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

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

2023-05-20 19:40:08 @rob_soko How it was. Things are better now, of course, but there is still plenty of sexism in academia, some of which we probably don't even recognize (just like back then, the assumption was that women weren't going to be "leaders").

2023-05-20 19:20:01 Another choice paragraph: https://t.co/nv1yRdjbYm

2023-05-20 19:18:54 This article (on the first women at Yale, 50 years ago) actually made me gasp in several parts. https://t.co/d5FI3RLb1l https://t.co/Mmr2a0aykX

2023-05-19 20:13:57 @YJernite @BigscienceW @BigCodeProject This is great -- thanks for the pointer!

2023-05-19 20:00:31 This would be extremely useful for understanding LLMs. It might be tricky to define "most similar documents". Has anyone done this with open-source / open-data LLMs? https://t.co/SQ8pUE9vLp

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

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

2023-04-23 22:50:05 @JosephLMcCord https://t.co/nnpL0lqOgf

2023-04-23 17:47:02 All NSF program managers (and other grant-giving organizations) should read this thread. It calls for deeper thinking about what diversity in academia entails. https://t.co/JyxBE3U4QG

2023-04-23 16:39:15 @ReflectiveAlt @DrTechlash @JoshuaBlake_ https://t.co/GTuPZYxIBc

2023-04-23 16:32:22 @MubasharSharif @ricard_sole @sfiscience @manlius84 @_AlexArenas @drmichaellevin Working on it!

2023-04-23 16:15:37 @dwallacewells @ezraklein

2023-04-23 15:38:47 Do half of AI researchers believe that there's a 10% chance AI will kill us all? I fact-check a widespread claim. (Thanks also to @DrTechlash) https://t.co/GTuPZYxIBc

2023-04-22 14:03:51 @WeisiG @ricard_sole https://t.co/G9xQyQvk2x

2023-04-22 14:03:42 @ricard_sole @sfiscience @manlius84 @_AlexArenas @drmichaellevin Ricard, thanks for the shout-out! For anyone interested, here's a non-paywalled version: https://t.co/G9xQyQvk2x

2023-04-21 18:32:36 @jeffclune @TonyZador And there will always be edge cases and uncertainty!

2023-04-21 18:32:28 @jeffclune @TonyZador Great question &

2023-04-21 18:23:28 @fchollet Next up is constant repetition of "why?" Same reason. Enjoy!

2023-04-21 04:50:36 @DrTechlash Seems like "50% of the 3% who answered this question" got translated into "50% of AI researchers". @Laurie_Garrett

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

2023-04-19 20:35:54 Thank you to the journalists who covered how @60minutes spread misinformation about AI: https://t.co/k61Ta49LnG https://t.co/rvyzv22Y2G https://t.co/uhGhpki5ll

2023-04-19 14:05:39 Looking forward to this salon on Friday! Join us. https://t.co/7L3bRe8IZy

2023-04-19 13:00:28 @conte_hal @zsk @tomdsingleton @60Minutes See also: https://t.co/rvyzv23vSe

2023-04-18 22:06:05 @dileeplearning @erikbryn @ceobillionaire Having fun with this. https://t.co/l0LsDVs59R

2023-04-18 02:20:51 See https://t.co/nX5Id4plb5 for fact-checking on the 60 Minutes claims.

2023-04-18 01:58:54 @enthusiasticdev @60Minutes Another thread on this topic: https://t.co/O7DiQNtEJY

2023-04-18 01:56:41 @enthusiasticdev @60Minutes Here's a good thread with more specifics: https://t.co/nX5Id4plb5

2023-04-17 21:33:50 @pbreit https://t.co/nX5Id4oNlx

2023-04-17 21:20:32 I sure hope some journalist does a review of the whole @60minutes segment on Google Bard as a case study in how *not* to cover AI.

2023-04-17 19:54:27 @luisortiz345 See excellent thread from@mmitchell_ai explaining just what is wrong with the headline. https://t.co/nX5Id4oNlx

2023-04-17 18:46:14 Some important fact-checking on that misleading @60Minutes headline: https://t.co/nX5Id4oNlx

2023-04-17 15:19:22 @IntuitMachine @philipcball We did not define "understanding" but discussed different possible ways to think about it.

2023-04-17 02:13:43 The whole clip. SMH.

2023-04-17 02:09:38 Not going there. https://t.co/b3jTPZG41Y

2023-04-17 02:09:06 This headline. https://t.co/4O9HVX2VGR

2023-04-16 17:50:24 @mer__edith I believe Neocognitron was first proposed in 1979.

2023-04-16 14:44:11 @pfau In the past, lots of people showed that earlier image-to-text systems were highly susceptible to AEs, but it didn't matter much since the image-to-text systems didn't work well in the first place. That might change with GPT-4, where AEs might turn into a real-world concern.

2023-04-16 14:43:46 @pfau I think we'll see renewed attention to adversarial examples in vision once OpenAI releases GPT-4 image-to-text component.

2023-04-15 16:39:45 @glupyan @mcxfrank We are in the midst of testing this kind of transfer on people using problems in the Abstraction and Reasoning Corpus domain, which I think is a better (more straightfoward) test of abstraction transfer than Raven's.

2023-04-15 16:38:26 @glupyan @mcxfrank No we didn't test on people (yet). Of course our experiments were on the procedurally generated "RAVEN" (not Raven's) problems, which tend to be much easier than the originals, so I suspect people would have an easier time.

2023-04-15 16:36:44 @glupyan @raphaelmilliere @ImageDeeply @QuantaMagazine @RTomMcCoy Yes, I agree.

2023-04-15 16:35:18 RT @AlisonGopnik: @mcxfrank We've done this! Use linguistic descriptions of blicket detectors to test LLM's and compare to children (recent…

2023-04-15 03:12:07 @anilananth Very good article. But I don't think the analogy problem is solved yet

2023-04-14 23:51:40 tl

2023-04-14 23:47:40 Thoughtful article about ChatGPT and its limitations. Some good analogies to explain how LLMs work. Worth reading (though might require subscription to New Yorker): https://t.co/4ltUubtCdl

2023-04-14 23:13:59 @mcxfrank This is exactly the kind of thing we were trying to get at in this preliminary paper (in the context of DNNs on Raven's Progressive Matrices, not LLMs): https://t.co/WhhlUlhp06 A much more in-depth paper is in progress!

2023-04-14 23:08:52 This is an truly insightful and useful thread. https://t.co/TA4eE5hVFR

2023-04-14 21:49:14 @RTomMcCoy @raphaelmilliere @ImageDeeply @QuantaMagazine Thx.

2023-04-14 21:46:13 @RTomMcCoy @raphaelmilliere @ImageDeeply @QuantaMagazine I saw your paper on this wrt RNNs --what's the reference for Transformers ?

2023-04-14 20:04:51 @ImageDeeply @QuantaMagazine My understanding of the difference is that the dimensionality of the vectors in hyperdimensional computing is so much higher than in, say, LLM embeddings. And the higher dimensionality gives rise to unexpected new properties. To quote Phil Anderson, "More is different".

2023-04-14 15:28:10 @geomblog @QuantaMagazine "Hypervector" seems to be shorthand for "hyperdimensional vector", in other words, *big*. From https://t.co/iDnvAdqTLR: https://t.co/Xi3M0xFxeD

2023-04-14 15:08:13 Including this wonderful photo: https://t.co/fFDxzGwwQY

2023-04-14 15:08:12 Article in @QuantaMagazine on two of the most brilliant thinkers I know: Bruno Olshausen and Pentti Kanerva: https://t.co/hQ1fpWhyEm

2023-04-14 03:03:31 @caddelljd @austinkleon @biblioracle Thank you, twice!

2023-04-13 23:23:30 @ProfLHunter Gotta start somewhere.

2023-04-13 23:16:37 @austinkleon @biblioracle I hope you continue to enjoy it!

2023-04-13 22:04:51 @yoavgo @Grady_Booch Emacs was the original AGI. It even had alignment -- the DWIM ("do what I mean") command.

2023-04-13 19:22:34 It's time to rethink how we evaluate AI systems. I'm proud to be a co-author of this new paper, which has some recommendations https://t.co/5llCrMurzJ

2023-04-13 16:22:20 @anecdotal mom here https://t.co/pIBVa1KGfK https://t.co/sJekVnNl7b https://t.co/wmga832Si9

2023-04-11 19:51:49 @davidWonderland I believe that weed jokes are Elon's main motivator in all decisions.

2023-04-11 19:31:54 For anyone who doesn't know why he chose 4/20 https://t.co/yuTKBAc6Wf

2023-04-11 19:31:04 In other words, "Final date for WEEDing out legacy Blue checks is 4/20". https://t.co/O9JTPIeInG

2023-04-11 15:37:17 RT @alondra: Had an engaging conversation with @EzraKlein about the #AIBillofRights and our unprecedented window of opportunity for broader…

2023-04-11 12:28:49 RT @DoctorJosh: Embodied AI, and @uvmvermont, on the front page of the @nytimes today. https://t.co/LbwZtNebD5

2023-04-09 18:49:00 In a dialogue in Gödel, Escher, Bach (1979), the Crab describes "a new...very flexible kind of computer": "they are called 'smart-stupids', since they are so flexible, and have the potential to be either smart or stupid, depending on how skillfully they are instructed." (1/2) https://t.co/7fk6kg5lCM

2023-04-09 16:10:50 'It is at once both smarter and dumber than any person you've ever met' Describes most big AI successes! https://t.co/nS2jHs2kJS

2023-04-08 21:08:24 Head of lettuce won again. Twitter block on Substack seems to be gone. https://t.co/wcdmAI4XYq

2023-04-08 13:41:56 RT @emilymbender: .@ChrisMurphyCT I'd like to set the record straight. I can understand how the reaction of the tech world to your tweet wa…

2023-04-08 02:31:35 @garg_arun @HernanAguirreD It's a screenshot, not original tweet.

2023-04-08 02:27:30 And I can't even reply to this tweet, since it contains link to substack. Maybe time to give up on Twitter? @HernanAguirreD https://t.co/lzoQDKasks

2023-04-07 21:51:47 For a microsecond, I was able to like this tweet. https://t.co/B4R0qoOTH9 https://t.co/xEFlKGenAL

2023-04-07 21:27:10 Twitter block of Substack versus head of lettuce, anyone?

2023-04-07 20:27:22 @sd_marlow @elonmusk https://t.co/MkhypskaSQ

2023-04-07 19:34:52 @ondergurcan @elonmusk https://t.co/Eohrpu5xWv

2023-04-07 19:26:30 Wow, Twitter is blocking likes, RTs, and replies to this tweet. WTF @elonmusk?? https://t.co/21gCjbr6XB

2023-04-07 15:05:21 RT @hamishmckenzie: Today we're announcing Substack Notes. Notes lets writers publish short-form posts and recommend almost anything: quo…

2023-04-07 13:41:19 Here's a link to my Substack, where I write about AI. Twitter is blocking you from liking or retweeting tweets that contain Substack links. Why??? But *for now* (how much longer?) you can still click on it. https://t.co/LHpY5WHgeR

2023-04-06 22:41:17 @michael_nielsen https://t.co/75WQte3pmZ

2023-04-06 19:35:44 @KMLewis903 Thanks for the kind words!

2023-04-06 14:47:11 Looking forward to reading this!! https://t.co/uJBab9f61I

2023-04-06 02:55:08 RT @UODataServices: Are you ready to explore the fascinating world of AI? Join our book club as we dive into 'Artificial Intelligence: A Gu…

2023-04-05 02:19:30 @mcxfrank Thanks for this excellent thread. Rich Shiffrin and I recently wrote a short commentary on this topic as well: https://t.co/sVeYdHheBb Doing psychological testing on AI systems is a tricky business!

2023-04-04 22:05:39 Excellent thread (and blog post) on why AI "doomerism" is ill-conceived. https://t.co/iyHChRoMHt

2023-04-04 20:20:07 "Hallucitation" -- brilliant! https://t.co/a93TSncSZ3

2023-04-04 04:03:11 RT @mmitchell_ai: New great piece from @Melissahei where she explains 3 serious security issues with modern chatbots -- and there aren't wo…

2023-04-03 20:50:49 My "thoughts on a crazy week in AI news". https://t.co/wmga833q7H

2023-04-03 16:03:10 @dileeplearning @pfau But if it does replicate --not clear what to conclude.

2023-04-03 15:40:50 @dileeplearning @pfau Plus, ChatGPT is likely updated constantly, using data from user conversations. https://t.co/YtOgRJQrmV

2023-04-03 15:39:23 @dileeplearning @pfau I agree with David's meta-opinion. But I don't think that replication by the likes of us outsiders is possible. The version of GPT-4 we get for $20 is almost certainly not the same one they were experimenting with.

2023-04-03 15:15:43 Law-enforcement use of facial recognition software should be banned. Period. https://t.co/MBy7yopeS9

2023-04-01 02:00:02 @Grady_Booch Is "Open AI" taken?

2023-03-31 17:03:40 @douglas_eck There's an app for that. https://t.co/H9qE1SFldi

2023-03-31 15:09:40 @pcastr @criticalneuro @Soccermatics Wow. I was trying to explain this dichotomy to someone just yesterday.

2023-03-30 22:05:51 I am so tired. https://t.co/jbHiMnV8iN

2023-03-30 20:59:25 @rasbt Thanks so much, Sebastian!

2023-03-30 17:42:46 @mmitchell_ai Oh no, I hope you are better soon.

2023-03-30 02:06:22 Dileep's thoughts on intelligence--natural and artificial--are always interesting and worth reading: https://t.co/PKoAsAzw2Q

2023-03-29 20:48:42 New job opening at the Santa Fe Institute (@sfiscience): Digital Media Specialist. https://t.co/XlLbQQ1UdX

2023-03-29 20:16:32 RT @TaylorWWebb: Major update to our paper on emergent analogy in LLMs, with a number of additional tests and behavioral experiments, and a…

2023-03-29 16:58:36 https://t.co/LbDY0hoA6M

2023-03-29 16:30:54 I didn't sign "the letter". Current AI poses lots of risks, but describing these systems as "ever more powerful digital minds" that no one can control is likely to make the problem even worse. What's needed: more transparency and better public discourse.

2023-03-29 15:39:49 @LarryChicago @ChrisMurphyCT Sorry, I'd have to pay $8 a month to Elon Musk to edit my typos. Instead, I'll live with the "typo shaming" . Cheers!

