Claims

2,653 claims by 33 public figures and organisations, checked against the public record. Each has its quote, source, verdict and evidence. Read them without a key; browse by person.

28,548 claims, sorted by newest said. Choose a claim to read it in full.
VerdictPersonKindSaidClaimTopic
Mostly true Nassim Taleb Factual 30 Mar 2024 The city of Jerusalem is not mentioned even once in the Pentateuch. Nassim Taleb · Factual · 30 Mar 2024 Philosophy, ethics and society
Not checkable Nassim Taleb Not checkable 30 Mar 2024 Others pushed it under the rug. Nassim Taleb · Not checkable · 30 Mar 2024
Unresolvable Nassim Taleb Factual 30 Mar 2024 Daniel Kahneman was the only psychologist who acknowledged the problem of fat tails. Nassim Taleb · Factual · 30 Mar 2024 Science and research methods
Not checkable Nassim Taleb Not checkable 30 Mar 2024 (The common academic belief that tail probabilities are "overpriced" makes no sense under F Tails, and he knew it.) Nassim Taleb · Not checkable · 30 Mar 2024
Not researched Yann LeCun Factual 30 Mar 2024 Anti-vaccine propaganda causes people to die needlessly. Yann LeCun · Factual · 30 Mar 2024 Philosophy, ethics and society
Unresolvable Ray Kurzweil Factual 29 Mar 2024 Every type of automation has eventually left people far better off. Ray Kurzweil · Factual · 29 Mar 2024 AI, the economy and work
Not researched Ray Kurzweil Factual 29 Mar 2024 Biological intelligence in the brain cannot currently be recovered if lost. Ray Kurzweil · Factual · 29 Mar 2024 Biosecurity
Mostly true Ray Kurzweil Factual 29 Mar 2024 About 300 years ago, life expectancy was around 28. Ray Kurzweil · Factual · 29 Mar 2024 Philosophy, ethics and society
Not checkable Eliezer Yudkowsky Not checkable 29 Mar 2024 apparently this battle is not quite over yet Eliezer Yudkowsky · Not checkable · 29 Mar 2024
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 Ultimately there will be more and more ways of extending human longevity. Ray Kurzweil · Prediction · 29 Mar 2024 Science and research methods
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 AI's added intelligence comes from people and will make humans smarter. Ray Kurzweil · Prediction · 29 Mar 2024 AI, the economy and work
Mostly true Ray Kurzweil Factual 29 Mar 2024 Eighty years ago, 100 million people died in Europe and Asia in World War II. Ray Kurzweil · Factual · 29 Mar 2024 Other global risks
Not researched Ray Kurzweil Factual 29 Mar 2024 Machines replacing jobs has not so far widened inequality or eliminated jobs overall. Ray Kurzweil · Factual · 29 Mar 2024 AI, the economy and work
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 Large language models will enable people to earn enough to live and spend time on things they enjoy. Ray Kurzweil · Prediction · 29 Mar 2024 AI, the economy and work
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 Humans will merge with computation and become so intelligent that it cannot be described now. Ray Kurzweil · Prediction · 29 Mar 2024 AI timelines and forecasting
Unresolvable Eliezer Yudkowsky Factual 29 Mar 2024 Current AIs are roughly as generally intelligent, as dangerous and as controllable as human 4-year-olds. Eliezer Yudkowsky · Factual · 29 Mar 2024 AI capabilities
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 The part of human intelligence that comes from machines will be backed up and recoverable. Ray Kurzweil · Prediction · 29 Mar 2024 Technology
Not checkable Ray Kurzweil Not checkable 29 Mar 2024 Now everybody’s heard of it, and people are very nervous about it. Ray Kurzweil · Not checkable · 29 Mar 2024
Mostly true Ray Kurzweil Factual 29 Mar 2024 AI progress is advancing very quickly, at an exponential rate. Ray Kurzweil · Factual · 29 Mar 2024 AI capabilities
