipynb: https://t.co/0j6HRlPd1p

2022-10-27 18:49:58 RT @untitled01ipynb: https://t.co/zYtKa0xUIC

2022-10-27 18:46:17 If you want to deepen your knowledge of the memes, here's the full videohttps://t.co/zwP3rINSCm

2022-10-27 18:44:47 I'm not sure what's worse- my entire carefully written lesson, watched at 2x, has been turned into a series of memes- the meme version might be just as educational as the original video https://t.co/UHzP8HdtAt

2022-10-27 05:38:24 RT @iScienceLuvr: This is currently the most important network in deep learning!From helping to power search for billions of users to bet…

2022-10-27 05:12:13 RT @radekosmulski: I am a bit late to the party, but the first lecture is a brilliant overview of all the components of Stable Diffusion.…

2022-10-27 01:46:27 @DSaience I will certainly do that

2022-10-26 23:27:27 @DSaience Oh and my daughter told me to say she thinks it's fun too!

2022-10-26 23:27:03 @DSaience I love it! :D

2022-10-26 19:47:38 RT @SimonWallace: Finally, a special thank you to @jeremyphoward, @HamelHusain, and @wasimlorgat for their work on nbdev. It’s an incredibl…

2022-10-26 19:47:35 RT @SimonWallace: Obiter is an open-source Python library of computational research tools. You can use the library to interact with @CanLii…

2022-10-26 17:33:26 RT @isaac_flath: What are programming things you think the data science community as a whole could use more or understand better to improve…

2022-10-26 17:15:07 RT @Robin_Tian: Amos is a new optimizer that we propose to pre-train large language models. It is more efficient and converges faster than…

2022-10-26 17:12:54 RT @alexmeshkin: I haven't disclosed this previously as I didn't want to get into the debate about the severity of COVID. Yes, I had a BA…

2022-10-26 17:10:32 RT @bot_fra: Here my attempt to apply Random Forest on a practical Industrial issue: classify Aluminium scraps boxes. I've figured out to b…

2022-10-26 17:03:24 @JayAlammar Thanks Jay!

2022-10-26 17:02:10 Interesting new podcast that discusses my tweet, questioning whether adding numpy-style broadcasting rules to APL would be a good idea https://t.co/ssWQD2Nj54

2022-10-26 16:59:04 @dyalogapl @RikedyP @a_brudz Thanks for the tip - I look forward to watching

2022-10-26 16:56:38 RT @agadmator: Magnus blunders a bishop on move 1! https://t.co/pzMFEt6idW

2022-10-24 06:02:28 RT @strickvl: This week in the @fastdotai part 2 course we started the exciting work of implementing our own deep learning library from scr…

2022-10-24 04:46:15 RT @JohnVial: How One University Researcher Used His Teaching Load To Create New Research https://t.co/SpFOWL7HYW

2022-10-24 02:47:55 @0Kiriyama @mhajabri @Fra_Pochetti @fastdotai @radekosmulski @marktenenholtz @iScienceLuvr some of it is out nowhttps://t.co/LiUu9HSflG

2022-10-24 01:35:11 RT @JACoates: Can we just talk about the reality of peer review for a second because so many scientists here seem completely oblivous to th…

2022-10-23 08:29:37 RT @TheZachMueller: Announcing preorders for Walk with @fastdotai, the missing pieces for success!If you want to take your fastai skills…

2022-10-23 02:10:10 @chrisalbon The bots will flock to reddit though, won't they?

2022-10-22 22:33:52 RT @DorsaAmir: How do you map out an ocean voyage without a map? Here’s an extremely cool example of a stick chart used by the Indigenous…

2022-10-22 21:10:26 This is a great summary of some of the main points from our first "Foundations to Stable Diffusion" lesson: https://t.co/neU9a7Txcy

2022-10-22 16:41:30 RT @TerribleMaps: The “international community” https://t.co/f9iCpZwD6H

2022-10-22 16:37:57 RT @TomLikesRobots: I've been testing the two new 1.5 checkpoints against the 1.4 versions at different steps.Faces are slightly more real…

2022-10-22 16:26:44 RT @quocleix: New open-source language model from Google AI: Flan-T5 Flan-T5 is instruction-finetuned on 1,800+ language tasks, leading…

2022-10-22 16:25:37 @_jasonwei @quocleix It got over 1k, so you gotta open something else now.

2022-10-22 16:23:33 RT @kprather88: Someone asked me why I feel one reason for the slow admission of #COVIDIsAirborne is due to a COI. Answer: Because the cha…

2022-10-22 16:18:18 RT @UrosZiv: @CDCDirector https://t.co/go5LAoPeDZ

2022-10-22 06:52:10 RT @johnowhitaker: This space is still new. Which means ideas like this, where smart people go 'this seems obvious, I wonder if anyone has…

2022-10-22 05:28:18 @kalpesh_ai @johnrobinsn Not really at the moment, no.

2022-10-22 03:52:18 I suggested to our students that they should try doing CLIP interpolation, &

2022-10-22 03:52:14 There's a trick we can use however - it's possible to use an init image as the starting point for image generation, instead of random noise. I explain the technique here:https://t.co/QpTLeALB9A

2022-10-22 03:52:13 The basic idea, illustrated below by @fastdotai student @cleavey1985, is to interpolate between the CLIP embeddings of multiple prompts, passing each to Stable Diffusion to generate images.But as you see, this leads to choppy output! https://t.co/2LKSYb3cCv

2022-10-22 03:52:10 Have a look at this amazing and beautiful video of a tree as it goes through the seasons.Can you guess how it was made? Read on! https://t.co/i5STjBI0sX

2022-10-22 02:11:47 RT @bhutanisanyam1: Some pillars of the @fastdotai community:@wasimlorgat, @radekosmulski, @HamelHusain, @jeremyphoward, @TheZachMueller…

2022-10-22 01:43:18 RT @ai_fast_track: You probably never heard of MMRotateHere are some reasons why you should be familiar with:• MMRotate is an open-so…

2022-10-21 22:27:54 RT @renegadesilicon: Friends, go read this post from my friend @Angelina_Magr on tools for data scientists — in particular the phenomenal n…

2022-10-21 22:24:57 @DienhoaT Key clarification: we subtract only a bit of the noise, then we re-run it. We do that a few times.

2022-10-21 18:59:14 RT @isaac_flath: I gave 1 of my Mom’s paintings to the Clip Interrogator. Part of the response was a name, Franklin Carmichael. I asked h…

2022-10-21 08:13:50 RT @iScienceLuvr: I'm glad to be part of the team developing the latest iteration of the @fastdotai Part 2 course, From Deep Learning Found…

2022-10-21 06:41:07 @SuryaGanguli Einstein said that relativity may have never been developed if it wasn't for Humehttps://t.co/hZX5pULQUf

2022-10-21 06:31:03 @JimGolden23 It's gonna need to be a big coffee...

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-20 00:19:59 RT @BriannaWu: This is amazing. Elno broke the copyright strike system, so you can just post entire films on Twitter now.

2022-11-19 18:22:55 RT @rahaeli: Hey, US folks newly running Mastodon instances: do Future You a *huge* favor, mitigate your potential liability, and register…

2022-11-19 03:25:23 RT @DataChaz: Still not sure about Mastodon? Come join @hamel's Discord instead! There's an ungodly amount of cool data peeps in there…

2022-11-19 03:17:44 RT @BlancheMinerva: I am very excited to announce that a joint proposal w/ @AiEleuther @Mila_Quebec and @laion_ai to train Foundation Model…

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-21 02:37:44 RT @jasoncbenn: LLMs are much more capable than anyone initially appreciated. GPT-3, by default, is playing some kind of weird improv game…

2022-11-21 02:30:59 RT @_akpiper: in my #AI and #Literature class we will be spending the next 2 weeks diving into AI co-writing tools. My first impression is…

2022-11-21 02:30:18 RT @EMostaque: Should we put a @paperswithcode Galactica 120bn demo back online for science? Assuming no objections from @MetaAI given lic…

2022-11-21 00:05:05 RT @ajankelo: Props to @HamelHusain for creating the most interesting discord for #MLOps and tooling. Excellent: discussion people…

2022-11-20 18:47:27 RT @untitled01ipynb: Can you fall in love with a monkey patch? https://t.co/sxVxRdEsmw https://t.co/NibMhD1Wjg

2022-11-20 00:19:59 RT @BriannaWu: This is amazing. Elno broke the copyright strike system, so you can just post entire films on Twitter now.