2023-03-28 22:55:23 @boneillhawk @ChrisMurphyCT

2023-03-28 22:00:18 @ChrisMurphyCT Please take my and others' comments in that light.

2023-03-28 22:00:10 @ChrisMurphyCT This is what I and my fellow AI researchers were worrying about in regard to your earlier tweet. We are all very anxious to educate the public about the actual abilities, limitations, and dangers of today's AI systems. Many of us are strongly in favor of regulation. (2/2)

2023-03-28 21:48:58 @ChrisMurphyCT Dear Senator, I am a big fan of yours and your work. Absolutely no shaming was intended. Given the heated rhetoric around AI, shorthand for the abilities of AI can (and will) be dangerously misinterpreted---I've written extensively about this issue. (1/2)

2023-03-28 18:43:49 @santiontanon Here's the arXiv version: https://t.co/G9xQyQuMcZ

2023-03-28 18:43:02 @LakeBrenden Both are bad, but they are an interesting microcosm of the opposing extremes of wrongness on this issue.

2023-03-28 15:30:54 @mmitchell_ai First year of undergrad. I signed up for intro CS class. Had trouble with an assignment. Went to professor for help. He said, "Maybe this subject isn't really for you." I dropped the class.

2023-03-27 23:48:02 Brilliant essay by my colleague, David Krakauer, on the question "Does GPT-4 Really Understand What We're Saying?" https://t.co/UZHlAvaVR1

2023-03-27 17:36:50 @JohnHaugeland Lots of good detailed replies in the thread. E.g., https://t.co/gtkPiIub0s

2023-03-27 13:20:03 @ChrisMurphyCT Senator, I'm an AI researcher. Your description of ChatGPT is dangerously misinformed. Every sentence is incorrect. I hope you will learn more about how this system actually works, how it was trained, and what it's limitations are.

2023-03-27 13:07:01 There is so much wrong in this tweet. Our political leaders need better AI literacy. https://t.co/gx5VM5tHrW

2023-03-26 00:06:56 @stanislavfort @deliprao @natfriedman Prompting via name-dropping

2023-03-25 20:40:23 @stanislavfort @deliprao @natfriedman I personally can't solve any of these.

2023-03-25 20:38:35 @stanislavfort @deliprao @natfriedman Lol. Gear wars -- humans vs. machines: https://t.co/bkAKX3xQ3B

2023-03-25 20:35:33 @deliprao @natfriedman Thanks for the interesting reply! I'm a big fan of analogy But a lot to unpack here -- too much for a sunny Saturday! More later.

2023-03-25 20:32:18 @stanislavfort @deliprao @natfriedman Thanks! This is more convincing.

2023-03-25 18:00:11 @deliprao @natfriedman See also: https://t.co/ihSmbbRGCJ

2023-03-25 17:58:46 @deliprao @natfriedman Maybe, but this was a widely published example from a well-known person, probably given to earlier versions of ChatGPT. I'm not saying GPT-4 can't do stuff like this but this specific example is not convincing. Need better ways to test.

2023-03-25 15:57:18 Daniel Ellsberg is a national hero. https://t.co/goyLPuay7K

2023-03-24 23:10:18 @dileeplearning

2023-03-24 23:05:08 @not2social @LakeBrenden ArXiv version here: https://t.co/G9xQyQuMcZ

2023-03-24 20:17:12 5:30pm Eastern (not 5:50pm) -- looking forward to watching this! https://t.co/1GIsCot5Ze

2023-03-24 14:54:40 @Leoagua1 @DKThomp @TomerUllman https://t.co/ihSmbbR8Nb

2023-03-24 14:52:48 Impossible to do reproducible experiments using ChatGPT. https://t.co/kn1IqyP7i5

2023-03-23 20:00:14 @phillipharr1s @Thom_Wolf Not clear that what we get with the ChatGPT (GPT-4) interface is from the same model that they were using. What we're getting now may be a watered-down version.

2023-03-23 17:22:38 @tdietterich @info_sprinkles @erikbryn But how much of humanity's knowledge is encoded in language and images that are used in training these systems? How much is missing?

2023-03-23 14:43:23 @Thom_Wolf Agreed! But anything mind-blowing in science begs for independent replication / probing!

2023-03-23 01:55:54 @DKThomp The study you cite was convincingly debunked by @TomerUllman. among others. https://t.co/Z9ab5YHXGB

2023-03-22 21:57:29 @AlisonGopnik Thanks so much, Alison!

2023-03-21 20:58:55 @YablokoU arXiv version here: https://t.co/G9xQyQuMcZ

2023-03-21 19:57:25 Perspective paper by myself and David Krakauer is now published in PNAS: "The Debate Over Understanding in AI's Large Language Models". https://t.co/jIsphK8IWD

2023-03-21 15:15:20 @miguelisolano @random_walker @tdietterich My point was not about any specific example, but about the need for evaluation of systematic vs. pointwise understanding.

2023-03-21 15:14:20 @miguelisolano @random_walker This gets to useful distinction made by @tdietterich btw "pointwise understanding" ("providing appropriate responses to individual queries") versus “systematic understanding” ("AI system is capable of providing appropriate responses across entire range of queries / situations.

2023-03-20 22:46:35 "GPT-4 and professional benchmarks: The wrong answer to the wrong question". https://t.co/cdoiKNLfYX

2023-03-19 20:48:52 This video on AI creativity is really good! https://t.co/qgqCNo007S

2023-03-19 19:11:19 @DavesBasilisk There's no agreed-upon definition of "AGI".

2023-03-19 18:42:41 @mmitchell_ai Even without being given the actual paper! https://t.co/Vi2KhqU4vL

2023-03-19 16:25:37 @davidchalmers42 Even if they "pass the bar exam" doesn't mean they can do the work of a lawyer.

2023-03-19 16:24:57 @davidchalmers42 Important clarification! Also, not clear what "near-human-level" performance is . Just because LLMs are near-human-level on benchmarks like SuperGLUE, etc. doesn't mean they are near-human-level on language understanding more generally.

2023-03-19 16:05:05 LLMs are an astounding technology but I think it's wrong to glibly assert that this will be a primary path to human-level AI. "Human-level AI" doesn't have a clear definition right now, but even so LLMs aren't (yet) obviously on the path to this end. https://t.co/8d6WyEKe9L

2023-03-16 23:19:24 @sotirov @cwarzel OMG

2023-03-16 22:02:43 Really good essay by @cwarzel on our current moment of "radical uncertainty" in AI. (I talked with him about my own struggle in trying to make sense of it all.) https://t.co/CIEVEMBhvO

2023-03-16 17:53:01 @fchollet When you say GPT-3 scored ~0 on ARC, was the prompt for a task simply the json file for that task? Is anything written on the GPT-3 results?

2023-03-16 16:53:26 First time I've seen a CAPTCHA using "same/different" task. Will multimodal GPT-4 be able to crack these? https://t.co/ZceT52w5y6

2023-03-16 14:52:02 @electrode3D @fchollet Can you send the prompt you gave to Bing and its response? Not sure what it means to try an ARC task on a text-only LLM.

2023-03-15 20:59:03 RT @sfiscience: Next week @MelMitchell1 &

2023-03-15 17:36:59 RT @mmitchell_ai: Are the AIs we're starting to see sentient? Conscious? All-knowing?! First, you must consider: The Eliza effect! A nice w…

2023-03-15 16:19:54 This was a terrific workshop, and now the talks are online https://t.co/7hoRQLa6X3

2023-03-15 14:13:53 RT @djleufer: Good piece on GPT-4, but I have an issue with this comment from one of the interviewees: “It’s exciting how evaluation is now…

2023-03-14 23:11:40 @mmitchell_ai Actually I've gotten a few where they think I'm you.

2023-03-14 23:10:34 @mmitchell_ai They think you're me. (JK )

2023-03-14 21:52:28 @yo_ean Note that Kurzweil and Kapor spent a huge amount of time (in their long bet) discussing the details -- in 2002.

2023-03-14 19:48:44 I have heard many people say "now we have AI that can pass the Turing test". But it's important to note that there is no such thing as *the* Turing test. The details matter very much! (1/2)

2023-03-14 18:03:19 @jjvincent The sentence about parameters now seems to be gone from the WP story!

2023-03-14 17:48:27 @Thom_Wolf Maybe, just maybe, there are limits to the progress we can expect from current scaling / RLHF approaches?

2023-03-14 17:01:50 RT @gdb: We’re releasing GPT-4 — a large multimodal model (image &

2023-03-14 15:38:14 Chomsky et al.'s NYT OpEd did one good thing -- it transformed the Connectionsts mailing list from a dull list of paper and job announcements into a fiery discussion forum on the nature of understanding.

2023-03-14 03:02:38 Also worth reading (and cited in the Noema article above), by Margaret Boden, written before the LLM revolution, but very relevant to how to think about today's AI: https://t.co/xgx9aj17nv

2023-03-14 02:59:32 This is an insightful article about LLMs (by @Jake_Browning00 and @ylecun): https://t.co/2vJPqjLLCV

2023-03-13 16:48:56 https://t.co/KIsLVqlfZK

2023-03-10 16:19:31 Just to clarify -- the ones in the first screenshot are all jokes created by humans (that ChatGPT has in its training data) but the one in the second screenshot was generated by ChatGPT.

2023-03-10 16:05:50 All the GPT-4 superpower predictions people are making is reminding me a bit of Chuck Norris jokes. I asked ChatGPT if it knows any Chuck Norris jokes, and then to make a Chuck-Norris-style joke about GPT-4. It did pretty well!! https://t.co/v6Qi4bDr0C

2023-03-10 14:35:21 @LTF_01 I believe the student in question was Gerry Sussman, who went on to a long and important research career!

2023-03-10 13:22:41 1966: Marvin Minsky "hired a first-year undergraduate and assigned him a problem to solve over the summer: connect a television camera to a computer and get the machine to describe what it sees." --Crevier, AI: The Tumultuous History of the Search for Artificial Intelligence

2023-03-08 15:37:36 @Turkey0guz Yes! See https://t.co/sHkcpL4nDw

2023-03-05 21:25:47 Be aware: AI deep-fake audio is making this classic scam much worse. https://t.co/9zkHmy4smc

2023-03-05 10:00:00 CAFIAC FIX

2023-03-02 22:00:00 CAFIAC FIX

2023-02-28 02:47:28 @ESYudkowsky @jim_rutt I hope you enjoyed reading Copycat's source code! I loved writing code in Lisp.

2023-02-28 02:46:42 @ESYudkowsky @jim_rutt Here's one: The notion that there is a strict division of human activities into "cognitive tasks" and other tasks. (See, e.g., https://t.co/9VKYTVNL48) I'm sure @glupyan has his own examples.

2023-02-27 21:59:16 @labenz Interesting thread! But...ChatGPT did not pass the bar exam. It did not pass the same medical licensing test that doctors take. See: https://t.co/NM83a7GcDy

2023-02-27 20:39:16 @greg_j_davis Pretty much!

2023-02-27 20:18:33 @Scott_E_Page Also doesn't finish carpentry require mathematics?

2023-02-27 20:17:30 @Scott_E_Page Doesn't scientific research often require carpentry / other "physical" tasks (e.g., building particle accelerators)? I'm not sure these can be separated so easily.

2023-02-27 19:19:26 @Dan_Rowinski I'm not sure what an "intellectual task" is either.

2023-02-27 19:03:39 What is the definition of a "cognitive task"? Serious question. https://t.co/9VKYTVNdeA

2023-02-27 18:36:29 This is very funny and actually quite informative about AI. (And I was thrilled to see my AI book's cover flashed on the screen for a microsecond ) https://t.co/uQQJW3QXII

2023-02-27 01:00:00 CAFIAC FIX

2023-02-19 03:11:39 @_onionesque Some info here: https://t.co/gaoNElIFbl

2023-02-17 22:19:01 RT @mpshanahan: I've just released another updated version of my paper "Talking About Large Language Models". This version has a whole new…

2023-02-17 20:51:08 @samfbiddle https://t.co/jpkbFR2WNv

2023-02-17 18:08:56 RT @NewYorker: Today’s Daily Cartoon, by Ivan Ehlers. #NewYorkerCartoons https://t.co/lAsPCofIot

2023-02-17 17:59:54 RT @mpshanahan: Not only should we not anthropomorphise chatbots, chatbots should not anthropomorphise themselves. That is to say, chatbots…

2023-02-17 16:25:31 @sharongoldman I agree -- the crazy behavior of Bing AI is likely a result of Microsoft training it to be as "engaging" as possible.

2023-02-17 16:24:30 @sharongoldman Given the strong pressures toward anthropomorphism, people need to be educated (esp. by journalists) on how to correctly / critically think about these systems. Not easy, but otherwise these systems are likely to be harmful, esp. to vulnerable people.

2023-02-17 16:18:09 @sharongoldman See https://t.co/EcIg9t0b8T

2023-02-17 16:05:29 This discourse gets dumber and dumber. Soon all news will be journalists asking LLMs what they think about stories by other journalists. Journalists: please stop anthropomorphizing these systems! https://t.co/WRsjN7hNoz

2023-02-17 15:17:06 If you read the other "LLM Theory of Mind" paper, or if you didn't, read this https://t.co/O8Wa6FnJUt

2023-02-16 21:06:14 @michael_nielsen I think you mean "Don't trust, and verify".

2023-02-16 02:21:57 @panyalarajeshw1 @sfiscience I'm happy that you enjoyed the course!!

2023-02-15 22:57:32 See also https://t.co/ybSAlH6eCv

2023-02-15 22:53:31 How search engines work: 1990s: Page Rank 2023: Rage Rank https://t.co/3lAEBwhUCe

2023-02-15 18:14:26 @jjvincent Is there a Bing chat vs. head of lettuce meme yet?

2023-02-15 16:12:36 Great piece by @Melissahei (and love the Pinocchio/search graphic!!) https://t.co/S79pPYj46C

2023-02-15 03:17:40 @zacharylipton @ChrisGPotts @chrisdonahuey How do you respond when a talk/conference organizer asks for your slides a week in advance?