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 Humans will be able to live indefinitely without aging, continuing to look as they do now. Ray Kurzweil · Prediction · 29 Mar 2024 Science and research methods
Not yet due Ray Kurzweil Prediction 29 Mar 2024 Within five years AI coding ability will be as good as humans'; currently it is not quite as good. Ray Kurzweil · Prediction · 29 Mar 2024 AI capabilities
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 Eventually people will be far better off than today despite job disruption from AI. Ray Kurzweil · Prediction · 29 Mar 2024 AI, the economy and work
Unresolvable Ray Kurzweil Factual 29 Mar 2024 No human being has expert knowledge across every topic; Einstein knew physics but not philosophy. Ray Kurzweil · Factual · 29 Mar 2024 AI capabilities
Not researched Ray Kurzweil Factual 29 Mar 2024 In his 1999 book The Age of Spiritual Machines, Kurzweil predicted the Turing test would be passed within 30 years, by 2029. Ray Kurzweil · Factual · 29 Mar 2024 AI timelines and forecasting
Mostly true Yann LeCun Factual 29 Mar 2024 Yann LeCun gave a lecture hosted by the Center of Mathematical Sciences and Applications at Harvard on 2024-03-28 to a full house. Yann LeCun · Factual · 29 Mar 2024 Their own work or organisation
Not researched Ray Kurzweil Factual 29 Mar 2024 Kurzweil has worked in artificial intelligence for 61 years, a record longevity in the field. Ray Kurzweil · Factual · 29 Mar 2024 Their own work or organisation
Mostly true Ray Kurzweil Factual 29 Mar 2024 Kurzweil's early computer music program produced melodies recognizable as in the style of Mozart or Chopin, though not quite as good. Ray Kurzweil · Factual · 29 Mar 2024 AI capabilities
Unresolvable Ray Kurzweil Factual 29 Mar 2024 Experts at a Stanford conference after 1999 felt passing the Turing test would take 100 years. Ray Kurzweil · Factual · 29 Mar 2024 AI timelines and forecasting
Not checkable Ray Kurzweil Not checkable 29 Mar 2024 It’s getting adopted very quickly, but it makes us smarter. Ray Kurzweil · Not checkable · 29 Mar 2024
Not checkable Yann LeCun Not checkable 29 Mar 2024 scientific and technical breakthroughs are unpredictable by management, and progress often comes from folks with unconventional… Yann LeCun · Not checkable · 29 Mar 2024
Not yet due Ray Kurzweil Prediction 29 Mar 2024 In five years one large language model will know everything any human being knows. Ray Kurzweil · Prediction · 29 Mar 2024 AI capabilities
Unresolvable Ray Kurzweil Factual 29 Mar 2024 Kurzweil said atomic weapons were invented when he was a small child, and that two nuclear weapons were exploded in anger within one week. Ray Kurzweil · Factual · 29 Mar 2024 Other global risks
Not checkable Ray Kurzweil Not checkable 29 Mar 2024 But millions of people are using it, and it is making us smarter. Ray Kurzweil · Not checkable · 29 Mar 2024
Not checkable Ray Kurzweil Not checkable 29 Mar 2024 I think religion is something we’ve made up because we didn’t have any alternative. Ray Kurzweil · Not checkable · 29 Mar 2024
Not checkable Ray Kurzweil Not checkable 29 Mar 2024 We will understand it when we get there, but right now we can’t really predict it. Ray Kurzweil · Not checkable · 29 Mar 2024
Not checkable Nassim Taleb Not checkable 29 Mar 2024 For an essay, read the text and skip the footnotes. Nassim Taleb · Not checkable · 29 Mar 2024
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 Human longevity will increase rather than decrease, though not guaranteed. Ray Kurzweil · Prediction · 29 Mar 2024 Science and research methods
Not yet due Ray Kurzweil Prediction 29 Mar 2024 Kurzweil said he thinks humanity will reach longevity escape velocity in five years, by 2029. Ray Kurzweil · Prediction · 29 Mar 2024 Science and research methods
Not yet due Ray Kurzweil Prediction 29 Mar 2024 Within a few years AI will be able to emulate the intelligence any human being can have. Ray Kurzweil · Prediction · 29 Mar 2024 AI timelines and forecasting