2022-11-19 18:22:55 RT @rahaeli: Hey, US folks newly running Mastodon instances: do Future You a *huge* favor, mitigate your potential liability, and register…

2022-11-19 03:25:23 RT @DataChaz: Still not sure about Mastodon? Come join @hamel's Discord instead! There's an ungodly amount of cool data peeps in there…

2022-11-19 03:17:44 RT @BlancheMinerva: I am very excited to announce that a joint proposal w/ @AiEleuther @Mila_Quebec and @laion_ai to train Foundation Model…

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-22 05:15:18 RT @ameliaelizalde: found out that ‘arvo’ is australian for ‘afternoon’. when will it end

2022-11-22 05:14:47 RT @RCoops1978: In case you get confused by Aussie responses to questions. .. Note, the simplest way is to just understand the final word…

2022-11-22 05:10:51 @Jousefm2 @keinname42 @jezza_art @jagarikin I don't understand why you're responding like this. You made a fairly minor mistake but your follow-up is not good at all.

2022-11-22 03:59:08 RT @JJitsev: We are in somewhat mild, but very pleasant shock - our LAION-5B paper has just received Outstanding Paper Award at NeurIPS 202…

2022-11-22 03:31:50 @realGeorgeHotz I have pinboard ingest my favorites so I get to search them. Would be nice if this was built in.

2022-11-22 03:29:26 RT @EricTopol: Speech and spread of #SARSCoV2 "Our results suggest that the quantity of speech aerosol, in particular in the ≥ 4-μm range,…

2022-11-21 23:47:49 RT @abouelatta_ali: This is one of the experiments we ran on the Duolingo Retention team : https://t.co/Dv9Wp377vT

2022-11-21 23:46:11 @renegadesilicon @amasad Has there been any twin studies that have looked at outstanding parenting?

2022-11-21 21:01:35 Obstructive sleep apnea (OSA) strongly correlates with snoring. Flu pneumonia hospitalization in highly CPAP-adherent OSA cohort is ~ 5 times less likely, but no correlation with OSA severity. This indicates CPAP benefit by eliminating snoring rather than by minimizing OSA https://t.co/rJaDvJJWZR

2022-11-21 21:01:34 Snoring only carries risk of causing pneumonia when nose/throat is infected. Risk can be strongly reduced by methods that prevent snoring: https://t.co/nqj4dKHXad

2022-11-21 21:01:32 The new study shows that the volume of inhaled snoring droplets is comparable to prior radiotracer studies, implying that these fluids enter the lungs through snoring rather than by “microaspiration”. https://t.co/pXdaP3We4c

2022-11-21 21:01:31 However, only microscopic particles carried by inhaled air, can enter the deep lung and cause viral pneumonia, which is the major cause of covid-19 deaths https://t.co/P3M5Fdussc

2022-11-21 21:01:30 Transfer of fluids from the pharynx into the lungs is common and is known to happen at night during sleep, as documented by radiotracer studies https://t.co/MVX6pTj9cQ

2022-11-21 21:01:28 Microaspiration during sleep is the accepted “hand-waving” mechanism for transfer of microbes from the oral cavity into the lung https://t.co/LdUHJan2MZ

2022-11-21 21:01:27 Infection of the lung with SARS-CoV-2 is a two-step process: first the nose / throat, then the lungs. Postulated, but physically implausible, mechanism for step 2 involves “microaspiration” https://t.co/UjppvrBSeY

2022-11-21 21:01:26 Intriguing new study from the amazing Adriaan Bax and team suggests that most covid deaths resulted from (preventable) snoring droplets rather than (unpreventable) microaspiration. This could be a game changer. No time for the paper? Then read this ! https://t.co/OhOWDSFveI

2022-11-21 07:56:57 @radekosmulski @fastdotai Here's a link to the docs for execnb :D https://t.co/tkLNiTtY3y

2022-11-21 07:55:29 RT @radekosmulski: I just gained a superpower Thank you so much @fastdotai for execnb! I can now: • treat my notebooks as inputs • m…

2022-11-21 02:37:44 RT @jasoncbenn: LLMs are much more capable than anyone initially appreciated. GPT-3, by default, is playing some kind of weird improv game…

2022-11-21 02:30:59 RT @_akpiper: in my #AI and #Literature class we will be spending the next 2 weeks diving into AI co-writing tools. My first impression is…

2022-11-21 02:30:18 RT @EMostaque: Should we put a @paperswithcode Galactica 120bn demo back online for science? Assuming no objections from @MetaAI given lic…

2022-11-21 00:05:05 RT @ajankelo: Props to @HamelHusain for creating the most interesting discord for #MLOps and tooling. Excellent: discussion people…

2022-11-20 18:47:27 RT @untitled01ipynb: Can you fall in love with a monkey patch? https://t.co/sxVxRdEsmw https://t.co/NibMhD1Wjg

2022-11-20 00:19:59 RT @BriannaWu: This is amazing. Elno broke the copyright strike system, so you can just post entire films on Twitter now.

2022-11-19 18:22:55 RT @rahaeli: Hey, US folks newly running Mastodon instances: do Future You a *huge* favor, mitigate your potential liability, and register…

2022-11-19 03:25:23 RT @DataChaz: Still not sure about Mastodon? Come join @hamel's Discord instead! There's an ungodly amount of cool data peeps in there…

2022-11-19 03:17:44 RT @BlancheMinerva: I am very excited to announce that a joint proposal w/ @AiEleuther @Mila_Quebec and @laion_ai to train Foundation Model…

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-22 05:15:18 RT @ameliaelizalde: found out that ‘arvo’ is australian for ‘afternoon’. when will it end

2022-11-22 05:14:47 RT @RCoops1978: In case you get confused by Aussie responses to questions. .. Note, the simplest way is to just understand the final word…

2022-11-22 05:10:51 @Jousefm2 @keinname42 @jezza_art @jagarikin I don't understand why you're responding like this. You made a fairly minor mistake but your follow-up is not good at all.

2022-11-22 03:59:08 RT @JJitsev: We are in somewhat mild, but very pleasant shock - our LAION-5B paper has just received Outstanding Paper Award at NeurIPS 202…

2022-11-22 03:31:50 @realGeorgeHotz I have pinboard ingest my favorites so I get to search them. Would be nice if this was built in.

2022-11-22 03:29:26 RT @EricTopol: Speech and spread of #SARSCoV2 "Our results suggest that the quantity of speech aerosol, in particular in the ≥ 4-μm range,…

2022-11-21 23:47:49 RT @abouelatta_ali: This is one of the experiments we ran on the Duolingo Retention team : https://t.co/Dv9Wp377vT

2022-11-21 23:46:11 @renegadesilicon @amasad Has there been any twin studies that have looked at outstanding parenting?