2023-02-15 00:44:29 @marvinvonhagen @mark_riedl "I will report you to the authorities" https://t.co/2JZtoGgNGo

2023-02-15 00:16:33 Need a German word for all the posts on the Bing / Bard failure cases. All I could come up with was "sCHATenfreude"

2023-02-14 20:53:22 @mmitchell_ai I would like to believe these are real transcripts but am a bit skeptical. Or is Bing AI really that ridiculous?

2023-02-13 16:53:50 Of interest, apropos of the Large Language Model "Theory of Mind" discussions (h/t @jaaanaru): https://t.co/AqdEGIRGoO

2023-02-13 01:00:36 And check out @belindazli 's interesting recent paper as well! https://t.co/LsuqwKCGUz https://t.co/WzsivlUKHB

2023-02-12 22:19:10 See also: https://t.co/wzbtek9EXs

2023-02-12 22:18:33 My take on the recent GPT 3.5 "theory of mind" paper. It's consistent w/ earlier results, e.g., from @Maxwell_Nye et al., that LLMs can keep track of variables &

2023-02-12 18:02:44 @beenwrekt @mraginsky @yoavgo Not as much as 280 characters

2023-02-12 18:00:52 @beenwrekt @yoavgo @mraginsky Yes, but it's often worth pointing out explicitly. I appreciated @mraginsky 's contribution to that.

2023-02-12 17:56:06 @beenwrekt @yoavgo @mraginsky Right, not a critique, but pointing out that the story is more complex and nuanced than as portrayed in the New Yorker piece.

2023-02-12 17:46:11 @beenwrekt @yoavgo https://t.co/RsFNKO551o https://t.co/hFyGFzmAE9 https://t.co/VL6DWYcNZu These are the ones I've seen so far. I think that lossy compression is part of what's going on in these models, but not the whole story by any means.

2023-02-12 17:12:05 @fchollet Did you see this paper, https://t.co/W3JCdkqW25, and blog posts on it https://t.co/sHkcpL4Vt4 (my commentary) https://t.co/QjcgIFrkUd (response from @TaylorWWebb) I found that GPT-3 (davinci-003) >

2023-02-11 20:33:35 @geomblog @BrownCSDept BTW, in case you didn't know, another great writer and Brown CS grad is @brianchristian !

2023-02-11 18:08:06 @willknight @WriteArthur See my blog posts on this, "Did ChatGPT Really Pass Graduate-Level Exams?" Part 1: https://t.co/HmYe8T95bo Part 2: https://t.co/NM83a7GKt6

2023-02-11 17:52:40 Part 2 of my blog on ChatGPT being given graduate-level exams: https://t.co/NM83a7GKt6 https://t.co/s7F8acuI6P

2023-02-10 17:38:14 RT @mraginsky: A thread on Ted Chiang's New Yorker piece about ChatGPT. WARNING: this will get into technical detail about lossy compressi…

2023-02-10 16:48:39 @doristsao I meant no one has figured out precisely *how* these LLMs can do arithmetic / math, and why they often succeed but sometimes fail.

2023-02-10 16:43:12 @doristsao No, I don't think anyone has figured this out.

2023-02-10 16:01:03 Good thread on the Ted Chiang article in the New Yorker: https://t.co/hFyGFzm2OB

2023-02-10 15:48:22 New on my blog: I ask "Did ChatGPT really pass MBA, law school, and medical school tests?" I look at the claims, and dig into what really happened, and what it means. Part 1 is here: https://t.co/HmYe8T8xlQ

2023-02-09 18:20:00 @sfiscience

2023-02-09 17:01:33 But there's a selection effect -- we only see the errors that got caught. https://t.co/lgN7FP7X8T

2023-02-09 15:44:12 The show will take questions from live callers as well as from Facebook and Twitter.

2023-02-09 15:42:21 I'll be on public radio's Science Friday (@scifri) tomorrow (Feb 9) along with @ruchowdh, live at 2:20pm Eastern Time. We'll be discussing and answering audience call-in questions about generative AI.

2023-02-08 15:21:31 @NeilRaden

2023-02-08 15:19:10 @AISupremacyNews Thanks! I'm overloaded on writing projects, unfortunately.

2023-02-07 18:35:33 @Grady_Booch Arise, fair Sundar, and kill the envious Bing Who is already perplexed in scale and grief, That thou, our Bard, art far more fair (&

2023-02-07 02:46:53 ChatGPT Forced To Take Bar Exam Even Though Dream Was To Be AI Art Bot https://t.co/XOtTpe7oSM

2023-02-07 02:32:24 Incredibly proud of what Complexity Explorer has accomplished. Check it out: https://t.co/eEfnlfX1Wl https://t.co/eOsSclLEGo

2023-02-05 14:41:52 @dileeplearning Interesting. Related point was made in https://t.co/LUca45YykL --- watermarking method didn't work well on training data text memorized by the LLM because probability of the exact text was too high.

2023-02-02 20:22:23 RT @TAGinDS: Excited to have @jacspringer speaking at the latest TAG-DS seminar on "What can adversarial examples tell us about similaritie…

2023-02-02 19:53:07 RT @mark_riedl: The AI research community has a lot of old baggage that needs to be thrown out https://t.co/vSYtU8cWjC

2023-02-02 19:43:26 @ZennaTavares Cool work described in the blog post. I look forward to reading the paper.

2023-02-02 19:23:38 @PublicHamilton Thanks!

2023-02-01 05:15:18 @michael_nielsen Not in SF, but whenever we visit family in L.A. we do some "secret stairs" walks, which are always wonderful. https://t.co/ies9kJ5cZB The same author now has a guide to secret stairs walks in the East Bay: https://t.co/YiZrpohVk9

2023-02-01 02:34:52 RT @_eric_mitchell_: Very excited to share this demo of DetectGPT! Looking forward to feedback of all kinds, even re: my questionable web d…

2023-01-31 21:33:20 @_eric_mitchell_ @yoonholeee @jwkirchenbauer @jonasgeiping @tomgoldsteincs

2023-01-31 21:29:06 New on my blog: Exploring some recent proposals for detecting text generated by large language models. https://t.co/SK2oN3EuGZ

2023-01-31 17:14:14 Very cool demo. https://t.co/ouEt9fSuzm

2023-01-31 16:15:04 Very exciting for us at SFI! https://t.co/2caPXZq1M7

2023-01-30 15:32:00 @isEdgarGalvan I'm so glad to hear that you are enjoying it!

2023-01-30 01:00:00 CAFIAC FIX

2023-01-13 16:06:54 RT @svscarpino: There's no better way to spend the summer than at the @sfiscience Complex Systems Summer School. My experience there shap…

2023-01-12 23:40:44 Thread and new paper about what #ExxonKnew about global warming, back in 1977. They predicted it and then actively worked to falsely convince the world it wasn't happening, for the sake of profit. https://t.co/JKtK5k4xnk

2023-01-11 20:47:53 New blog post: @TaylorWWebb, co-author of fascinating paper on GPT-3's analogy-making ability, responds to my perspective: https://t.co/QjcgIFrkUd

2023-01-10 19:55:10 RT @sfiscience: Join us this summer for the 2023 Complex Systems Summer School, a 4-week residential research program hosted by SFI &

2023-01-10 18:35:12 RT @criticalneuro: Friends of the #learningSalon, please join us for 2023's 2nd session: Fri Jan 13th 4 PM ET Myself, @blamlab @MelM…

2023-01-07 23:50:47 @PessoaBrain @GaryMarcus At least it was sorry for its mistake

2023-01-06 20:49:58 International @sfiscience summer school on Intelligence and Representation For PhD students, in Cambridge, UK, August 13-25, 2023. Tuition is free! Apply by March 1, 2023. https://t.co/kQzvB3mesb

2023-01-06 17:38:44 Interesting discussion of the question "what are large language models good for?" https://t.co/5kdzEHKu6N

2023-01-04 23:14:47 @aviadt @TaylorWWebb Interesting! I hope someone will be able to explain all this someday.

2023-01-04 22:02:31 @KameronDHarris Both academic and a good read. But definitely academic.

2023-01-04 21:41:50 Top of my reading list: https://t.co/CDeqEFDAWT

2023-01-04 02:23:23 @mutant_liberal No, but there are a lot of theories / experiments that have shed light on this.

2023-01-04 02:22:56 @schulzb589 I don't understand the logic here.

2023-01-04 00:23:35 Very clear and interesting post about LLMs . Unlike Yoav, I *do* think it matters for their performance whether and how they "understand". But I really enjoyed reading this. https://t.co/ymR3IcEv4K

2023-01-03 23:45:57 @spaniel_bashir @gradientpub Yes, very interesting! But as you'll see, I'm not on board with their broad claims.

2023-01-03 23:07:24 @TaylorWWebb @dileeplearning

2023-01-03 23:06:21 One of my new years resolutions is to blog (from time to time) about interesting work in AI. I'm trying out Substack for this. My first post is a perspective on recent paper by Webb et al., "Emergent Analogical Reasoning in Large Language Models". https://t.co/sHkcpL4nDw

2023-01-03 18:53:55 RT @criticalneuro: Friends of the #learningSalon, start the year off right &

2023-01-02 19:21:40 @C4COMPUTATION @cigitalgem Yow, sorry to hear this! Have you contacted police / FBI?

2023-01-02 19:08:55 @Plinz Yup, unlike Google Search, ChatGPT will confidently find you things that don't exist yet

2023-01-02 19:05:50 @C4COMPUTATION @cigitalgem Any advice for Jessica?

2023-01-02 19:03:56 .@bendee983 explains why ChatGPT will not work as a search engine anytime soon: https://t.co/ndPMD0Qn3O

2023-01-02 18:42:46 @petrussola Thank you!

2023-01-01 21:11:07 Happy new year and brave new world! https://t.co/22ya7crvGq

2023-01-01 18:59:29 @Parisa__Rashidi @fchollet Yes, it seems that Minecraft is more appealing to girls than many other video games!

2023-01-01 18:56:56 @npparikh @fchollet I wish I knew! I personally was never interested in computer / video games

2023-01-01 18:45:54 @fchollet Playing video games was definitely the impetus for my two sons to learn to code at a young age. I don't think video games have enticed as many girls into programming, unfortunately.

2022-12-30 19:30:15 The Learning Salon is a weekly forum in which we explore bridges and contentions in biological and artificial learning, with amazing guests. Events are livestreamed every other Friday at 4pm EST. Go to https://t.co/u6LOHbWCW0 to register and to see upcoming guests. (2/4)

2022-12-30 19:30:14 I have thoroughly enjoyed attending The Learning Salon for the past two years, and now I get to co-host with @criticalneuro and @blamlab ! (1/4)

2022-12-28 19:45:32 @SSTaylorBooks Silver lining!

2022-12-28 18:08:40 Me: It will be so nice to finally fly home today after long visit with family. Southwest Airlines: lol

2022-12-26 19:15:16 "The engineers of the future will, in a few keystrokes, fire up an instance of a four-quintillion-parameter model that already encodes the full extent of human knowledge (and then some)"

2022-12-26 19:15:15 Very interesting and provocative article (not sure I agree with conclusions, but they are worth considering seriously). https://t.co/AsIHz0s35D

2022-12-23 22:26:07 Lol https://t.co/Iuaib9PAsi https://t.co/vIE1EzKGDO

2022-12-23 17:00:45 @DataScienceHarp I'm so happy to hear this!

2022-12-23 16:59:46 @EricHallahan @evanthebouncy https://t.co/KQMD32wr7S

2022-12-23 15:31:13 @WiringTheBrain @LiteraryVienna @SMcfarnell Analogies (like this one) are the highways thoughts drive on.

2022-12-23 03:28:51 @barrygoldman1 Read my book to find out! :-)

2022-12-23 02:00:30 Not a new edition but I will likely just put new chapters on my website. Stay tuned!

2022-12-22 22:22:10 RT @emollick: Tom Lehrer has put all his songs online (including lyrics &

2022-12-22 22:10:45 How does modern AI work, what is its history, and how does it compare with human intelligence? My book gives an in-depth, accessible guide for lay people &

2022-12-22 02:46:19 @AndrewLampinen I'll post some thoughts soon.

2022-12-21 17:39:09 So many interesting LLM papers to read on "vacation" over the holidays! https://t.co/HakineL0yR

2022-12-21 17:18:17 @ncasenmare I'll read this thread (and also new paper https://t.co/wDXO6yYcNo) in next few days and post some thoughts. Cheers!

2022-12-21 02:56:25 RT @TaylorWWebb: Here's one to add to the LLM debates: in a new paper, we found that GPT-3 matches or exceeds human performance on zero-sho…

2022-12-21 02:48:55 Some holiday reading from @TaylorWWebb et al https://t.co/W3JCdkHZ45

2022-12-17 00:01:45 Hi @twitter, I just wanted to let you know that I bravely ignored your warning and went to the link, and guess what --- it's safe! So it's okay to stop warning all of us about this! https://t.co/KFn9fjxJHx

2022-12-16 20:02:10 Looking forward to this! https://t.co/mdNDPyXkYZ

2022-12-16 15:57:34 This is today, Friday Dec. 16, 4pm EST. Hope to see you "there"! https://t.co/vatRrPABpR

2022-12-16 03:42:08 Where else to find me: s i g m o i d .s o c i a l/@melaniemitchell https://t.co/ntQPCL85Yq

2022-12-16 03:35:09 Twitter is trying their best to block anything to do with Mastodon. Here's what happened when I tried to update my Twitter bio with a link to my Mastodon account: https://t.co/lUeXA2AbFI

2022-12-16 03:20:12 RT @gradientpub: Sigmoid Social is approaching 5k active users and has hosted 125k interactions so far! If you're in AI, this is a Mastodo…

2022-12-16 03:17:53 Things seem to be going downhill very fast @twitter.

2022-12-16 01:15:26 @mmitchell_ai @not2social Follow all the M. Mitchells in AI! And the K. Mitchells too (@WiringTheBrain)! And follow us on Mastodon as well, since who knows how long we will be able to stand it over here.

2022-12-15 17:52:17 I was lucky to get a chance to talk with @spaniel_bashir for the Gradient podcast! https://t.co/qlgZ1AhHU2

2022-12-15 04:08:42 @FelixHill84 @davidchalmers42 It would be so interesting to be able to trace back LLM outputs to training data -- I wish there was a way to do this.

2022-12-15 04:06:39 @FelixHill84 @davidchalmers42 Indeed, it's clever! But the prompt did ask for a linux code block to be included in the poem. If that hadn't been explicitly asked for, I'd have been more blown away. Also, seems like "sudo apt-get install [food item]" is a thing on the internet. E.g. https://t.co/I7Lly9KuQX)

2022-12-13 21:40:18 Here's the real ChatGPT song about "EM in the style of Dylan" https://t.co/g6BUEPEP9w

2022-12-13 21:30:34 @WeidmannMark Did Elon himself write that song? Sounds like it.

2022-12-13 17:52:30 The Learning Salon (https://t.co/dTxWdW6osD) is back! And I am thrilled to be co-hosting. Fridays at 4pm Eastern. https://t.co/vatRrPABpR

2022-12-13 14:59:08 Though I am not convinced yet this was really written by ChatGPT...

2022-12-13 14:57:30 My newest column for @QuantaMagazine: What Does It Mean to Align AI With Human Values? https://t.co/g002PRuNG9

2022-12-13 01:15:58 AI vs. @alyankovic

2022-12-13 00:58:31 Omg https://t.co/SCaafWDWJL

2022-12-13 00:51:47 This is pretty brilliant. https://t.co/n3sdGfn1ud

2022-12-10 20:32:40 @dmonett Same!

2022-12-10 19:37:02 Good essay from @mpshanahan on the perils of using anthropomorphic language to describe the achievements of large language models. https://t.co/Z3UvwZyOa3 https://t.co/xi39FxpELL

2022-12-09 15:12:50 @yoavartzi Can someone get a Turing award posthumously?

2022-12-09 00:56:08 @fchollet Totally a blast! https://t.co/AJ69b3VN48

2022-12-09 00:17:02 Breaking news for language models used in search engines: Different people have the same name! https://t.co/xVxFfCEABr

2022-12-09 00:12:57 @AravSrinivas @perplexity_ai Disappointing. I don't think this way is going to be very effective.

2022-12-08 23:59:14 @AravSrinivas @perplexity_ai Thanks for looking into it. How are you going to get your system to return more reliable information in the future?

2022-12-08 22:19:51 @PWGTennant @random_walker Maybe I forgot to tweet that day.

2022-12-08 19:46:57 @random_walker At least it doesn't think you're dead https://t.co/GYse3nLOZg

2022-12-08 19:45:21 .@perplexity_ai, you are not making my day a happy one https://t.co/WnLdqo0MUO https://t.co/XjJaGNhIPd

2022-12-08 16:55:22 RT @daveyalba: New from me: ChatGPT is awe-inspiring tech. It wrote a wonderful poem about my cats! But it's plagued by the same issues of…

2022-12-08 16:09:46 Pretty close vote. Only 7 years to wait for the final result. https://t.co/nL6KCQLX61

2022-12-08 13:00:00 CAFIAC FIX

2022-12-07 08:00:00 CAFIAC FIX

2022-11-14 16:16:04 Important discussion by Ernie Davis on problems in Big-Bench common-sense language benchmarks: https://t.co/rMDtWg1l9F

2022-11-12 17:56:54 RT @C4COMPUTATION: COLLECTIVE INTELLIGENCE (@Collective_CI) is recruiting a Social Media Fellow. If you have strong writing + conceptual…

2022-11-12 16:18:22 @bestofdyingtwit To thine self-own be true.

2022-11-11 16:18:10 RT @brianbuccola: Can baboons do compositionality? That is, can they form compositional representations like "not blue"? In a new paper wit…

2022-11-10 16:56:50 @NicoleCRust Thank you -- I'm so happy to hear that you enjoyed the book!

2022-11-10 16:55:54 RT @NicoleCRust: This book by @MelMitchell1 is one of the most gripping science books out there (and I've read a lot). Touches on what make…

2022-11-09 19:03:58 A Twitter checking account with $8 fee for blue checks? https://t.co/RoMifpxo8Y

2022-11-08 16:46:43 @DoctorJosh @Kriegmerica @uvmvermont Congratulations, Sam!

2022-11-08 15:33:13 Some things that are not actually blue: https://t.co/rw7Y7YP6NG

2022-11-07 15:58:52 Interesting essay competition from https://t.co/DDdNqtu5Cy with ~$2,000 prize: "What fundamental principles of intelligence must be considered in the successful design of artificial intelligence?"https://t.co/gkyNo9DaIh

2022-11-07 15:16:28 @wildebees https://t.co/C9jSJjLDTG

2022-11-07 03:49:01 Funny that virtually every post on my Twitter timeline is about Twitter itself. And every post I've seen on Mastodon is about Mastodon. Social media has literally become a *meta*verse.

2022-11-07 03:36:16 Mastodon tip: I used https://t.co/jFg2g6s4DE to automatically go through the list of people I follow on Twitter &

2022-11-06 19:32:19 @melaniemitchell@sigmoid.social

2022-11-06 19:30:20 I'm staying on Twitter for the time being but am also trying out @gradientpub's new Mastodon instance -- you can find me at: https://t.co/OIKTFCYW7v

2022-11-02 21:13:33 SFI's In memoriam for Chuck Stevens: https://t.co/g3dAZY8ucT

2022-11-01 17:16:42 @pomeranian99 “Think how hard physics would be if particles could think.” -- Murray Gell-Mann

2022-11-01 15:12:16 Per chi parla italiano: https://t.co/1KF7FwCjxm https://t.co/RwtRX19Gxz

2022-10-31 18:48:16 Very fun conversation with Ben and Vaden: https://t.co/246tbYGf3q

2022-10-31 01:24:37 Finally some good news in the world. https://t.co/HRUx1hEAWY

2022-10-28 15:26:53 @TiernanRayTech @ylecun @GaryMarcus Lots of people care, of course -- many went into the field precisely to work on those big questions.

2022-10-28 15:02:25 @TiernanRayTech @ylecun @GaryMarcus https://t.co/G9xQyQdaOp

2022-10-27 00:25:24 @TheWill Hope you get better soon!

2022-10-26 17:34:00 Back in April, SFI hosted a fantastic workshop on "Embodied, Situated, and Grounded Intelligence". We wrote a report about it: https://t.co/mB6aHJUUH1Also, all talk videos are here: https://t.co/eB12lkU4Od https://t.co/mB6aHJUUH1

2022-10-26 14:08:48 @C4COMPUTATION @NadineSpychala @jlizier @RomainBrette @GaelVaroquaux @KordingLab Ditto!!

2022-10-26 14:05:25 New: A perspective piece by myself and David Krakauer about large language models and "understanding". https://t.co/G9xQyQdaOp

2022-10-26 14:03:01 @KordingLab @jlizier @RomainBrette @GaelVaroquaux @C4COMPUTATION I don't think AI is hitting a wall. My talk was about the many reasons I think people are over-optimistic about near-term human-level AI, and about some core aspects of human intelligence that are still missing in AI.

2022-10-25 15:57:41 @tyrell_turing @NSERC_CRSNG Great to see this -- congratulations!

2022-10-24 16:03:04 RT @bendee983: My review of "The Book of Minds" by @philipcball.Ball gives you a framework for looking past your instinctive tendency to d…

2022-10-24 15:50:55 @tdietterich @tinaeliassi @davecliff @arxiv Any advice for figuring out what the holdup is?

2022-10-24 15:06:38 @ankurhandos @arxiv Frustrating.

2022-10-24 15:01:45 @svscarpino @arxiv Thanks! I think I messed up by forgetting to choose a primary category (CS) and instead just choosing crosslists (AI/ML). Now they won't let me update it....sigh.

2022-10-24 14:53:19 @ankurhandos @arxiv Yup, I got the same reply.

2022-10-24 14:49:03 I submitted a new preprint to @arxiv 10 days ago --- it is still "on hold". No information except "Your submission was identified by arXiv staff or moderators as needing further attention." Anyone else have this experience? Can anything be done to find out what hold-up is?

2022-10-22 19:18:35 I'll also be doing a book signing after the panel -- come by and talk about complexity and/or AI!https://t.co/W0duysckWB https://t.co/ON0sZrp9pj

2022-10-22 19:17:09 If you're in Santa Fe and interested in intelligence (biological or artificial), come by SITE Santa Fe to hear our panel discussion: https://t.co/8cxjM4rnsRIf not in Santa Fe, follow on the YouTube livestream: https://t.co/aMfdexZRBehttps://t.co/8cxjM4rnsR

2022-10-22 18:08:06 Very sad. Chuck was a regular visitor to @sfiscience for many years and I was fortunate to get to know him. He was brilliant and interested in everything. I learned a lot from him and will miss him very much. https://t.co/RkfvzudNqe

2022-10-21 18:32:45 @cigitalgem Indeed!

2022-10-20 22:59:21 @AtomicAdMan Reminds me of when I tried using GPT-3 to generate horoscopes: https://t.co/Oj0Kp1faPm

2022-10-20 20:33:43 @TonyZador

2022-10-20 20:33:21 Good article on why "neuroAI" isimportant: https://t.co/vurFzbjFeF. I'm glad this was written. But it should have cited the Animal AI Olympics by Crosby, @mpshanahan, @LucyCheke and others, an attempt to achieve exactly the kind of "embodied Turing Test" the paper proposes.

2022-10-20 20:14:35 @deliprao @Penn @ccb Congrats!!

2022-10-20 13:10:04 @spiantado @FelixHill84 And something about lettuce. Seems important.

2022-10-19 19:16:00 @GaryMarcus Way optimistic even if you count in Mars-years.

2022-10-19 16:48:20 I found a positive way to channel my anxiety about the upcoming midterm elections: volunteering to write letters to citizens in key states, encouraging them to vote: https://t.co/l66v7oSb32

2022-10-18 22:03:47 @criticalneuro I had bad reaction to every covid shot but zero reaction to 2022 flu shot, much less than normal ¯\_()_/¯

2022-10-14 16:18:39 @B_Lindstroem @EikoFried Thank you!

2022-10-13 20:22:46 @rasbt So is driving a car in a city more like making coffee or like predicting protein structure?

2022-10-13 16:55:21 "'People interested in AI are very keen on anything that can play Pong,” Potter says. 'That was a brilliant decision."Yup, Pong is the new Turing Test https://t.co/kSW9pEzB2O

2022-10-12 23:52:29 @brianchristian @theNASEM @SchmidtFutures Congrats!!!

2022-10-12 22:21:58 @erikbryn @jackclarkSF @dashstander @fchollet Lots of papers on this in NLP , e.g., https://t.co/tB97aoiVlfhttps://t.co/8GJfBc4CguAnd workshops like this one: https://t.co/8DjQ6aK5be(2/2)

2022-10-12 22:21:44 @erikbryn @jackclarkSF @dashstander To answer your first question, lots of work on this recently, but not all in agreement with one another: @fchollet wrote about this at length in https://t.co/pRwx4NwqVaI also wrote about this at the end of https://t.co/JEDUOzQ8BW(1/2)

2022-10-12 21:52:37 @jackclarkSF @dashstander I was disagreeing with your analogy of "road running out". But I like your proposal for a general system (not trained for or fine-tuned on a particular dataset) being applied to the Perception test.

2022-10-12 21:51:09 @jackclarkSF @dashstander Sorry for bad analogy. My point was that ML systems to date can do really well on specific benchmarks (by fitting their gazillion parameters to some kind of statistical function) but then they often don't generalize well.

2022-10-12 21:46:09 @jackclarkSF @dashstander To warp the analogy further: a trillion parameters is a lot of statistical "fuel", and can make a car run perfectly on any particular type of road. It's not that the road runs out, but that car won't work on roads in general.

2022-10-12 21:11:18 @jackclarkSF Any particular benchmark will be "solved", but the question will always be in generalization and how it is measured.

2022-10-12 17:18:32 RT @sfiscience: Revolutionary technologies requires fundamentally new operational principles &

2022-10-12 16:44:47 Congratulations to @YejinChoinka on being awarded a MacArthur Fellowship!! https://t.co/s3id2uRRES

2022-10-12 02:32:14 RT @sfiscience: #IPFest returns on Oct 22-23 for a full weekend of panels, keynotes, avant-garde performances, &

2022-10-11 13:50:21 @grbradsk Indeed. Just as we tend to be unaware that our lives are going to become data of reinterpretation. Hopefully reinterpreting the past will help us look around us and reinterpret the present as well.

2022-10-10 21:48:50 I'm looking forward to participating in this event! https://t.co/tEGLAaSntK

2022-10-10 21:37:02 The book actually looks interesting. Fun note from the review: https://t.co/InT4NRLDxb

2022-10-10 21:27:19 Did women even exist back then? https://t.co/3JlpR0SSRU

2022-10-10 18:52:52 https://t.co/fqka6zVh6E

2022-10-10 18:51:47 Very good discussion of DeepMind's AlphaTensor. https://t.co/K8oxlnNVfd

2022-10-07 21:16:48 New method for Python error visualization. https://t.co/3MPt5WOldk

2022-10-07 17:18:54 Prospective postdocs in AI / Cogsci: Deadline is Sunday October 16 https://t.co/pPfgz9qyGR

2022-10-07 17:17:34 Deadline is Friday, October 14 https://t.co/Ts9BkHy7IJ

2022-10-06 16:52:59 RT @DimaKrotov: Are you a PhD student who wants to work on Transformers, Hopfield Networks, NeuroAI? Apply to our summer internship program…

2022-10-05 20:52:17 @svscarpino I'm in!

2022-10-04 12:19:49 @bleepbeepbzzz @GaryMarcus Good point!

2022-10-03 04:01:37 @GaryMarcus Let's think step by step. That should help! https://t.co/Kh2uUe8L5W

2022-09-30 22:20:07 @hardmaru @googlejapan Can't wait to see what you do next!

2022-09-30 15:07:42 @BryanCoxwell Very interesting! Thanks for posting. I wonder if my posing these questions to GPT-3 a while ago means that they were included in its ongoing training data....

2022-09-30 14:36:53 Very good post. (Note: they are not "AI Critics", but rather "AI Hype Critics".) https://t.co/MiyNyWg7dS

2022-09-29 17:18:31 @yoavgo Here's a link that should work: https://t.co/Xv4mmYJrbI

2022-09-29 14:26:25 @TylerMarghetis @UCMCogSci @sfiscience

2022-09-29 14:25:22 I'm very grateful for this honor from UC Merced Cogsci and looking forward to visiting campus next week! https://t.co/epJD1LKWkA

2022-09-28 23:34:03 @deliprao Weird2Vec

2022-09-28 21:55:13 @PaulTopping @math_dandy @GaryMarcus Some links here: https://t.co/vvejZrIe8i

2022-09-28 21:32:03 @AndrewLampinen @math_dandy @GaryMarcus Agree, it's clear that symbolic methods alone won't scale.