2022-11-21 21:01:35 Obstructive sleep apnea (OSA) strongly correlates with snoring. Flu pneumonia hospitalization in highly CPAP-adherent OSA cohort is ~ 5 times less likely, but no correlation with OSA severity. This indicates CPAP benefit by eliminating snoring rather than by minimizing OSA https://t.co/rJaDvJJWZR

2022-11-21 21:01:34 Snoring only carries risk of causing pneumonia when nose/throat is infected. Risk can be strongly reduced by methods that prevent snoring: https://t.co/nqj4dKHXad

2022-11-21 21:01:32 The new study shows that the volume of inhaled snoring droplets is comparable to prior radiotracer studies, implying that these fluids enter the lungs through snoring rather than by “microaspiration”. https://t.co/pXdaP3We4c

2022-11-21 21:01:31 However, only microscopic particles carried by inhaled air, can enter the deep lung and cause viral pneumonia, which is the major cause of covid-19 deaths https://t.co/P3M5Fdussc

2022-11-21 21:01:30 Transfer of fluids from the pharynx into the lungs is common and is known to happen at night during sleep, as documented by radiotracer studies https://t.co/MVX6pTj9cQ

2022-11-21 21:01:28 Microaspiration during sleep is the accepted “hand-waving” mechanism for transfer of microbes from the oral cavity into the lung https://t.co/LdUHJan2MZ

2022-11-21 21:01:27 Infection of the lung with SARS-CoV-2 is a two-step process: first the nose / throat, then the lungs. Postulated, but physically implausible, mechanism for step 2 involves “microaspiration” https://t.co/UjppvrBSeY

2022-11-21 21:01:26 Intriguing new study from the amazing Adriaan Bax and team suggests that most covid deaths resulted from (preventable) snoring droplets rather than (unpreventable) microaspiration. This could be a game changer. No time for the paper? Then read this ! https://t.co/OhOWDSFveI

2022-11-21 07:56:57 @radekosmulski @fastdotai Here's a link to the docs for execnb :D https://t.co/tkLNiTtY3y

2022-11-21 07:55:29 RT @radekosmulski: I just gained a superpower Thank you so much @fastdotai for execnb! I can now: • treat my notebooks as inputs • m…

2022-11-21 02:37:44 RT @jasoncbenn: LLMs are much more capable than anyone initially appreciated. GPT-3, by default, is playing some kind of weird improv game…

2022-11-21 02:30:59 RT @_akpiper: in my #AI and #Literature class we will be spending the next 2 weeks diving into AI co-writing tools. My first impression is…

2022-11-21 02:30:18 RT @EMostaque: Should we put a @paperswithcode Galactica 120bn demo back online for science? Assuming no objections from @MetaAI given lic…

2022-11-21 00:05:05 RT @ajankelo: Props to @HamelHusain for creating the most interesting discord for #MLOps and tooling. Excellent: discussion people…

2022-11-20 18:47:27 RT @untitled01ipynb: Can you fall in love with a monkey patch? https://t.co/sxVxRdEsmw https://t.co/NibMhD1Wjg

2022-11-20 00:19:59 RT @BriannaWu: This is amazing. Elno broke the copyright strike system, so you can just post entire films on Twitter now.

2022-11-19 18:22:55 RT @rahaeli: Hey, US folks newly running Mastodon instances: do Future You a *huge* favor, mitigate your potential liability, and register…

2022-11-19 03:25:23 RT @DataChaz: Still not sure about Mastodon? Come join @hamel's Discord instead! There's an ungodly amount of cool data peeps in there…

2022-11-19 03:17:44 RT @BlancheMinerva: I am very excited to announce that a joint proposal w/ @AiEleuther @Mila_Quebec and @laion_ai to train Foundation Model…

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-25 04:06:55 RT @fede_gr: The entirety of @nixtlainc's time series ecosystem is built on top of #nbdev. We code, test, publish and release in notebooks.…

2022-11-25 04:03:19 RT @SteveStuWill: Wow! Scientists have observed a single-cell alga evolve in real time into a multicellular organism. The transition took a…

2022-11-25 03:57:30 RT @johnowhitaker: These models can make such amazing things. But they've also trained on lots of 'meh' stock photos etc. Something like th…

2022-11-25 00:46:11 RT @lukefinleyedit: @ginnyhogan_ https://t.co/o3YJJVR3nU

2022-11-25 00:44:32 RT @mikevanbus: This is a “promoted tweet” on my timeline: https://t.co/PJiO4iZyZv

2022-11-25 00:43:50 RT @iScienceLuvr: "Thanksgiving Dinner" generated by @StableDiffusion 2.0 https://t.co/JkRogEbNv7

2022-11-25 00:42:39 RT @LyskovAndrei: Haven’t run my own experiments yet but seems like commentary about the new @StabilityAI‘s 2.0 model is negative. Thread…

2022-11-24 23:11:20 RT @aronchick: my god GraphQL is unusable. I am so fucking frustrated the @github folks moved over from REST to it without giving us an SDK…

2022-11-24 23:03:37 @radekosmulski @shendusuipian @fastdotai Have you checked out the ability also to inject code into a cell to parameterise notebooks?

2022-11-24 23:00:56 RT @DelJohnsonVC: U.S. Venture capital hasn't been able to outperform European VC for the past 20 years. The Thiel Andreessen generation of…

2022-11-24 07:33:54 RT @_akhaliq: Thinking of retiring from paper tweets it’s pretty time intensive on top of a full time job, I had a good run so far. Plus al…

2022-11-22 05:15:18 RT @ameliaelizalde: found out that ‘arvo’ is australian for ‘afternoon’. when will it end

2022-11-22 05:14:47 RT @RCoops1978: In case you get confused by Aussie responses to questions. .. Note, the simplest way is to just understand the final word…

2022-11-22 05:10:51 @Jousefm2 @keinname42 @jezza_art @jagarikin I don't understand why you're responding like this. You made a fairly minor mistake but your follow-up is not good at all.

2022-11-22 03:59:08 RT @JJitsev: We are in somewhat mild, but very pleasant shock - our LAION-5B paper has just received Outstanding Paper Award at NeurIPS 202…

2022-11-22 03:31:50 @realGeorgeHotz I have pinboard ingest my favorites so I get to search them. Would be nice if this was built in.

2022-11-22 03:29:26 RT @EricTopol: Speech and spread of #SARSCoV2 "Our results suggest that the quantity of speech aerosol, in particular in the ≥ 4-μm range,…

2022-11-21 23:47:49 RT @abouelatta_ali: This is one of the experiments we ran on the Duolingo Retention team : https://t.co/Dv9Wp377vT

2022-11-21 23:46:11 @renegadesilicon @amasad Has there been any twin studies that have looked at outstanding parenting?

2022-11-21 21:01:35 Obstructive sleep apnea (OSA) strongly correlates with snoring. Flu pneumonia hospitalization in highly CPAP-adherent OSA cohort is ~ 5 times less likely, but no correlation with OSA severity. This indicates CPAP benefit by eliminating snoring rather than by minimizing OSA https://t.co/rJaDvJJWZR

2022-11-21 21:01:34 Snoring only carries risk of causing pneumonia when nose/throat is infected. Risk can be strongly reduced by methods that prevent snoring: https://t.co/nqj4dKHXad

2022-11-21 21:01:32 The new study shows that the volume of inhaled snoring droplets is comparable to prior radiotracer studies, implying that these fluids enter the lungs through snoring rather than by “microaspiration”. https://t.co/pXdaP3We4c

2022-11-21 21:01:31 However, only microscopic particles carried by inhaled air, can enter the deep lung and cause viral pneumonia, which is the major cause of covid-19 deaths https://t.co/P3M5Fdussc

2022-11-21 21:01:30 Transfer of fluids from the pharynx into the lungs is common and is known to happen at night during sleep, as documented by radiotracer studies https://t.co/MVX6pTj9cQ

2022-11-21 21:01:28 Microaspiration during sleep is the accepted “hand-waving” mechanism for transfer of microbes from the oral cavity into the lung https://t.co/LdUHJan2MZ

2022-11-21 21:01:27 Infection of the lung with SARS-CoV-2 is a two-step process: first the nose / throat, then the lungs. Postulated, but physically implausible, mechanism for step 2 involves “microaspiration” https://t.co/UjppvrBSeY

2022-11-21 21:01:26 Intriguing new study from the amazing Adriaan Bax and team suggests that most covid deaths resulted from (preventable) snoring droplets rather than (unpreventable) microaspiration. This could be a game changer. No time for the paper? Then read this ! https://t.co/OhOWDSFveI

2022-11-21 07:56:57 @radekosmulski @fastdotai Here's a link to the docs for execnb :D https://t.co/tkLNiTtY3y

2022-11-21 07:55:29 RT @radekosmulski: I just gained a superpower Thank you so much @fastdotai for execnb! I can now: • treat my notebooks as inputs • m…

2022-11-21 02:37:44 RT @jasoncbenn: LLMs are much more capable than anyone initially appreciated. GPT-3, by default, is playing some kind of weird improv game…

2022-11-21 02:30:59 RT @_akpiper: in my #AI and #Literature class we will be spending the next 2 weeks diving into AI co-writing tools. My first impression is…

2022-11-21 02:30:18 RT @EMostaque: Should we put a @paperswithcode Galactica 120bn demo back online for science? Assuming no objections from @MetaAI given lic…

2022-11-21 00:05:05 RT @ajankelo: Props to @HamelHusain for creating the most interesting discord for #MLOps and tooling. Excellent: discussion people…

2022-11-20 18:47:27 RT @untitled01ipynb: Can you fall in love with a monkey patch? https://t.co/sxVxRdEsmw https://t.co/NibMhD1Wjg

2022-11-20 00:19:59 RT @BriannaWu: This is amazing. Elno broke the copyright strike system, so you can just post entire films on Twitter now.