2022-09-28 21:20:17 @AndrewLampinen @math_dandy @GaryMarcus That being said, "toy" domains like ARC (&

2022-09-28 21:15:23 @AndrewLampinen @math_dandy @GaryMarcus How to define "nuanced context"? I think ARC captures some aspects of whatever that is. I agree about rich inputs &

2022-09-28 17:46:13 Come to SFI's Interplanetary Festival, October 22-23, either in person in Santa Fe or virtually, anywhere in the galaxy!https://t.co/FeJXtlCeaG

2022-09-27 22:04:25 @math_dandy @GaryMarcus See https://t.co/pRwx4NeQ3CAlso https://t.co/a6x1UlFfSUhttps://t.co/L6Rg54mund

2022-09-27 21:46:29 @math_dandy @GaryMarcus Perhaps this? https://t.co/33GkiWKHwB

2022-09-27 17:30:47 I just signed this open letter and encourage all scholars who support open publishing and research to do so. https://t.co/uFCb3k4lN9

2022-09-27 14:33:57 Annual AI twitter debate on "what is a symbol" is on again. Does it matter? https://t.co/IdLAXog3K0

2022-09-26 00:02:16 RT @DanLarremore: The SFI postdoc program is absolutely fantastic.• Supports risk-taking and creativity without losing rigor or community…

2022-09-24 16:18:41 RT @sfiscience: Announcing acclaimed novelist Tom McCarthy as our next Miller Scholar!“Tom is one of those lyrically synthesizing minds t…

2022-09-24 16:18:28 RT @mpshanahan: If you're an #AI PhD student with an interest in either a) reasoning with LLMs or b) learning abstract representations, the…

2022-09-24 00:16:41 @GeePawHill Unfortunately no authorized e-book versions. Sorry!

2022-09-21 15:03:53 Yes, I'm subtweeting https://t.co/RuE1Uu1oDB and https://t.co/UfhnIQ8834

2022-09-21 15:00:30 Like in all sciences, it's really important to have constructive debates about AI research, and to have scholars who write critically about specific results in the field . It's not helpful to brand these scholars quite generally (and negatively) as "AI Critics" or "AI Deniers".

2022-09-20 23:07:06 Thread on language understanding in LLMs: https://t.co/8edEpCQ783

2022-09-20 14:47:44 RT @C4COMPUTATION: Please join @sfiscience @TheLensic Tue+Wed, Sep 20+21, for our Ulam Lectures. @Cornell math prof + gifted science commun…

2022-09-20 02:52:09 @RRejeleene @sfiscience Sorry, I misunderstood your original question. The position is for people who have completed a PhD or will complete one soon.

2022-09-19 15:18:04 @criticalneuro @behrenstimb @Tim_Dettmers @WiringTheBrain @blamlab @tyrell_turing @jcrwhittington Thanks. I wasn't criticizing the paper at all, just asking about the claims in the Quanta tweet, which many more people will see than people will read the paper.

2022-09-18 18:11:42 "Hippocampus is a transformer" -- is this (1) true and amazing, (2) true and trivial, (3) untrue but useful metaphor, (4) nonsense, or (5) ???@WiringTheBrain @blamlab @criticalneuro @tyrell_turing https://t.co/uoyiF3Xan2

2022-09-17 15:43:54 Amazing and hilarious thread about simple ways to manipulate LLMs and the products that use them. https://t.co/c1I8NKwzaE

2022-09-17 14:41:20 Important article that thoroughly lays out the "what" and "why" of current threats to US democracy. Terrifying. https://t.co/4KACPE4KkO

2022-09-16 15:10:51 @falcon7201 Not all AI people focus on ML :-)

2022-09-16 15:04:38 Re-upping this also. Postdoc opportunities at SFI for early-career scientists interested in any aspect of complex systems. Apply by 10/14/2022. Please RT. https://t.co/wMoLsY3Ijx

2022-09-16 15:02:28 Re-upping this. Postdoc opportunity at SFI for AI/ML/cogsci people interested in abstract concepts, reasoning and analogy-making. Apply by 10/16/2022. https://t.co/pPfgz99vER

2022-09-15 17:46:48 Counterpoint to Atlantic article linked below: NNs &

2022-09-15 16:50:41 @roydanroy @EmtiyazKhan @_joaogui1 @maosbot Calling NNs or LLMs "magic" is not a good thing for scientific understanding, or even for general public understanding of what these technologies can do.

2022-09-15 16:49:56 @roydanroy @EmtiyazKhan @_joaogui1 @maosbot For example, "we" (the scientific community) don't fully understand the mechanisms of gravity, but we don't think of gravity as a "miracle", but rather as a natural phenomenon that will eventually be explained by science.

2022-09-15 16:49:43 @roydanroy @EmtiyazKhan @_joaogui1 @maosbot I don't think that's the way the general public interprets the words "magic" and "miracle".

2022-09-15 16:18:16 @EmtiyazKhan @roydanroy @_joaogui1 @maosbot More: (From https://t.co/geJxnO6Z7g) https://t.co/CPIrfXyclw

2022-09-15 16:11:20 @EmtiyazKhan @roydanroy @_joaogui1 @maosbot From an article published today in the Atlantic https://t.co/2zliXYBPfC

2022-09-14 22:06:12 When I see the claim that "AGI" will (or should) be at the same time generally intelligent AND more like a computer than like a human on tasks like arithmetic, I am reminded by this speculation from Gödel, Escher, Bach, which I think is probably correct. https://t.co/1lLWceIxqH https://t.co/Qd5VFXZZsO

2022-09-14 16:53:55 @chazfirestone @davidchalmers Top right looks like a chicken in a pear tree.

2022-09-14 16:52:43 @chazfirestone @davidchalmers and..... https://t.co/MwAzlPrhhf

2022-09-14 16:24:08 BBC covers really important energy study by Doyne Farmer and others: "Switching from fossil fuels to renewable energy could save the world as much as $12tn (£10.2tn) by 2050, an Oxford University study says."https://t.co/fzdHHY54bx

2022-09-14 16:04:28 US Media: Let's do nonstop coverage of the Queen.US Media: We've run out of things to say about the Queen. US Media: Let's do nonstop meta-coverage of the Queen. https://t.co/2FBOEdAjd2

2022-09-13 19:41:08 RT @bendee983: I spoke to @MelMitchell1 on concepts, abstractions, and analogies, three pieces that are missing from deep learning systems.…

2022-09-12 17:45:26 RT @sfiscience: Mathematical Stories with @stevenstrogatz! Our Ulam Memorial Lecture Series is back – tune in on YouTube or attend in per…

2022-09-12 16:53:17 RT @MelMitchell1: Today I went on a beautiful hike starting from Los Alamos ski area. After the hike I spent some time looking at exhibi…

2022-09-12 14:34:23 @KameronDHarris OMG.

2022-09-12 13:06:50 @curiouswavefn Yes!

2022-09-12 03:13:52 In 1943, when the ski club was formed, virtually no one knew of Los Alamos's existence. Two years later came the bombing of Hiroshima and Nagasaki. So many of these physicists later became activists against nuclear weapons. /end

2022-09-12 03:11:07 Extra credit for identifying the energetic Nordic skier from a ~1948 photo. /3 https://t.co/sIXLaaLTQ8

2022-09-12 03:11:04 Part of roster of initial members and how much they paid for their membership. A few more scientific legends here. /2 https://t.co/thnGbrZ6H6

2022-09-12 03:11:00 Today I went on a beautiful hike starting from Los Alamos ski area. After the hike I spent some time looking at exhibit in the ski lodge on the "Los Alamos Ski Club" from its beginning 1943/1944. Below:part of original sign-up sheet. See any Manhattan Project luminaries? /1 https://t.co/adS9fJ8PYF

2022-09-11 04:00:55 @RRejeleene @sfiscience Yes.

2022-09-09 18:10:14 Really looking forward to @stevenstrogatz's two lectures in Santa Fe (this year's Ulam Memorial Lectures series), September 20-21. If you can't attend in person, the lectures will be live streamed.https://t.co/Jvfq9jsWQDhttps://t.co/G95m8ylEPC

2022-09-09 16:16:53 RT @LongFormMath: When you were a carefree tenured professor living your best life, and then are suddenly forced to chair your department h…

2022-09-08 14:48:56 This will be fun and interesting!! Open to everyone. https://t.co/FcuDcRhfOZ

2022-09-07 14:50:52 @Abebab Congrats!!

2022-09-04 19:42:19 @databacked101 Also check out the work of Charles Ofria https://t.co/R9GiZFxwUm and Chris Adami: https://t.co/igL4VH6wWV

2022-09-04 19:33:27 @databacked101 You might look at the work of Lee Altenberg (https://t.co/W0L8xAjFhy) who has worked on applying GA-type models to biological questions. Also an interesting paper linking GAs to biology is this one: https://t.co/6tXINJJXbt

2022-09-01 20:32:01 @kthai1618 @sfiscience Yes, we occasionally have graduate interns, depending on funding. Email me about this and I can send a more detailed response.

2022-09-01 20:30:11 @ocramz_yo Depends what you mean by "meaningful." I think SOTA is now 30% correct on the test set (given three guesses for each task). However, I don't think any of the approaches so far will generalize very well.

2022-09-01 16:21:03 Exciting initiative for anyone interested in AI https://t.co/mNIxC6VINa

2022-08-31 19:31:25 This non-paywall link seems to work: https://t.co/TGgjiSOw1N https://t.co/BvAS8nKVg2

2022-08-31 19:19:24 @WiringTheBrain https://t.co/rgnIR64t3E

2022-08-31 17:52:30 FT interview with me about large language models. Sorry it is behind paywall. I am trying to get a non-restrictive link I can share. https://t.co/Mb0Jof2CgF

2022-08-31 17:51:00 RT @PWGTennant: .@MelMitchell1 again adds an important dose of reality to discussions about the current state of #AI! Very timely given I'm…

2022-08-31 16:50:54 RT @fchollet: A postdoc position where you will get to enjoy the New Mexico weather and views -- and work on solving ARC at the same time!

2022-08-31 15:28:48 Important nerditry question of the day: https://t.co/xTJIdCXsQ7

2022-08-31 13:06:38 RT @TonyZador: Are you trained in AI, and interested in doing original research at intersection of Neuroscience and AI? Come to CSHL and…

2022-08-31 01:19:54 @davidchalmers42 @rgblong I'd heard of "AI safety" but now I know what it actually means!

2022-08-30 23:24:17 @ev_fedorenko @mpshanahan @gershbrain @TomerUllman @LakeBrenden @todd_gureckis @neuranna @MelanieEMoses Please share with anyone you know who might be interested. Thanks! https://t.co/pPfgz98XPj

2022-08-30 19:45:46 @davidchalmers42 @rgblong Sneaky. https://t.co/rOMRMy8gzT

2022-08-30 19:12:17 RT @Kriegmerica: The https://t.co/LadozAqMdA is hiring @NorthwesternEng! We are looking for a postdoc to help us explore compositionality…

2022-08-30 17:17:51 New postdoc position in my group at @sfiscience to collaborate on developing and evaluating AI models of abstraction and analogy. https://t.co/pDm2jYQez5 Application deadline 10/16. SFI offers fantastic intellectual opportunities!(Also see views from our campus below.) https://t.co/LHW7X3INKM

2022-08-24 16:15:59 @philipcball Very well-deserved. Congratulations!!

2022-08-23 19:52:48 Yup, I agree 100%. See https://t.co/lqzCDdRPbvby @mpshanahan et al. for similar arguments. (I also wrote about this in my last book.) https://t.co/8gTWvj51nz

2022-08-23 17:47:49 @Vannatikk @kischiman I tried the same prompts in Craiyon -- honestly, it was not much better than SD.

2022-08-23 16:48:52 @akbirthko My tweet wasn't meant as criticism, just an observation. But not obvious to me that the system could have learned these abstract concepts robustly even if trained on examples. I'd love to see experiments on this!

2022-08-23 16:43:50 Related paper on DALLE-2 by Colin Conwell and @TomerUllman https://t.co/7XUb62gfAg

2022-08-23 16:41:07 A few more prompts:A: One cube on top of another cube (correct on one output)B: A small cube to the left of a large cubeC: A red cube below a green cubeI wonder if it could reliably learn these relationships if it had examples (with relevant captions) in training set. https://t.co/k5lM074jkb

2022-08-23 16:10:28 *Prepositions are hard.*Stable diffusion demo (https://t.co/HTkzn6LpGH)Prompt A: A small green cubePrompt B: A large red cubePrompt C: A small green cube on top of a large red cube https://t.co/4U8oKTfZUY

2022-08-18 15:57:24 "Why I am not an effective altruist" by @erikphoel Very thoughtful, and IMO correct. Worth reading. https://t.co/i5D6qD0XC9

2022-08-18 14:28:50 @ehud

2022-08-16 17:53:27 @WiringTheBrain Sent it to you via email.

2022-08-16 14:22:06 @CoolsHannes Thank you for the shout out!

2022-08-15 20:37:32 Can't argue with Twitter's "trending" algorithm https://t.co/jgz7BXLnb3

2022-08-15 15:01:21 @aaronclauset @CUBoulder @CUEngineering @BioFrontiers Congrats!!

2022-08-14 21:02:09 Lol https://t.co/ojFgJM9NuE

2022-08-12 15:49:48 Thinking about this made me painfully aware of how much of my writing time I actively spend *not writing* -- that is, doing things only for the purpose of avoiding writing.

2022-08-12 15:48:36 My favorite advice on how to write: "Don't not write."

2022-08-10 15:43:02 @spiantado @Zergylord @FelixHill84 @KarlKraatz @DeepMind If one's understanding of an abstract concept like "justice" consists of a simulation of a situation (or regenerated memory), would you consider that concept to have a "referent"?

2022-08-10 14:54:48 @Zergylord @FelixHill84 @KarlKraatz @spiantado @DeepMind While @spiantado and @FelixHill84 say there are "many terms that are meaningful to us but have no discernible referent at all, such as abstract words like 'justice' and 'wit'", Barsalou's view: these indeed have referents via situation-based simulations. Seems compelling to me.

2022-08-10 14:52:36 @Zergylord @FelixHill84 @KarlKraatz @spiantado @DeepMind Can picture theory of meaning (reference) refer not just to pictures but to entire mental simulations (a la Barsalou)? In his view, abstract concepts like "justice" or "wit" would be represented via mental simulations grounded in experience/situations.

2022-08-10 14:43:41 SFI's Postdoctoral Fellowships offer early-career scientists unparalleled intellectual freedom to pursue insights about complex systems that matter most for science and society. Apply now at https://t.co/l0fLuhOqHjAlso, fantastic views from our campus on a hill!(Pls RT) https://t.co/0TZbyyJkCo

2022-08-09 20:18:44 @jacspringer https://t.co/fBYPcwQe53

2022-08-08 22:16:35 Interesting profile of "Effective Altruism" movement, and why many of its followers became obsessed with "superintelligence" and "AI alignment". https://t.co/sU4kGiqqXm

2022-08-02 16:36:11 RT @SebastienBubeck: If you want to learn about adversarial examples and the current community consensus (or lack thereof) on its importanc…

2022-08-02 14:56:32 @gualtieropicc @PessoaBrain Thanks for the pointer!

2022-08-02 13:08:18 Yup, relationships are hard.Even for DALL-E 2. https://t.co/wVeBC8zTzw

2022-08-01 21:23:53 @fchollet Indeed, their arguments were not sound. But they sure generated a lot of discussion! I'm just wondering if anyone has come up with any new arguments, or if everyone now believes it's possible in principle.

2022-08-01 21:13:05 Tweeps: an AI / philosophy question: Several people (e.g., Searle, Penrose) have argued that machine intelligence is not just hard but impossible *in principle*. Has anyone made an (interesting) argument for that position more recently (e.g., in last 5 years)?

2022-08-01 17:25:00 If you're looking for a postdoc position, there is no better one than this one at @sfiscience: https://t.co/4KXWsB6a6fInterdisciplinary scientists in any field (including those in AI / ML / Cogsci) are encouraged to apply! Applications due October 14, 2020. https://t.co/QbJPGZBcKg

2022-07-31 15:27:22 @pomeranian99 Good essay. See also https://t.co/uFwFPOkmig

2022-07-27 14:56:45 @roydanroy @Abebab "cymbal-ic" AI

2022-07-25 22:16:19 @MaxBoot Excellent article. But what can ordinary folks do with the "wake up, people" message? We're already awake, but feel powerless.

2022-07-25 21:54:43 @seanmcarroll @juniperlov @C4COMPUTATION @mikha_ehl @aaronclauset @danieldennett

2022-07-25 21:54:42 StuBot, Digi-Dan -- what'll be the next SFI faculty bot?Maybe "Geoffrey West-world"? https://t.co/sRpjL28QuY

2022-07-25 03:38:07 "OpenAI's DALL-E 2 helps kids bring their world-changing ideas to life". For example https://t.co/9JRaB3Ssb1

2022-07-24 18:47:29 @mims Unexpected ending: Brin "is currently writing a physics textbook."

2022-07-24 14:50:24 If true, this is astounding and profoundly distressing: https://t.co/MVgblA2HYv

2022-07-22 16:13:19 Today, 10am PDT (1pm EDT) https://t.co/QMOIu0E4Xm

2022-07-22 15:57:47 RT @anilananth: Today, 10AM PT: Adversarial examples in deep learning? What are their dangers? Are they useful in making sense of deep neur…

2022-07-20 18:38:38 @DigPhysics See https://t.co/8zyFdvXP6N

2022-07-18 23:19:53 @rgblong That project was completely different from my dissertation topic, so it felt a bit like cheating to work on it. But it was also refreshing (when my actual dissertation work was hopelessly bogged down.)

2022-07-18 23:16:36 @rgblong It was also the start of a long and productive collaboration.

2022-07-18 23:16:19 @rgblong Forrest, S. and Mitchell, M. (1993). What makes a problem hard for a genetic algorithm? Some anomalousresults and their explanation. Machine Learning, 13, 285-319.It was really fun to work on this paper with Stephanie, with minimal input from any of our professors!

2022-07-18 23:14:44 @rgblong During graduate school, my fellow grad student and good friend Stephanie Forrest worked on a project together trying to explain surprising results from another student's dissertation. We ended up publishing the paper in Machine Learning.

2022-07-18 00:32:28 RT @MelMitchell1: @TMHolding Recommended: The Alignment Problem by @brianchristian

2022-07-18 00:30:57 @TMHolding Recommended: The Alignment Problem by @brianchristian

2022-07-18 00:28:34 RT @TMHolding: I don't read many books cover-to-cover but @MelMitchell1's Artificial Intelligence: A guide for Thinking Humans was great! E…

2022-07-18 00:12:24 @TMHolding I'm so happy to hear that you enjoyed it!

2022-07-17 15:29:05 @zacharylipton 18 months to AI Pioneer but at least 24 months to AI Godfather.

2022-07-15 21:32:41 @chriswolfvision I think she's saying LaMDA is more like the Berkeley library than it is like an intelligent agent that lives in and interacts with the world. It's hard to find the right metaphors / ways to think about these models, and I found this way interesting and maybe useful.

2022-07-15 21:14:28 Interesting take by @AlisonGopnik : Large language models are "cultural technologies" -- they are not themselves intelligent agents, but, like Wikipedia and libraries, they are technologies that allow humans to access the intelligence of other humans. https://t.co/92hswMk1D6

2022-07-15 19:15:18 @FelixHill84 @pmddomingos This reminds me of an interaction in grad school:Student: I know you generally don't like Beethoven, but is there any piece of his that you enjoy?Doug Hofstadter: I'll take the fifth.

2022-07-15 19:10:55 @FelixHill84 @pmddomingos Good answer.

2022-07-15 19:07:50 @FelixHill84 @pmddomingos I also think this is fair. But it begs the following question: Mozart: Bach :: Beatles :: ?

2022-07-13 20:35:42 Modern AI is alchemy https://t.co/kcUIGfvQWe

2022-07-13 17:08:00 I found this critique of Harari very interesting but depressing. It seems that being careful about the truth and limitations of the science you are communicating is often not an optimal path to bestselling books.https://t.co/6814wIoHpL

2022-07-12 16:26:57 Related articles by @blaiseaguera: https://t.co/9VnIwIAXZYhttps://t.co/j23k3VGHglRelated articles by me: https://t.co/IwkbWxtqZOhttps://t.co/uFwFPOBXGQ

2022-07-12 15:08:57 This Thursday, in person and on livestream: "How Close Are We to AI?" (And what does it all mean?) https://t.co/yY1wenaDtk

2022-07-12 15:06:31 The great @NaomiOreskes and @ErikMConway take on the myth of the "free market"! https://t.co/zZDig8gy4A

2022-07-09 19:12:06 Dear fellow Monetizable Daily Active Users (yes, that's you!),This is a very fun and informative article about the Musk / Twitter situation. Recommended!https://t.co/wuX1HIMceZ

2022-07-07 14:44:47 Upcoming talk on Zoom: https://t.co/6zALQKnSr6

2022-07-05 18:41:47 @protienking @raphaelmilliere "Though analogy is often misleading, it is the least misleading thing we have."– Samuel Butler

2022-07-05 16:51:02 @raphaelmilliere Yes, I agree, it is clearly not just mimicry. We definitely need a good metaphor to help people (and ourselves) understand what these models are and what they can/cannot do.

2022-07-05 16:19:30 Excellent, thoughtful analysis on what large pre-trained models are and what they can do from @raphaelmilliere.I agree that "stochastic parrots" is not the right metaphor -- he proposes "stochastic chameleons". Still not quite right, but closer. https://t.co/P1aXUktzSn

2022-07-05 16:04:12 Good thread on NeuroAI citations. https://t.co/NpVfpUGMlo

2022-07-01 14:48:43 @AjdDavison @ankurhandos I'm very happy to hear that you enjoyed it!

2022-06-30 18:00:22 If you live in the US and this worries you -- call your representatives and ask them to act --- first steps are to get rid of filibuster and expand Supreme Court. https://t.co/XD3Li56Qqx

2022-06-30 16:36:17 The discussion will also be livestreamed.

2022-06-30 16:33:19 Seattle folks: Town Hall Seattle @THSEAdiscussion / debate with @blaiseagueraand me has been rescheduled for Thursday, July 14, 7:30pm. Topic: "How Close Are We To AI?"It will be https://t.co/yY1wenaDtk

2022-06-29 17:32:08 @_KarenHao @MilesKruppa Is there a non-paywalled version?

2022-06-29 15:50:24 This paper will be published in the proceedings of the AI Evaluation Beyond Metrics workshop (at IJCAI-ECAI 2022): https://t.co/8DjQ6aK5be.Thanks to my co-author Victor Odouard, who is a recent graduate of Cornell and a current intern at SFI. Thanks also to @NSF for support!

2022-06-29 15:50:23 We then did something similar for the ARC (Abstraction and Reasoning Corpus) dataset of @fchollet.

2022-06-29 15:50:22 We tested two of the best (previously published) RAVEN solvers on our variations to see if they did as well on our concept variations as predicted by their accuracy on the overall dataset.

2022-06-29 15:50:21 https://t.co/dPyP0AjtAg

2022-06-29 15:50:20 https://t.co/SqmS4vGijn

2022-06-29 15:50:19 In this paper we created concept variations in the form of RAVEN-like problems on two concepts: "sameness" and "progression". These concepts appeared in the RAVEN dataset, but we created new variations. https://t.co/x9cnWAnOLQ

2022-06-29 15:50:18 We propose that a better evaluation is one that is *concept-based* -- that is, identify the underlying concepts, and probe the system with variations on these concepts, to see how well the concepts are actually understood.

2022-06-29 15:50:17 RAVEN has many tens of thousands of training and test examples. If an AI system (e.g., DNN) does well on this dataset, does it really *understand* abstract concepts like "progression" (e.g., in number of sides, as in the above example).

2022-06-29 15:50:16 Below is a sample problem from the RAVEN dataset (which is procedurally generated). On the left is a matrix of 8 shapes with different attributes. The objective is to choose the candidate answer (on the right) that best fits the 9th matrix element. Here the answer is 6. https://t.co/wNDgiONMkL

2022-06-29 15:50:15 New preprint from me and Victor Odouard (SFI intern): "Evaluating Understanding on Conceptual Abstraction Benchmarks"https://t.co/7f6S2m6ghAbelow.

2022-06-28 17:25:16 @mpshanahan And excerpt from Newell's 1982 paper "Intellectual Issues in the History of Artificial Intelligence", where he talks about what separates "artificial intelligence" from "other approaches to machine intelligence" https://t.co/i4hwglG2gu

2022-06-28 17:23:57 @mpshanahan The definition of AI as a field (and its proper area of study) was fluid, of course. But early pioneers made it clear in their papers that "artificial intelligence" was separate from the study of NNs. E.g., telling footnote from Minsky's 1961 paper "Steps Toward AI": https://t.co/FBXA82R43j

2022-06-28 16:33:33 @EthanJPerez @hardmaru @GaryMarcus @douwekiela @kchonyc Interesting, thanks!

2022-06-28 15:42:14 @deliprao @hardmaru @EthanJPerez @GaryMarcus What paper is this from?

2022-06-28 15:33:48 @hardmaru @EthanJPerez @GaryMarcus I wonder if prompt-sensitivity might be a good "Turing Test" for LLMs vs. humans, even very large LLMs.

2022-06-28 15:27:18 @hardmaru @EthanJPerez @GaryMarcus LLMs like GPT-3 are very "prompt-sensitive" -- i.e., their performance depends sensitively on details of prompt. Is there any evidence that scaling (or inverse scaling) makes LLMs less or more prompt sensitive? Also, any measures of "prompt sensitivity"?

2022-06-27 22:59:52 @RonJeffries I'm very happy to hear that you enjoyed it!

2022-06-27 19:15:19 RT @tdverstynen: @MelMitchell1 @Abebab This sort of thing always makes me think of this classic study.https://t.co/ytDVbIIXjx

2022-06-27 19:05:21 @Abebab I read the whole paper

2022-06-27 18:55:29 At that time, non-symbolic advocates started derisively referred to the original definition of AI as "GOFAI". Now the meaning of the term AI has gone 180 degrees --- nowadays it is almost synonymous with NNs.

2022-06-27 18:55:28 Interesting thread abt coverage of NNs in AI textbooks over the years. But important to point out that "Artificial Intelligence" was explicitly named and defined (by McCarthy. Minsky et al.) to refer to "heuristic programming" which by design did not include NNs. https://t.co/PXz3Gz1WgT

2022-06-24 14:44:30 Like many others horrified by SCOTUS decisions, I'm tempted to scream out my anguish here on Twitter. But better to do something concrete, like support organizations that promote voting, like @votesaveamerica, @fairfightaction, and @swingleft . I will be doing that today.

2022-06-24 02:23:56 RT @SimonsInstitute: CORRECTION: This event doesn't take place this week, but in a month!Adversarial Examples in Deep LearningTheoretica…

2022-06-24 00:17:48 RT @mmitchell_ai: Appreciate this piece. Covers a lot of ground wrt what's happened in AI over past few weeks &

2022-06-23 16:39:39 Looking forward to reading this book in its entirety! I read a few chapters in draft form, which were fascinating. https://t.co/aWPuTRRPtf

2022-06-22 20:34:11 @SimonsInstitute @anilananth @SebastienBubeck Friday July 22 (not *this* Friday )

2022-06-22 00:36:08 @mark_riedl That's exactly how these tools will be used by designers in the future. AI assistance, not AI autonomy.

2022-06-21 14:53:31 RT @bendee983: The "sentience" discussion is entertaining, but here are three important things large language models need as they enter rea…

2022-06-20 14:55:07 @LeilaKincaid @ZeeshanAleem it is simply using its weight values to compute probabilities over what word to output next. Once the conversation is over, It doesn't have any memory that carries over to the next conversation.

2022-06-20 14:54:49 @LeilaKincaid @ZeeshanAleem What I meant is that during its training, LaMDA memorizes (by setting weight values) much of its training data (human-created sentences). But when a person is "conversing" with it, it is no longer learning (or memorizing anything)....