2022-11-19 18:22:55 RT @rahaeli: Hey, US folks newly running Mastodon instances: do Future You a *huge* favor, mitigate your potential liability, and register…

2022-11-19 03:25:23 RT @DataChaz: Still not sure about Mastodon? Come join @hamel's Discord instead! There's an ungodly amount of cool data peeps in there…

2022-11-19 03:17:44 RT @BlancheMinerva: I am very excited to announce that a joint proposal w/ @AiEleuther @Mila_Quebec and @laion_ai to train Foundation Model…

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-26 00:47:44 https://t.co/ClyXR3RfXE

2022-11-26 00:46:59 @ximix84 @capetorch @wasimlorgat Me three

2022-11-26 00:42:53 RT @giffmana: Anyone used FocalLoss from RetinaNet successfully for highly imbalanced binary task? Besides detection which it was originall…

2022-11-25 15:28:48 https://t.co/U3t9PMl0ak

2022-11-25 15:27:16 RT @bhutanisanyam1: We will be re-starting Top @kaggle Solutions Study group this weekend Join me on Sunday to learn from: “Open Proble…

2022-11-25 13:30:48 RT @ducha_aiki: I am in dark mood today, as continues terrorist strikes on . However let’s talk about academia. Out of 93700 followe…

2022-11-25 13:30:10 This is amazing - a single awesome dev in his spare time has massively upgraded a giant e-commerce company's search engine, thanks to deep learning and #nbdev https://t.co/pANohhY7mI

2022-11-25 13:26:10 RT @JayAlammar: Big update to "The Illustrated Stable Diffusion" post https://t.co/sbjKHP5ndT 14 new and updated visuals. The biggest u…

2022-11-25 04:06:55 RT @fede_gr: The entirety of @nixtlainc's time series ecosystem is built on top of #nbdev. We code, test, publish and release in notebooks.…

2022-11-25 04:03:19 RT @SteveStuWill: Wow! Scientists have observed a single-cell alga evolve in real time into a multicellular organism. The transition took a…

2022-11-25 03:57:30 RT @johnowhitaker: These models can make such amazing things. But they've also trained on lots of 'meh' stock photos etc. Something like th…

2022-11-25 00:46:11 RT @lukefinleyedit: @ginnyhogan_ https://t.co/o3YJJVR3nU

2022-11-25 00:44:32 RT @mikevanbus: This is a “promoted tweet” on my timeline: https://t.co/PJiO4iZyZv

2022-11-25 00:43:50 RT @iScienceLuvr: "Thanksgiving Dinner" generated by @StableDiffusion 2.0 https://t.co/JkRogEbNv7

2022-11-25 00:42:39 RT @LyskovAndrei: Haven’t run my own experiments yet but seems like commentary about the new @StabilityAI‘s 2.0 model is negative. Thread…

2022-11-24 23:11:20 RT @aronchick: my god GraphQL is unusable. I am so fucking frustrated the @github folks moved over from REST to it without giving us an SDK…

2022-11-24 23:03:37 @radekosmulski @shendusuipian @fastdotai Have you checked out the ability also to inject code into a cell to parameterise notebooks?

2022-11-24 23:00:56 RT @DelJohnsonVC: U.S. Venture capital hasn't been able to outperform European VC for the past 20 years. The Thiel Andreessen generation of…

2022-11-24 07:33:54 RT @_akhaliq: Thinking of retiring from paper tweets it’s pretty time intensive on top of a full time job, I had a good run so far. Plus al…

2022-11-22 05:15:18 RT @ameliaelizalde: found out that ‘arvo’ is australian for ‘afternoon’. when will it end

2022-11-22 05:14:47 RT @RCoops1978: In case you get confused by Aussie responses to questions. .. Note, the simplest way is to just understand the final word…

2022-11-22 05:10:51 @Jousefm2 @keinname42 @jezza_art @jagarikin I don't understand why you're responding like this. You made a fairly minor mistake but your follow-up is not good at all.

2022-11-22 03:59:08 RT @JJitsev: We are in somewhat mild, but very pleasant shock - our LAION-5B paper has just received Outstanding Paper Award at NeurIPS 202…

2022-11-22 03:31:50 @realGeorgeHotz I have pinboard ingest my favorites so I get to search them. Would be nice if this was built in.

2022-11-22 03:29:26 RT @EricTopol: Speech and spread of #SARSCoV2 "Our results suggest that the quantity of speech aerosol, in particular in the ≥ 4-μm range,…

2022-11-21 23:47:49 RT @abouelatta_ali: This is one of the experiments we ran on the Duolingo Retention team : https://t.co/Dv9Wp377vT

2022-11-21 23:46:11 @renegadesilicon @amasad Has there been any twin studies that have looked at outstanding parenting?

2022-11-21 21:01:35 Obstructive sleep apnea (OSA) strongly correlates with snoring. Flu pneumonia hospitalization in highly CPAP-adherent OSA cohort is ~ 5 times less likely, but no correlation with OSA severity. This indicates CPAP benefit by eliminating snoring rather than by minimizing OSA https://t.co/rJaDvJJWZR

2022-11-21 21:01:34 Snoring only carries risk of causing pneumonia when nose/throat is infected. Risk can be strongly reduced by methods that prevent snoring: https://t.co/nqj4dKHXad

2022-11-21 21:01:32 The new study shows that the volume of inhaled snoring droplets is comparable to prior radiotracer studies, implying that these fluids enter the lungs through snoring rather than by “microaspiration”. https://t.co/pXdaP3We4c

2022-11-21 21:01:31 However, only microscopic particles carried by inhaled air, can enter the deep lung and cause viral pneumonia, which is the major cause of covid-19 deaths https://t.co/P3M5Fdussc

2022-11-21 21:01:30 Transfer of fluids from the pharynx into the lungs is common and is known to happen at night during sleep, as documented by radiotracer studies https://t.co/MVX6pTj9cQ

2022-11-21 21:01:28 Microaspiration during sleep is the accepted “hand-waving” mechanism for transfer of microbes from the oral cavity into the lung https://t.co/LdUHJan2MZ

2022-11-21 21:01:27 Infection of the lung with SARS-CoV-2 is a two-step process: first the nose / throat, then the lungs. Postulated, but physically implausible, mechanism for step 2 involves “microaspiration” https://t.co/UjppvrBSeY

2022-11-21 21:01:26 Intriguing new study from the amazing Adriaan Bax and team suggests that most covid deaths resulted from (preventable) snoring droplets rather than (unpreventable) microaspiration. This could be a game changer. No time for the paper? Then read this ! https://t.co/OhOWDSFveI

2022-11-21 07:56:57 @radekosmulski @fastdotai Here's a link to the docs for execnb :D https://t.co/tkLNiTtY3y

2022-11-21 07:55:29 RT @radekosmulski: I just gained a superpower Thank you so much @fastdotai for execnb! I can now: • treat my notebooks as inputs • m…

2022-11-21 02:37:44 RT @jasoncbenn: LLMs are much more capable than anyone initially appreciated. GPT-3, by default, is playing some kind of weird improv game…

2022-11-21 02:30:59 RT @_akpiper: in my #AI and #Literature class we will be spending the next 2 weeks diving into AI co-writing tools. My first impression is…

2022-11-21 02:30:18 RT @EMostaque: Should we put a @paperswithcode Galactica 120bn demo back online for science? Assuming no objections from @MetaAI given lic…

2022-11-21 00:05:05 RT @ajankelo: Props to @HamelHusain for creating the most interesting discord for #MLOps and tooling. Excellent: discussion people…

2022-11-20 18:47:27 RT @untitled01ipynb: Can you fall in love with a monkey patch? https://t.co/sxVxRdEsmw https://t.co/NibMhD1Wjg

2022-11-20 00:19:59 RT @BriannaWu: This is amazing. Elno broke the copyright strike system, so you can just post entire films on Twitter now.