2022-06-19 23:25:32 @mmitchell_ai Sad. Peer review isn't perfect but it's still essential for science. So is replication. Much of AI "research" these days eschews these key elements.

2022-06-19 19:48:52 One more note on the statement that "PaLM can perform reasoning". Indeed, it is better on some reasoning benchmarks than previous large language models, but the figure below from Google's paper on PaLM shows that it is still not very good at reasoning.... @StephenMarche https://t.co/SshhPW5o5w

2022-06-19 17:43:08 @TeemoDoScience That doesn't mean that the term is meaningless, or that anything could be called "sentient". I think the term has enough agreed-upon meaning that we can say programs like LaMDA are not sentient.

2022-06-19 17:41:48 @TeemoDoScience There is definitely no agreed-upon definition.

2022-06-19 16:50:01 I liked the skepticism of the Atlantic article vis a vis "sentience" but I wish the author had talked to AI experts outside of Google before reporting unqualified grandiose claims about PaLM. /end

2022-06-19 16:50:00 I worry about claims such as "PaLM can reason. Or, to be more precise—and precision very much matters here—PaLM can perform reason." Many researchers have shown limitations of "reasoning" in GPT-3, which is accessible to the public. /7

2022-06-19 16:49:59 that the tasks can be done reliably, with generality, (and maybe at level of humans)---not really accurate. Readers should be given a clearer picture of the abilities and limitations of such models, and the benchmarks they are tested on. /5

2022-06-19 16:49:58 This is an interesting article on Google's PaLM model. The author reports on his discussions with Google researchers. But I think he reports too credulously on what they say. /2

2022-06-19 16:49:57 I have a few comments on an article that appeared in today's Atlantic magazine, written by @StephenMarche: https://t.co/M9sMWdgHfD /1

2022-06-19 15:57:44 Fantastic memoir by his son, George Dyson.

2022-06-19 15:55:44 Freeman Dyson, as a young child during naptime https://t.co/Qh4mXqHazc(h/t @edyson) https://t.co/WxipR4sxUE

2022-06-17 16:38:13 RT @mmitchell_ai: New piece from @timnitGebru and me, with editing help from @katzish Thanks to WaPo for the opportunity! https://t.co/0p…

2022-06-17 16:34:47 If you can stand one more take on the sentient AI story.... from my chat with MSNBC writer @ZeeshanAleem. https://t.co/qjnyGMWMqI

2022-06-16 16:10:53 @CoralieLemke Hello, can you send your questions to me by email (mm@santafe.edu)?

2022-06-15 19:58:03 @etzioni @geekwire Congrats on building this important institute -- Can't wait to see what you do next!

2022-06-15 19:49:51 I enjoyed reading this thoughtful article by @StephenMarche:https://t.co/A25quByC5G

2022-06-15 02:17:46 @emilymbender Same. Let's talk about something (anything) else!

2022-06-14 23:22:14 RT @BTG_ie: "How does modern #AI work

2022-06-14 20:19:22 Great essay by @emilymbender . Looking forward to changing the topic from "sentience" to transparency. https://t.co/mcl8opvD8m

2022-06-13 21:54:54 @jamessseattle @Kdawg5000 @jeffsussna Consciousness is not well-defined, but it is not completely without definition. Many things can be definitively ruled out, if we want the word to mean anything at all.

2022-06-13 21:43:04 RT @philipcball: So is AI now sentient? Ah, if only there was a book to tell us. Oh, wait, look what drops in 10 days...https://t.co/q6KQA…

2022-06-13 20:19:53 @philipcball @WiringTheBrain Funny how every news organization in the world is now reporting on *consciousness*. Just in time for your new book. Could that be a coincidence?

2022-06-13 19:59:29 Me talking to CNN about (non) sentient AI https://t.co/vvR0T2rCwO

2022-06-13 19:43:50 @mark_riedl This bit from the article was useful for contextualizing: Lemoine was "ordained as a mystic Christian priest, and served in the Army before studying the occult...He concluded LaMDA was a person in his capacity as a priest, not a scientist...."

2022-06-13 18:56:24 This is a good take on the whole "sentient AI" thing. https://t.co/7g3FOpq87V

2022-06-12 21:34:29 @ChrSzegedy LMs don't have *experiences*, in any meaningful use of the term.

2022-06-12 21:30:26 Both still true. Hearing #2 is Monday at 10am Eastern time. https://t.co/lgaWCzHl3d

2022-06-11 15:06:16 "Lemoine worked ... to present evidence to Google that LaMDA was sentient...Google vice president Blaise Aguera y Arcas...looked into his claims and dismissed them." Dear Google, how do you "look into claims" that an AI system is sentient? Just asking for the entire field :-) https://t.co/Iv7hODNK63

2022-06-11 14:24:45 See https://t.co/0j3UFe26GA

2022-06-11 14:21:53 Such a strange article. It's been known for *forever* that humans are predisposed to anthropomorphize even with only the shallowest of signals (cf. ELIZA). Google engineers are human too, and not immune. https://t.co/dECTixuSmq

2022-06-11 00:53:53 Finally got it! https://t.co/TsRJ99VlOA

2022-06-11 00:51:38 Um.... https://t.co/SmXMAQNVHo

2022-06-10 23:21:26 Let's think step by step. https://t.co/H8IpXW25Mi

2022-06-10 23:06:00 Playing with reasoning questions in GPT-3. I found out it gets confused about dates in February, just like the rest of us. https://t.co/C0jOmqdSjD

2022-06-10 21:34:04 @GaryMarcus I got stuck at "its brain"

2022-06-10 20:49:21 @dileeplearning @neuroecology @GaryMarcus @slatestarcodex @ErnestSDavis @rogerkmoore "A concept is a package of analogies."-- Douglas Hofstadter

2022-06-10 20:48:10 @dileeplearning @neuroecology @GaryMarcus @slatestarcodex @ErnestSDavis @rogerkmoore "A concept is a competence or disposition for generating infinite conceptualizations of a category." -- Lawrence Barsalou

2022-06-10 17:06:43 @livcomp Two personal heroes!

2022-06-09 23:01:50 @charleswangb @C4COMPUTATION @sfiscience

2022-06-09 22:28:30 @C4COMPUTATION @sfiscience !!!!

2022-06-09 22:02:25 AI twitter is more exhausting than usual these days. Gonna watch the January 6 hearings to relax.

2022-06-09 12:48:23 @BeebsMemes The term has slipped, so now it is common to say things like "untrained model", "pretrained model" and similar.

2022-06-09 12:40:07 @mmitchell_ai The Economist magazine thinks "foundation models" = "Foundational AI".(https://t.co/IeeBlutOBS) https://t.co/e0uO0L77Wu

2022-06-08 23:38:55 Of course there is also the question of why we call every AI system a "model". It seems like the word "model" has lost its meaning. https://t.co/AF65JVRmex

2022-06-08 02:58:21 RT @WiringTheBrain: The evolution of meaning – from pragmatic couplings to semantic representations https://t.co/0cHmASfkzr https://t.co/SY…

2022-06-06 20:47:54 @zacharylipton @geoffreyhinton He finally figured out what Twitter is good for.

2022-06-06 17:12:46 @ch402 @FelixHill84 Oy, Twitter.

2022-06-06 17:11:29 @FelixHill84 @ch402 https://t.co/F8tDLGh4nw

2022-06-06 17:10:50 @TacoCohen @FelixHill84 @ch402

2022-06-06 17:10:24 @FelixHill84 @ch402 https://t.co/24PXhpMyrU

2022-06-06 17:09:08 @FelixHill84 @ch402 Sorry -- I guess there were more replies to Chris's thread than I saw.

2022-06-06 16:11:29 @FelixHill84 No misrepresentation, no polarization, just scientific discussion. @ch402

2022-06-06 16:11:15 @FelixHill84 Indeed, I wrote a whole book about analogies between fields in which simple processes generate complexity. It's important to be clear in each case what aspects of the process are actually responsible for generating the complexity we observe.

2022-06-06 16:07:03 @FelixHill84 pointing out the differences between evolution as a "simple" process to generate complexity and GD as such a process. No views were misrepresented.

2022-06-06 16:06:18 @FelixHill84 I feel subtweeted here. Actually, Chris Olah said explicitly "I'm trying to make a rough analogy to evolution because I'm interested in what it means to have a field centered around a simple process that generates complexity."I totally get that. I was just ....

2022-06-06 15:23:24 @IntuitMachine "Taking heat" = presenting an idea for discussion, and people discussing?

2022-06-06 02:56:41 @criticalneuro @blamlab I will ask him! Will DM you about hosting.

2022-06-05 21:52:18 @santoroAI @ch402 https://t.co/KbMlVw62Hj

2022-06-05 21:51:33 @santoroAI @ch402 https://t.co/kbFhpbkkAI

2022-06-05 21:49:22 @santoroAI @ch402 Here is some old work one of my students did on random weights in the HMAX model: https://t.co/R8JZVMM5gg

2022-06-05 21:49:13 @santoroAI @ch402 A few examples: Work on Reservoir Computing with random weightsLottery HypothesisThis paper by Uber AI Labs shows how surprisingly well random search does for NN weights: https://t.co/Ho5WfxAImK

2022-06-05 21:48:35 @santoroAI @ch402 Of course GD is typically better than random, but there is a lot of interesting work on how surprisingly powerful random weights / random search can be.

2022-06-05 21:01:13 Great interview with the amazing Vijay Balasubramanian, who is studying the foundations of space in the universe and the foundations of spatial cognition in neuroscience!(@criticalneuro @blamlab -- have him on the Learning Salon sometime?)https://t.co/XBPDhvCO70

2022-06-05 20:30:12 @ch402 I agree---the work on RL + self-play (going back to Arthur Samuel!) is the best ML/evolutionary analogy.Yes, probably topology is overrated. But GD might be also. Random weights and random search can have surprisingly good performance!

2022-06-05 19:17:53 Also, GD has human-created gradient (i.e., loss function). Evolution doesn't have that (or any knowledge of gradient at all).

2022-06-05 19:00:16 It seems the mind-boggling structure and behavior of DNNs is not from GD alone but also requires substantial input from humans in form of massive human-created data, NN topology, &

2022-06-05 18:57:08 I like this analogy, but seems incomplete. Evolution continually builds up complexity, by creating hierarchy, modularity, complex genotype/phenotype mappings via development, etc. Open-ended creation of complexity is the real power of evolution. What corresponds to that in GD? https://t.co/Do0yJ7VpGc

2022-06-02 13:03:45 @DylanRobertCope I'm so happy to hear you enjoyed it!

2022-06-01 16:26:35 RT @ComplexExplorer: Join us for Intro to Agent-Based Modeling beginning on June 7! Watch the welcome video below to learn more about the…

2022-06-01 00:18:04 @IgorBrigadir @GaryMarcus @elonmusk Yes, https://t.co/60hTI7KvRE

2022-05-31 15:55:08 @umeshpersad Unfortunately no, but here is a related talk I gave: https://t.co/rLagEjrabD

2022-05-30 20:28:47 Kurzweil &

2022-05-30 20:20:14 @GaryMarcus https://t.co/60hTI7sV06

2022-05-30 17:15:53 Okay, who's going to do DALL-E 2 version: "Queen Elizabeth projected onto Stonehenge" ? https://t.co/kQNURy9f2f

2022-05-30 17:09:29 Even creepier is the fact that Stonehenge has a Twitter account. https://t.co/AhM8d44hh6

2022-05-30 02:13:29 Fascinating read: @rodneyabrooks' adventures in taking Cruise self-driving taxis in SF: https://t.co/Fk7RZ6dNRv

2022-05-29 14:10:43 @TalKachman Wow, thanks for the kind words!

2022-05-27 17:29:37 Excellent thread. https://t.co/212lWUNGGq

2022-05-27 16:00:05 @raphaelmilliere Of course all VCs are male!

2022-05-26 02:55:11 RT @AMarch4OurLives: Together, we rose up 4 years ago. 1 million of us demanded change. We built a movement. We voted for new leaders.And…

2022-05-21 20:54:53 @dshoskes @UnitedAirlines See also: https://t.co/y46Xtv2pod

2022-05-21 20:51:14 Received today: New book on an essential topic for AI. Thanks Hector Levesque and @ronbrachman ! https://t.co/UAoF1QSqwX

2022-05-21 14:23:17 @mlittmancs Excellent choice by @nsf!!!

2022-05-20 08:11:00 CAFIAC FIX

2022-10-28 15:26:53 @TiernanRayTech @ylecun @GaryMarcus Lots of people care, of course -- many went into the field precisely to work on those big questions.

2022-10-28 15:02:25 @TiernanRayTech @ylecun @GaryMarcus https://t.co/G9xQyQdaOp

2022-10-27 00:25:24 @TheWill Hope you get better soon!

2022-10-26 17:34:00 Back in April, SFI hosted a fantastic workshop on "Embodied, Situated, and Grounded Intelligence". We wrote a report about it: https://t.co/mB6aHJUUH1Also, all talk videos are here: https://t.co/eB12lkU4Od https://t.co/mB6aHJUUH1

2022-10-26 14:08:48 @C4COMPUTATION @NadineSpychala @jlizier @RomainBrette @GaelVaroquaux @KordingLab Ditto!!

2022-10-26 14:05:25 New: A perspective piece by myself and David Krakauer about large language models and "understanding". https://t.co/G9xQyQdaOp

2022-10-26 14:03:01 @KordingLab @jlizier @RomainBrette @GaelVaroquaux @C4COMPUTATION I don't think AI is hitting a wall. My talk was about the many reasons I think people are over-optimistic about near-term human-level AI, and about some core aspects of human intelligence that are still missing in AI.

2022-10-25 15:57:41 @tyrell_turing @NSERC_CRSNG Great to see this -- congratulations!

2022-10-24 16:03:04 RT @bendee983: My review of "The Book of Minds" by @philipcball.Ball gives you a framework for looking past your instinctive tendency to d…

2022-10-24 15:50:55 @tdietterich @tinaeliassi @davecliff @arxiv Any advice for figuring out what the holdup is?

2022-10-24 15:06:38 @ankurhandos @arxiv Frustrating.

2022-10-24 15:01:45 @svscarpino @arxiv Thanks! I think I messed up by forgetting to choose a primary category (CS) and instead just choosing crosslists (AI/ML). Now they won't let me update it....sigh.