2022-11-19 18:22:55 RT @rahaeli: Hey, US folks newly running Mastodon instances: do Future You a *huge* favor, mitigate your potential liability, and register…

2022-11-19 03:25:23 RT @DataChaz: Still not sure about Mastodon? Come join @hamel's Discord instead! There's an ungodly amount of cool data peeps in there…

2022-11-19 03:17:44 RT @BlancheMinerva: I am very excited to announce that a joint proposal w/ @AiEleuther @Mila_Quebec and @laion_ai to train Foundation Model…

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-28 07:21:32 RT @Mylovanov: Good evening. Day 5 after the latest Russian attack on Kyiv. Day 277 of the war. I am president of the Kyiv School of Econom…

2022-11-28 05:39:23 RT @rdmorris: Overwhelming evidence of outdoor transmission of Omicron. At least 32 unmasked people were infected by a man jogging through…

2022-11-28 05:35:26 RT @Suhail: Headed on a week long trip to completely focus on and study AI tomorrow. Would love any links, code, papers to read related to…

2022-11-28 05:35:09 RT @radekosmulski: What is a sure way to improve the performance of your ML solution? Ensembling And what is the easiest way to introdu…

2022-11-27 19:58:33 RT @PromptArtApp: Got some nice results for human faces using SD v2. Had to play around with the prompts and negative prompts. Facial feat…

2022-11-27 19:58:01 RT @EMostaque: So today I am seeing if #StableDiffusion2 can do character art as some have complained about this. I think it can? What do…

2022-11-27 19:55:43 RT @mattturck: VCs switching from crypto to generative AI https://t.co/Z2lM8m50iI

2022-11-26 20:14:51 RT @radekosmulski: Well, this is becoming even more fun Instead of parallelizing the code inside a notebook (which can get tedious), why…

2022-11-26 20:09:45 RT @TheZachMueller: Walk with @fastdotai is still in preorder but as of yesterday the website has gone live! While only the first half of c…

2022-11-26 20:07:20 RT @radekosmulski: The @kaggle OTTO RecSys competition is underway! I share everything that I know So far: • 13x medals for posts •…

2022-11-26 00:47:44 https://t.co/ClyXR3RfXE

2022-11-26 00:46:59 @ximix84 @capetorch @wasimlorgat Me three

2022-11-26 00:42:53 RT @giffmana: Anyone used FocalLoss from RetinaNet successfully for highly imbalanced binary task? Besides detection which it was originall…

2022-11-25 15:28:48 https://t.co/U3t9PMl0ak

2022-11-25 15:27:16 RT @bhutanisanyam1: We will be re-starting Top @kaggle Solutions Study group this weekend Join me on Sunday to learn from: “Open Proble…

2022-11-25 13:30:48 RT @ducha_aiki: I am in dark mood today, as continues terrorist strikes on . However let’s talk about academia. Out of 93700 followe…

2022-11-25 13:30:10 This is amazing - a single awesome dev in his spare time has massively upgraded a giant e-commerce company's search engine, thanks to deep learning and #nbdev https://t.co/pANohhY7mI

2022-11-25 13:26:10 RT @JayAlammar: Big update to "The Illustrated Stable Diffusion" post https://t.co/sbjKHP5ndT 14 new and updated visuals. The biggest u…

2022-11-25 04:06:55 RT @fede_gr: The entirety of @nixtlainc's time series ecosystem is built on top of #nbdev. We code, test, publish and release in notebooks.…

2022-11-25 04:03:19 RT @SteveStuWill: Wow! Scientists have observed a single-cell alga evolve in real time into a multicellular organism. The transition took a…

2022-11-25 03:57:30 RT @johnowhitaker: These models can make such amazing things. But they've also trained on lots of 'meh' stock photos etc. Something like th…

2022-11-25 00:46:11 RT @lukefinleyedit: @ginnyhogan_ https://t.co/o3YJJVR3nU

2022-11-25 00:44:32 RT @mikevanbus: This is a “promoted tweet” on my timeline: https://t.co/PJiO4iZyZv

2022-11-25 00:43:50 RT @iScienceLuvr: "Thanksgiving Dinner" generated by @StableDiffusion 2.0 https://t.co/JkRogEbNv7

2022-11-25 00:42:39 RT @LyskovAndrei: Haven’t run my own experiments yet but seems like commentary about the new @StabilityAI‘s 2.0 model is negative. Thread…

2022-11-24 23:11:20 RT @aronchick: my god GraphQL is unusable. I am so fucking frustrated the @github folks moved over from REST to it without giving us an SDK…

2022-11-24 23:03:37 @radekosmulski @shendusuipian @fastdotai Have you checked out the ability also to inject code into a cell to parameterise notebooks?

2022-11-24 23:00:56 RT @DelJohnsonVC: U.S. Venture capital hasn't been able to outperform European VC for the past 20 years. The Thiel Andreessen generation of…

2022-11-24 07:33:54 RT @_akhaliq: Thinking of retiring from paper tweets it’s pretty time intensive on top of a full time job, I had a good run so far. Plus al…

2022-11-22 05:15:18 RT @ameliaelizalde: found out that ‘arvo’ is australian for ‘afternoon’. when will it end

2022-11-22 05:14:47 RT @RCoops1978: In case you get confused by Aussie responses to questions. .. Note, the simplest way is to just understand the final word…

2022-11-22 05:10:51 @Jousefm2 @keinname42 @jezza_art @jagarikin I don't understand why you're responding like this. You made a fairly minor mistake but your follow-up is not good at all.

2022-11-22 03:59:08 RT @JJitsev: We are in somewhat mild, but very pleasant shock - our LAION-5B paper has just received Outstanding Paper Award at NeurIPS 202…

2022-11-22 03:31:50 @realGeorgeHotz I have pinboard ingest my favorites so I get to search them. Would be nice if this was built in.

2022-11-22 03:29:26 RT @EricTopol: Speech and spread of #SARSCoV2 "Our results suggest that the quantity of speech aerosol, in particular in the ≥ 4-μm range,…

2022-11-21 23:47:49 RT @abouelatta_ali: This is one of the experiments we ran on the Duolingo Retention team : https://t.co/Dv9Wp377vT

2022-11-21 23:46:11 @renegadesilicon @amasad Has there been any twin studies that have looked at outstanding parenting?

2022-11-21 21:01:35 Obstructive sleep apnea (OSA) strongly correlates with snoring. Flu pneumonia hospitalization in highly CPAP-adherent OSA cohort is ~ 5 times less likely, but no correlation with OSA severity. This indicates CPAP benefit by eliminating snoring rather than by minimizing OSA https://t.co/rJaDvJJWZR

2022-11-21 21:01:34 Snoring only carries risk of causing pneumonia when nose/throat is infected. Risk can be strongly reduced by methods that prevent snoring: https://t.co/nqj4dKHXad

2022-11-21 21:01:32 The new study shows that the volume of inhaled snoring droplets is comparable to prior radiotracer studies, implying that these fluids enter the lungs through snoring rather than by “microaspiration”. https://t.co/pXdaP3We4c

2022-11-21 21:01:31 However, only microscopic particles carried by inhaled air, can enter the deep lung and cause viral pneumonia, which is the major cause of covid-19 deaths https://t.co/P3M5Fdussc

2022-11-21 21:01:30 Transfer of fluids from the pharynx into the lungs is common and is known to happen at night during sleep, as documented by radiotracer studies https://t.co/MVX6pTj9cQ

2022-11-21 21:01:28 Microaspiration during sleep is the accepted “hand-waving” mechanism for transfer of microbes from the oral cavity into the lung https://t.co/LdUHJan2MZ

2022-11-21 21:01:27 Infection of the lung with SARS-CoV-2 is a two-step process: first the nose / throat, then the lungs. Postulated, but physically implausible, mechanism for step 2 involves “microaspiration” https://t.co/UjppvrBSeY

2022-11-21 21:01:26 Intriguing new study from the amazing Adriaan Bax and team suggests that most covid deaths resulted from (preventable) snoring droplets rather than (unpreventable) microaspiration. This could be a game changer. No time for the paper? Then read this ! https://t.co/OhOWDSFveI

2022-11-21 07:56:57 @radekosmulski @fastdotai Here's a link to the docs for execnb :D https://t.co/tkLNiTtY3y

2022-11-21 07:55:29 RT @radekosmulski: I just gained a superpower Thank you so much @fastdotai for execnb! I can now: • treat my notebooks as inputs • m…

2022-11-21 02:37:44 RT @jasoncbenn: LLMs are much more capable than anyone initially appreciated. GPT-3, by default, is playing some kind of weird improv game…

2022-11-21 02:30:59 RT @_akpiper: in my #AI and #Literature class we will be spending the next 2 weeks diving into AI co-writing tools. My first impression is…

2022-11-21 02:30:18 RT @EMostaque: Should we put a @paperswithcode Galactica 120bn demo back online for science? Assuming no objections from @MetaAI given lic…

2022-11-21 00:05:05 RT @ajankelo: Props to @HamelHusain for creating the most interesting discord for #MLOps and tooling. Excellent: discussion people…

2022-11-20 18:47:27 RT @untitled01ipynb: Can you fall in love with a monkey patch? https://t.co/sxVxRdEsmw https://t.co/NibMhD1Wjg

2022-11-20 00:19:59 RT @BriannaWu: This is amazing. Elno broke the copyright strike system, so you can just post entire films on Twitter now.

2022-11-19 18:22:55 RT @rahaeli: Hey, US folks newly running Mastodon instances: do Future You a *huge* favor, mitigate your potential liability, and register…

2022-11-19 03:25:23 RT @DataChaz: Still not sure about Mastodon? Come join @hamel's Discord instead! There's an ungodly amount of cool data peeps in there…

2022-11-19 03:17:44 RT @BlancheMinerva: I am very excited to announce that a joint proposal w/ @AiEleuther @Mila_Quebec and @laion_ai to train Foundation Model…

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-11-29 06:49:02 @RickNTA That's a very good question! I haven't actually tried it TBH... :D

2022-11-29 06:47:54 @simonw @unsynchronized Awesome just sent a preview request :) I want to create a Factorio recipe DB to use with my daughter. Very important application!

2022-11-29 06:08:29 @Just_Janes @nocodb That's not an online service AFAIK.

2022-11-29 06:07:30 @unsynchronized Datasette is great (thanks @simonw!) but there isn't really a collaborative online service for it AFAIK, and forms and reports would require a bunch extra.

2022-11-29 06:05:55 @unsynchronized Thanks for the ideas. The Google one doesn't do joins.

2022-11-29 05:19:57 What online service is as similar as possible to MS Access? I.e supports sql, joins, basic forms and reports? (airtable doesn't do sql/joins IIUC)

2022-11-29 01:16:17 @pcuenq What's a Spaces lounge?

2022-11-29 01:15:29 RT @enzokro_: Put together a quick post for #stablediffusion v2, checking whether dynamic classifier-free guidance helps like it did with v…

2022-11-28 07:21:32 RT @Mylovanov: Good evening. Day 5 after the latest Russian attack on Kyiv. Day 277 of the war. I am president of the Kyiv School of Econom…

2022-11-28 05:39:23 RT @rdmorris: Overwhelming evidence of outdoor transmission of Omicron. At least 32 unmasked people were infected by a man jogging through…

2022-11-28 05:35:26 RT @Suhail: Headed on a week long trip to completely focus on and study AI tomorrow. Would love any links, code, papers to read related to…

2022-11-28 05:35:09 RT @radekosmulski: What is a sure way to improve the performance of your ML solution? Ensembling And what is the easiest way to introdu…

2022-11-27 19:58:33 RT @PromptArtApp: Got some nice results for human faces using SD v2. Had to play around with the prompts and negative prompts. Facial feat…

2022-11-27 19:58:01 RT @EMostaque: So today I am seeing if #StableDiffusion2 can do character art as some have complained about this. I think it can? What do…

2022-11-27 19:55:43 RT @mattturck: VCs switching from crypto to generative AI https://t.co/Z2lM8m50iI

2022-11-26 20:14:51 RT @radekosmulski: Well, this is becoming even more fun Instead of parallelizing the code inside a notebook (which can get tedious), why…

2022-11-26 20:09:45 RT @TheZachMueller: Walk with @fastdotai is still in preorder but as of yesterday the website has gone live! While only the first half of c…

2022-11-26 20:07:20 RT @radekosmulski: The @kaggle OTTO RecSys competition is underway! I share everything that I know So far: • 13x medals for posts •…

2022-11-26 00:47:44 https://t.co/ClyXR3RfXE

2022-11-26 00:46:59 @ximix84 @capetorch @wasimlorgat Me three

2022-11-26 00:42:53 RT @giffmana: Anyone used FocalLoss from RetinaNet successfully for highly imbalanced binary task? Besides detection which it was originall…

2022-11-25 15:28:48 https://t.co/U3t9PMl0ak

2022-11-25 15:27:16 RT @bhutanisanyam1: We will be re-starting Top @kaggle Solutions Study group this weekend Join me on Sunday to learn from: “Open Proble…

2022-11-25 13:30:48 RT @ducha_aiki: I am in dark mood today, as continues terrorist strikes on . However let’s talk about academia. Out of 93700 followe…

2022-11-25 13:30:10 This is amazing - a single awesome dev in his spare time has massively upgraded a giant e-commerce company's search engine, thanks to deep learning and #nbdev https://t.co/pANohhY7mI

2022-11-25 13:26:10 RT @JayAlammar: Big update to "The Illustrated Stable Diffusion" post https://t.co/sbjKHP5ndT 14 new and updated visuals. The biggest u…

2022-11-25 04:06:55 RT @fede_gr: The entirety of @nixtlainc's time series ecosystem is built on top of #nbdev. We code, test, publish and release in notebooks.…

2022-11-25 04:03:19 RT @SteveStuWill: Wow! Scientists have observed a single-cell alga evolve in real time into a multicellular organism. The transition took a…

2022-11-25 03:57:30 RT @johnowhitaker: These models can make such amazing things. But they've also trained on lots of 'meh' stock photos etc. Something like th…

2022-11-25 00:46:11 RT @lukefinleyedit: @ginnyhogan_ https://t.co/o3YJJVR3nU

2022-11-25 00:44:32 RT @mikevanbus: This is a “promoted tweet” on my timeline: https://t.co/PJiO4iZyZv

2022-11-25 00:43:50 RT @iScienceLuvr: "Thanksgiving Dinner" generated by @StableDiffusion 2.0 https://t.co/JkRogEbNv7

2022-11-25 00:42:39 RT @LyskovAndrei: Haven’t run my own experiments yet but seems like commentary about the new @StabilityAI‘s 2.0 model is negative. Thread…

2022-11-24 23:11:20 RT @aronchick: my god GraphQL is unusable. I am so fucking frustrated the @github folks moved over from REST to it without giving us an SDK…

2022-11-24 23:03:37 @radekosmulski @shendusuipian @fastdotai Have you checked out the ability also to inject code into a cell to parameterise notebooks?