2022-10-24 14:53:19 @ankurhandos @arxiv Yup, I got the same reply.

2022-10-24 14:49:03 I submitted a new preprint to @arxiv 10 days ago --- it is still "on hold". No information except "Your submission was identified by arXiv staff or moderators as needing further attention." Anyone else have this experience? Can anything be done to find out what hold-up is?

2022-10-22 19:18:35 I'll also be doing a book signing after the panel -- come by and talk about complexity and/or AI!https://t.co/W0duysckWB https://t.co/ON0sZrp9pj

2022-10-22 19:17:09 If you're in Santa Fe and interested in intelligence (biological or artificial), come by SITE Santa Fe to hear our panel discussion: https://t.co/8cxjM4rnsRIf not in Santa Fe, follow on the YouTube livestream: https://t.co/aMfdexZRBehttps://t.co/8cxjM4rnsR

2022-10-22 18:08:06 Very sad. Chuck was a regular visitor to @sfiscience for many years and I was fortunate to get to know him. He was brilliant and interested in everything. I learned a lot from him and will miss him very much. https://t.co/RkfvzudNqe

2022-10-21 18:32:45 @cigitalgem Indeed!

2022-10-28 15:26:53 @TiernanRayTech @ylecun @GaryMarcus Lots of people care, of course -- many went into the field precisely to work on those big questions.

2022-10-28 15:02:25 @TiernanRayTech @ylecun @GaryMarcus https://t.co/G9xQyQdaOp

2022-10-27 00:25:24 @TheWill Hope you get better soon!

2022-10-26 17:34:00 Back in April, SFI hosted a fantastic workshop on "Embodied, Situated, and Grounded Intelligence". We wrote a report about it: https://t.co/mB6aHJUUH1Also, all talk videos are here: https://t.co/eB12lkU4Od https://t.co/mB6aHJUUH1

2022-10-26 14:08:48 @C4COMPUTATION @NadineSpychala @jlizier @RomainBrette @GaelVaroquaux @KordingLab Ditto!!

2022-10-26 14:05:25 New: A perspective piece by myself and David Krakauer about large language models and "understanding". https://t.co/G9xQyQdaOp

2022-10-26 14:03:01 @KordingLab @jlizier @RomainBrette @GaelVaroquaux @C4COMPUTATION I don't think AI is hitting a wall. My talk was about the many reasons I think people are over-optimistic about near-term human-level AI, and about some core aspects of human intelligence that are still missing in AI.

2022-10-25 15:57:41 @tyrell_turing @NSERC_CRSNG Great to see this -- congratulations!

2022-10-24 16:03:04 RT @bendee983: My review of "The Book of Minds" by @philipcball.Ball gives you a framework for looking past your instinctive tendency to d…

2022-10-24 15:50:55 @tdietterich @tinaeliassi @davecliff @arxiv Any advice for figuring out what the holdup is?

2022-10-24 15:06:38 @ankurhandos @arxiv Frustrating.

2022-10-24 15:01:45 @svscarpino @arxiv Thanks! I think I messed up by forgetting to choose a primary category (CS) and instead just choosing crosslists (AI/ML). Now they won't let me update it....sigh.

2022-10-24 14:53:19 @ankurhandos @arxiv Yup, I got the same reply.

2022-10-24 14:49:03 I submitted a new preprint to @arxiv 10 days ago --- it is still "on hold". No information except "Your submission was identified by arXiv staff or moderators as needing further attention." Anyone else have this experience? Can anything be done to find out what hold-up is?

2022-10-22 19:18:35 I'll also be doing a book signing after the panel -- come by and talk about complexity and/or AI!https://t.co/W0duysckWB https://t.co/ON0sZrp9pj

2022-10-22 19:17:09 If you're in Santa Fe and interested in intelligence (biological or artificial), come by SITE Santa Fe to hear our panel discussion: https://t.co/8cxjM4rnsRIf not in Santa Fe, follow on the YouTube livestream: https://t.co/aMfdexZRBehttps://t.co/8cxjM4rnsR

2022-10-22 18:08:06 Very sad. Chuck was a regular visitor to @sfiscience for many years and I was fortunate to get to know him. He was brilliant and interested in everything. I learned a lot from him and will miss him very much. https://t.co/RkfvzudNqe

2022-10-21 18:32:45 @cigitalgem Indeed!

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-24 16:57:24 I'm excited to participate in this workshop! https://t.co/93lznCMh5P

2022-11-23 17:20:54 RT @davidbkinney: The next run of @faha_sfi, a free, @NEH_ODH-funded online course in core digital humanities techniques, begins January 17…

2022-11-23 15:40:04 Undergrads!!! This is a wonderful research internship opportunity. https://t.co/zvsbldhbrn

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-25 22:26:54 RT @arthur_spirling: “long-termism” means commenting your code.

2022-11-24 16:57:24 I'm excited to participate in this workshop! https://t.co/93lznCMh5P

2022-11-23 17:20:54 RT @davidbkinney: The next run of @faha_sfi, a free, @NEH_ODH-funded online course in core digital humanities techniques, begins January 17…

2022-11-23 15:40:04 Undergrads!!! This is a wonderful research internship opportunity. https://t.co/zvsbldhbrn

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-28 18:39:32 Kudos to @Northeastern, for choosing Sam to lead this initiative! https://t.co/vYRdWWoCMX

2022-11-28 18:36:09 @svscarpino @Northeastern @usamaf @Experiential_AI Congrats! Very exciting initiative!

2022-11-28 15:15:22 @BitchbotDE I'm not sure, but you can email education@santafe.edu with questions.

2022-11-28 00:47:29 @sfiscience @shreyunsh Email education@santafe.edu with any questions.

2022-11-27 19:53:08 PhD Students from any field --- are you interested in understanding the nature of intelligence across disciplines? Apply to join us this summer in Cambridge UK for the SFI's two-week international summer school on *Intelligence &

2022-11-25 22:26:54 RT @arthur_spirling: “long-termism” means commenting your code.

2022-11-24 16:57:24 I'm excited to participate in this workshop! https://t.co/93lznCMh5P

2022-11-23 17:20:54 RT @davidbkinney: The next run of @faha_sfi, a free, @NEH_ODH-funded online course in core digital humanities techniques, begins January 17…

2022-11-23 15:40:04 Undergrads!!! This is a wonderful research internship opportunity. https://t.co/zvsbldhbrn

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-11-29 02:55:42 @hardmaru Could happen, if this gets funded: https://t.co/I4gnIUv2yz

2022-11-28 18:39:32 Kudos to @Northeastern, for choosing Sam to lead this initiative! https://t.co/vYRdWWoCMX

2022-11-28 18:36:09 @svscarpino @Northeastern @usamaf @Experiential_AI Congrats! Very exciting initiative!

2022-11-28 15:15:22 @BitchbotDE I'm not sure, but you can email education@santafe.edu with questions.

2022-11-28 00:47:29 @sfiscience @shreyunsh Email education@santafe.edu with any questions.

2022-11-27 19:53:08 PhD Students from any field --- are you interested in understanding the nature of intelligence across disciplines? Apply to join us this summer in Cambridge UK for the SFI's two-week international summer school on *Intelligence &

2022-11-25 22:26:54 RT @arthur_spirling: “long-termism” means commenting your code.

2022-11-24 16:57:24 I'm excited to participate in this workshop! https://t.co/93lznCMh5P

2022-11-23 17:20:54 RT @davidbkinney: The next run of @faha_sfi, a free, @NEH_ODH-funded online course in core digital humanities techniques, begins January 17…

2022-11-23 15:40:04 Undergrads!!! This is a wonderful research internship opportunity. https://t.co/zvsbldhbrn

2022-11-16 23:40:15 @yoavgo Darn!

2022-11-16 23:37:06 Science FTW https://t.co/hFxiU7NvHK

2022-11-16 21:46:25 @amahabal Fortunately comedy is now legal on Twitter.

2022-11-16 21:34:37 @amahabal @GaryMarcus @ylecun @MetaAI @paperswithcode TBH that doesn't sound too far off from some human-written "consciousness studies" papers. No coincidence, I guess...

2022-11-16 20:58:54 Feeling like my job as a scientist is still secure. https://t.co/3ezAkgDcGQ

2022-12-07 17:15:12 @rgblong Did ChatGPT write that? (JK)

2022-12-07 16:50:14 @mkapor I'm still betting on you winning, given the sophisticated judges and requirement for two-hour interviews.

2022-12-07 16:23:28 Note that judges will not be random people but will be experts picked by the committee. Also, the interviews will be *two hours* each! By 2029 will there be an AI system that can fool human experts over the course of two hours? I'm not saying impossible, just seems very hard.

2022-12-07 16:08:57 What do you think?

2022-12-07 16:08:56 Given furor over ChatGPT, time to remind people about the #longbet between @mkapor and Kurzweil: “By 2029 no computer or "machine intelligence"will have passed Turing Test.” Before voting in poll (next tweet), see strict terms for the debate: https://t.co/7dXzwCa82O #chatgpt

2022-12-07 17:15:12 @rgblong Did ChatGPT write that? (JK)

2022-12-07 16:50:14 @mkapor I'm still betting on you winning, given the sophisticated judges and requirement for two-hour interviews.

2022-12-07 16:23:28 Note that judges will not be random people but will be experts picked by the committee. Also, the interviews will be *two hours* each! By 2029 will there be an AI system that can fool human experts over the course of two hours? I'm not saying impossible, just seems very hard.

2022-12-07 16:08:57 What do you think?

2022-12-07 16:08:56 Given furor over ChatGPT, time to remind people about the #longbet between @mkapor and Kurzweil: “By 2029 no computer or "machine intelligence"will have passed Turing Test.” Before voting in poll (next tweet), see strict terms for the debate: https://t.co/7dXzwCa82O #chatgpt

2022-12-08 22:19:51 @PWGTennant @random_walker Maybe I forgot to tweet that day.

2022-12-08 19:46:57 @random_walker At least it doesn't think you're dead https://t.co/GYse3nLOZg

2022-12-08 19:45:21 .@perplexity_ai, you are not making my day a happy one https://t.co/WnLdqo0MUO https://t.co/XjJaGNhIPd

2022-12-08 16:55:22 RT @daveyalba: New from me: ChatGPT is awe-inspiring tech. It wrote a wonderful poem about my cats! But it's plagued by the same issues of…

2022-12-08 16:09:46 Pretty close vote. Only 7 years to wait for the final result. https://t.co/nL6KCQLX61

2022-12-07 17:15:12 @rgblong Did ChatGPT write that? (JK)

2022-12-07 16:50:14 @mkapor I'm still betting on you winning, given the sophisticated judges and requirement for two-hour interviews.

2022-12-07 16:23:28 Note that judges will not be random people but will be experts picked by the committee. Also, the interviews will be *two hours* each! By 2029 will there be an AI system that can fool human experts over the course of two hours? I'm not saying impossible, just seems very hard.

2022-12-07 16:08:57 What do you think?

2022-12-07 16:08:56 Given furor over ChatGPT, time to remind people about the #longbet between @mkapor and Kurzweil: “By 2029 no computer or "machine intelligence"will have passed Turing Test.” Before voting in poll (next tweet), see strict terms for the debate: https://t.co/7dXzwCa82O #chatgpt

2022-12-09 00:56:08 @fchollet Totally a blast! https://t.co/AJ69b3VN48

2022-12-09 00:17:02 Breaking news for language models used in search engines: Different people have the same name! https://t.co/xVxFfCEABr

2022-12-09 00:12:57 @AravSrinivas @perplexity_ai Disappointing. I don't think this way is going to be very effective.

2022-12-08 23:59:14 @AravSrinivas @perplexity_ai Thanks for looking into it. How are you going to get your system to return more reliable information in the future?

2022-12-08 22:19:51 @PWGTennant @random_walker Maybe I forgot to tweet that day.

2022-12-08 19:46:57 @random_walker At least it doesn't think you're dead https://t.co/GYse3nLOZg

2022-12-08 19:45:21 .@perplexity_ai, you are not making my day a happy one https://t.co/WnLdqo0MUO https://t.co/XjJaGNhIPd

2022-12-08 16:55:22 RT @daveyalba: New from me: ChatGPT is awe-inspiring tech. It wrote a wonderful poem about my cats! But it's plagued by the same issues of…

2022-12-08 16:09:46 Pretty close vote. Only 7 years to wait for the final result. https://t.co/nL6KCQLX61

2022-12-07 17:15:12 @rgblong Did ChatGPT write that? (JK)

2022-12-07 16:50:14 @mkapor I'm still betting on you winning, given the sophisticated judges and requirement for two-hour interviews.

2022-12-07 16:23:28 Note that judges will not be random people but will be experts picked by the committee. Also, the interviews will be *two hours* each! By 2029 will there be an AI system that can fool human experts over the course of two hours? I'm not saying impossible, just seems very hard.

2022-12-07 16:08:57 What do you think?

2022-12-07 16:08:56 Given furor over ChatGPT, time to remind people about the #longbet between @mkapor and Kurzweil: “By 2029 no computer or "machine intelligence"will have passed Turing Test.” Before voting in poll (next tweet), see strict terms for the debate: https://t.co/7dXzwCa82O #chatgpt

2022-12-09 00:56:08 @fchollet Totally a blast! https://t.co/AJ69b3VN48

2022-12-09 00:17:02 Breaking news for language models used in search engines: Different people have the same name! https://t.co/xVxFfCEABr

2022-12-09 00:12:57 @AravSrinivas @perplexity_ai Disappointing. I don't think this way is going to be very effective.

2022-12-08 23:59:14 @AravSrinivas @perplexity_ai Thanks for looking into it. How are you going to get your system to return more reliable information in the future?

2022-12-08 22:19:51 @PWGTennant @random_walker Maybe I forgot to tweet that day.

2022-12-08 19:46:57 @random_walker At least it doesn't think you're dead https://t.co/GYse3nLOZg

2022-12-08 19:45:21 .@perplexity_ai, you are not making my day a happy one https://t.co/WnLdqo0MUO https://t.co/XjJaGNhIPd

2022-12-08 16:55:22 RT @daveyalba: New from me: ChatGPT is awe-inspiring tech. It wrote a wonderful poem about my cats! But it's plagued by the same issues of…

2022-12-08 16:09:46 Pretty close vote. Only 7 years to wait for the final result. https://t.co/nL6KCQLX61