2022-11-24 23:00:56 RT @DelJohnsonVC: U.S. Venture capital hasn't been able to outperform European VC for the past 20 years. The Thiel Andreessen generation of…

2022-11-24 07:33:54 RT @_akhaliq: Thinking of retiring from paper tweets it’s pretty time intensive on top of a full time job, I had a good run so far. Plus al…

2022-11-22 05:15:18 RT @ameliaelizalde: found out that ‘arvo’ is australian for ‘afternoon’. when will it end

2022-11-22 05:14:47 RT @RCoops1978: In case you get confused by Aussie responses to questions. .. Note, the simplest way is to just understand the final word…

2022-11-22 05:10:51 @Jousefm2 @keinname42 @jezza_art @jagarikin I don't understand why you're responding like this. You made a fairly minor mistake but your follow-up is not good at all.

2022-11-22 03:59:08 RT @JJitsev: We are in somewhat mild, but very pleasant shock - our LAION-5B paper has just received Outstanding Paper Award at NeurIPS 202…

2022-11-22 03:31:50 @realGeorgeHotz I have pinboard ingest my favorites so I get to search them. Would be nice if this was built in.

2022-11-22 03:29:26 RT @EricTopol: Speech and spread of #SARSCoV2 "Our results suggest that the quantity of speech aerosol, in particular in the ≥ 4-μm range,…

2022-11-21 23:47:49 RT @abouelatta_ali: This is one of the experiments we ran on the Duolingo Retention team : https://t.co/Dv9Wp377vT

2022-11-21 23:46:11 @renegadesilicon @amasad Has there been any twin studies that have looked at outstanding parenting?

2022-11-21 21:01:35 Obstructive sleep apnea (OSA) strongly correlates with snoring. Flu pneumonia hospitalization in highly CPAP-adherent OSA cohort is ~ 5 times less likely, but no correlation with OSA severity. This indicates CPAP benefit by eliminating snoring rather than by minimizing OSA https://t.co/rJaDvJJWZR

2022-11-21 21:01:34 Snoring only carries risk of causing pneumonia when nose/throat is infected. Risk can be strongly reduced by methods that prevent snoring: https://t.co/nqj4dKHXad

2022-11-21 21:01:32 The new study shows that the volume of inhaled snoring droplets is comparable to prior radiotracer studies, implying that these fluids enter the lungs through snoring rather than by “microaspiration”. https://t.co/pXdaP3We4c

2022-11-21 21:01:31 However, only microscopic particles carried by inhaled air, can enter the deep lung and cause viral pneumonia, which is the major cause of covid-19 deaths https://t.co/P3M5Fdussc

2022-11-21 21:01:30 Transfer of fluids from the pharynx into the lungs is common and is known to happen at night during sleep, as documented by radiotracer studies https://t.co/MVX6pTj9cQ

2022-11-21 21:01:28 Microaspiration during sleep is the accepted “hand-waving” mechanism for transfer of microbes from the oral cavity into the lung https://t.co/LdUHJan2MZ

2022-11-21 21:01:27 Infection of the lung with SARS-CoV-2 is a two-step process: first the nose / throat, then the lungs. Postulated, but physically implausible, mechanism for step 2 involves “microaspiration” https://t.co/UjppvrBSeY

2022-11-21 21:01:26 Intriguing new study from the amazing Adriaan Bax and team suggests that most covid deaths resulted from (preventable) snoring droplets rather than (unpreventable) microaspiration. This could be a game changer. No time for the paper? Then read this ! https://t.co/OhOWDSFveI

2022-11-21 07:56:57 @radekosmulski @fastdotai Here's a link to the docs for execnb :D https://t.co/tkLNiTtY3y

2022-11-21 07:55:29 RT @radekosmulski: I just gained a superpower Thank you so much @fastdotai for execnb! I can now: • treat my notebooks as inputs • m…

2022-11-21 02:37:44 RT @jasoncbenn: LLMs are much more capable than anyone initially appreciated. GPT-3, by default, is playing some kind of weird improv game…

2022-11-21 02:30:59 RT @_akpiper: in my #AI and #Literature class we will be spending the next 2 weeks diving into AI co-writing tools. My first impression is…

2022-11-21 02:30:18 RT @EMostaque: Should we put a @paperswithcode Galactica 120bn demo back online for science? Assuming no objections from @MetaAI given lic…

2022-11-21 00:05:05 RT @ajankelo: Props to @HamelHusain for creating the most interesting discord for #MLOps and tooling. Excellent: discussion people…

2022-11-20 18:47:27 RT @untitled01ipynb: Can you fall in love with a monkey patch? https://t.co/sxVxRdEsmw https://t.co/NibMhD1Wjg

2022-11-20 00:19:59 RT @BriannaWu: This is amazing. Elno broke the copyright strike system, so you can just post entire films on Twitter now.

2022-11-19 18:22:55 RT @rahaeli: Hey, US folks newly running Mastodon instances: do Future You a *huge* favor, mitigate your potential liability, and register…

2022-11-19 03:25:23 RT @DataChaz: Still not sure about Mastodon? Come join @hamel's Discord instead! There's an ungodly amount of cool data peeps in there…

2022-11-19 03:17:44 RT @BlancheMinerva: I am very excited to announce that a joint proposal w/ @AiEleuther @Mila_Quebec and @laion_ai to train Foundation Model…

2022-11-18 20:00:46 RT @qlhoest: Using Hugging Face Datasets 2.7 you can now leverage multiprocessing to load any of the 10k datasets even faster ! Load…

2022-11-18 19:20:22 RT @simonw: Since it's feeling particularly relevant for some reason right now, here's my Mastodon explainer targeted at internet users of…

2022-11-18 14:38:05 RT @yvrjsharma: Yesterday @Suhail shared his dataset from @playgroundAI of liked images (links) and other metadata. I have built a fun lit…

2022-11-18 14:31:14 RT @growing_daniel: Most difficult five days of my posting career but we got him https://t.co/EdmMr7Z1UP

2022-11-18 14:14:12 RT @HeidyKhlaaf: I know *not all* EA's explicitly believe in eugenics, but if these are the communities you're harbouring in your space (…

2022-11-18 13:48:52 RT @peterclowes: Why I left @twitter or rather why I did not sign up for “extremely hardcore” Twitter 2.0

2022-11-18 10:16:54 RT @vinnguini: @Vinncent One of the main philosophers of "Effective Altruism" proposed a eugenics based on economic productivity. He was al…

2022-11-18 10:16:45 RT @Vinncent: I admit I just learned about "effective altruism" this month but, at the risk of being too on-brand, I have to say that "I ne…

2022-11-18 10:15:34 RT @parismarx: Effective altruist Sam Bankman-Fried talks to effective altruist Kelsey Piper for Vox’s effective altruist vertical Future P…

2022-11-18 09:57:26 RT @SeattleiteLeo: 5/ With masks, we can directly test and measure their efficacy via controlled particulate generators and sensors. This i…

2022-11-18 09:35:47 RT @Buntworthy: Comparing my SD upscaler and rivers': - original - @RiversHaveWings latent upscaler - My SD upscaler https://t.co/4VOU37f…

2022-11-18 08:45:30 RT @jamescham: A terrific meditation on simulations as a--among other things--game, tool for reasoning about the world, and way to deal wit…

2022-11-18 07:03:51 @BlackHC @TheZachMueller There's an article here that covers some of the issues https://t.co/nX49O9OmPE

2022-11-18 07:02:24 @BlackHC @TheZachMueller Yeah I'm hoping there's some problem with the methodology. But I've heard from immunologists that have expertise in this area and they tell me they've been considering this a likely outcome.

2022-11-18 06:55:58 RT @radekosmulski: If you'd like to take the new Merlin Dataloader library for a spin I created a @kaggle kernel that trains a matrix fa…

2022-11-18 04:25:39 @TheZachMueller @BlackHC Yeah I read the study pretty carefully and thought it seemed reasonable, although they didn't share code or data or model details so hard to verify

2022-11-17 21:25:11 RT @LyskovAndrei: @mickeyxfriedman @fastdotai w/ @jeremyphoward was better than any course I took in college. I can maybe remember a hand…

2022-11-17 21:23:30 RT @GergelyOrosz: Scoop: I am hearing far fewer than expected devs hit "yes". Elon sent out an email relaxing remote working from the form…

2022-11-17 20:08:21 RT @TheZachMueller: Looks like nbdev-extensions hit #1 trending on @github. Cool!!! https://t.co/v7Y74sG1vq https://t.co/f3ovNopkla

2022-11-17 20:02:43 RT @arxiv: We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly…

2022-11-17 19:53:10 @johnowhitaker Oh no!

2022-11-17 19:42:02 RT @CDCgov_parody: Immunity debt is NOT a thing.

2022-11-17 19:40:57 RT @ai_fast_track: My mentee @gunaratne_rahel and I are super pleased: We won first place at the @lablabai Stable Diffusion AI Hackathon…

2022-11-17 08:06:36 RT @benjamin_warner: PyTorch's fused foreach methods, used in their multi-tensor optimizers, are surprisingly undocumented given how useful…

2022-11-17 06:35:48 RT @radekosmulski: Merlin Dataloader is 119x faster than my own PyTorch Dataset + Dataloader combo! This is revolutionary for tabular data…

2022-11-17 02:35:52 RT @wightmanr: For ages timm had 2 models that were at or past 87% on ImageNet. Today there's ~18 of them. Prepping a big update while fini…

2022-11-17 01:50:02 RT @Angelina_Magr: I will be hiring Data scientists and ML engineers, so please reach out if you are interested. Fully remote in the US. Bo…

2022-11-17 01:40:54 This VQGAN tutorial is really impressive - a hidden gem! :D https://t.co/Vj0sYvAB6F

2022-11-17 01:40:18 RT @hardmaru: Our children will grow up taking for granted that computers can “simply” produce realistic images, entire stories, or chat wi…

2022-11-17 01:36:41 RT @Suhail: An open dataset of all the liked images and their prompts on playgroundai\.com: https://t.co/24YIpCoGFi Have fun. Let us know…

2022-11-17 01:34:28 RT @isaac_flath: What's your favorite course or resource for data scientists to learn crucial skills that are not data-science specific? Or…

2022-11-17 00:11:55 RT @EMostaque: Nothing major, just take brain recordings and reconstruct what folk are seeing at the time from them. Exponentials eh.

2022-11-17 00:10:28 RT @Tara_MacAulay: Quick facts: I started Alameda Research with Sam in 2017. In April 2018, I and a group of others all quit, in part due t…

2022-11-17 00:08:38 RT @savvyRL: Thoughts? https://t.co/qGXLMQye5n

2022-11-16 23:25:16 Here's a very cool project - and written using #nbdev, which really helps ensure a high quality piece of software. https://t.co/xqcfAgv52N

2022-11-16 19:02:14 RT @iScienceLuvr: I will attempt to explain the basic idea of how diffusion models work! ... in only 15 tweets! Let's get started ↓

2022-11-16 06:13:28 @karpathy @johnowhitaker Yes - this feels like what I've been looking for for a while now!

2022-11-16 06:13:05 RT @johnowhitaker: This looks great! Simple idea, nice clean code implementation... Takes a key idea from diffusion (iterative improvement…

2022-11-16 02:28:25 @rosstaylor90 @misterkardas Love it :)

2022-11-16 02:25:19 RT @jachiam0: The fundamental problem with EA is that it wants to be a non-geographic nation state, with its progressive taxes, secular hum…

2022-11-15 22:36:09 @Mascobot @HamelHusain Oops!

2022-11-15 22:36:04 https://t.co/8v7K3syVzS

2022-11-15 21:39:00 Looks like https://t.co/pCXO2erTLe had a good plan for a fastai notebook, although I can't get it to go further than producing this intro! :D https://t.co/VmxreWedBc

2022-11-15 21:11:47 RT @GergelyOrosz: Musk is mocking the sw engineers he’s fired at Twitter on the way out. Tech is a pretty small community. These aren’t fac…

2022-11-15 21:05:37 RT @BillHiggins: I’ve been unpleasantly surprised that some people seem to condone the public criticism of employees and/or the public fi…

2022-11-15 21:02:15 RT @CaseyNewton: NEW: Employees who have criticized Elon Musk in Twitter’s Slack channels were fired overnight over email. “We regret to…

2022-11-15 20:58:18 I'm so sick of mask-shaming. I gave a talk in a big lecture theater full of hundreds of people, and still got mask-shamed. Somehow this person decided I'm actually on a video call on my own?!? If you see someone wearing a mask, they have a reason for it. Don't mask-shame. https://t.co/4gjFzcQszE

2022-12-08 19:53:43 RT @DGBassani: I suspect it is no coincidence that the rate of COVID-19 admissions among the cohort of children born 6-12 months ago aligns…

2022-12-08 19:52:00 RT @JeromeAdamsMD: An alternative to the “no germs” theory of respiratory infections: Over 80% of the population has now been exposed to…

2022-12-08 19:51:05 @eugeneyan Fastcore.script

2022-12-08 19:47:20 RT @marksaroufim: Forget about dragons for a second, yes compilers are complex because perf outweighs complexity but picture a happy panda…

2022-12-08 06:29:47 Another great diffusion model paper explainer from our wonderful @fastdotai students! https://t.co/TtLHPopVXY

2022-12-08 05:24:56 RT @Suhail: It kinda feels like Apple is asleep at the AI wheel and I think that could be the moment we look back on and think: why didn't…

2022-12-08 05:24:42 RT @LyskovAndrei: Looks like @OpenAI is starting to nerf ChatGPT, I'm glad we launched @ai_sages within a few days of launch, otherwise, we…

2022-12-08 05:11:02 RT @HamelHusain: We partnered w/ W&

2022-12-08 05:10:49 RT @renegadesilicon: I can’t help but think that ChatGPT is going to be the most disruptive technology in quite some time. For the last 2…

2022-12-08 19:53:43 RT @DGBassani: I suspect it is no coincidence that the rate of COVID-19 admissions among the cohort of children born 6-12 months ago aligns…

2022-12-08 19:52:00 RT @JeromeAdamsMD: An alternative to the “no germs” theory of respiratory infections: Over 80% of the population has now been exposed to…

2022-12-08 19:51:05 @eugeneyan Fastcore.script

2022-12-08 19:47:20 RT @marksaroufim: Forget about dragons for a second, yes compilers are complex because perf outweighs complexity but picture a happy panda…

2022-12-08 06:29:47 Another great diffusion model paper explainer from our wonderful @fastdotai students! https://t.co/TtLHPopVXY

2022-12-08 05:24:56 RT @Suhail: It kinda feels like Apple is asleep at the AI wheel and I think that could be the moment we look back on and think: why didn't…

2022-12-08 05:24:42 RT @LyskovAndrei: Looks like @OpenAI is starting to nerf ChatGPT, I'm glad we launched @ai_sages within a few days of launch, otherwise, we…

2022-12-08 05:11:02 RT @HamelHusain: We partnered w/ W&

2022-12-08 05:10:49 RT @renegadesilicon: I can’t help but think that ChatGPT is going to be the most disruptive technology in quite some time. For the last 2…

2022-12-09 03:58:51 RT @iScienceLuvr: I think there is some confusion in the #AiArt community about Stable Diffusion weights. Some folks think non-EMA is for i…

2022-12-09 03:53:50 @andfanilo Great stuff!

2022-12-08 19:53:43 RT @DGBassani: I suspect it is no coincidence that the rate of COVID-19 admissions among the cohort of children born 6-12 months ago aligns…

2022-12-08 19:52:00 RT @JeromeAdamsMD: An alternative to the “no germs” theory of respiratory infections: Over 80% of the population has now been exposed to…

2022-12-08 19:51:05 @eugeneyan Fastcore.script

2022-12-08 19:47:20 RT @marksaroufim: Forget about dragons for a second, yes compilers are complex because perf outweighs complexity but picture a happy panda…

an ideal talk to me :)