Not checkable Ray Kurzweil Not checkable 29 Mar 2024 I believe in innovation. Ray Kurzweil · Not checkable · 29 Mar 2024
Not checkable Ray Kurzweil Not checkable 29 Mar 2024 If we make ourselves even smarter, that’ll be even more beneficial, and that’s what this is all about. Ray Kurzweil · Not checkable · 29 Mar 2024
Not researched Ray Kurzweil Factual 29 Mar 2024 There are more jobs today than at the time of the Luddites 200 years ago, despite machines doing jobs. Ray Kurzweil · Factual · 29 Mar 2024 AI, the economy and work
Not yet due Ray Kurzweil Prediction 29 Mar 2024 By 2029 scientific progress will give people back a full year of life expectancy for each year lived. Ray Kurzweil · Prediction · 29 Mar 2024 Science and research methods
Mostly false Ray Kurzweil Factual 29 Mar 2024 Computation/AI capability has increased 20 quadrillion-fold over 80 years. Ray Kurzweil · Factual · 29 Mar 2024 AI compute, hardware and data
Resolved true Ray Kurzweil Factual 29 Mar 2024 In the roughly 80 years the speaker refers to, the event he calls "it" has not happened, although atomic weapons are held by people he describes as "not the nicest people in the world". Ray Kurzweil · Factual · 29 Mar 2024 Other global risks
Not yet due Ray Kurzweil Prediction 29 Mar 2024 After 2029, people will gain back more than a year of longevity per year, effectively going backwards in time regarding longevity. Ray Kurzweil · Prediction · 29 Mar 2024 Science and research methods
Not checkable Yann LeCun Not checkable 29 Mar 2024 It's astonishing how certain types of managers will prefer less-talented-but-more-controlable scientists and engineers over more brilliant… Yann LeCun · Not checkable · 29 Mar 2024
Unresolvable Eliezer Yudkowsky Prediction 29 Mar 2024 If Yudkowsky's policy proposals are not adopted and a distorted version is used to justify government takeover, a government rather than a private company might kill everyone on Earth. Eliezer Yudkowsky · Prediction · 29 Mar 2024 AI policy and governance
Not yet due Ray Kurzweil Prediction 29 Mar 2024 Kurzweil predicts that in five years (by 2029) one large language model will know everything that any human being knows. Ray Kurzweil · Prediction · 29 Mar 2024 AI timelines and forecasting
Unresolvable Ray Kurzweil Factual 29 Mar 2024 People only want to end their lives when in unbearable pain. Ray Kurzweil · Factual · 29 Mar 2024 Philosophy, ethics and society
Mostly false Ray Kurzweil Factual 29 Mar 2024 Human-caused destruction is declining as humans have grown more intelligent; current wars are less deadly than past ones. Ray Kurzweil · Factual · 29 Mar 2024 Philosophy, ethics and society
Mostly true Ray Kurzweil Factual 29 Mar 2024 People today earn about 10 times as much in constant dollars as 100 years ago. Ray Kurzweil · Factual · 29 Mar 2024 Economics and markets
Unresolvable Ray Kurzweil Factual 29 Mar 2024 Life expectancy was 20 a thousand years ago, 35 in 1800, and is now nearly 80. Ray Kurzweil · Factual · 29 Mar 2024 Philosophy, ethics and society
Unresolvable Ray Kurzweil Prediction 29 Mar 2024 Despite job losses to automation, people will continue to be better off. Ray Kurzweil · Prediction · 29 Mar 2024 AI, the economy and work
Not checkable Nassim Taleb Not checkable 28 Mar 2024 It's all about him. It's not about the victims. Nassim Taleb · Not checkable · 28 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 28 Mar 2024 Interesting bounty format. Eliezer Yudkowsky · Not checkable · 28 Mar 2024
Unresolvable Yann LeCun Factual 28 Mar 2024 The linked document is Meta's official response to the NTIA request for comments about AI, whose recommendation number 1 stresses the importance of open foundation models. Yann LeCun · Factual · 28 Mar 2024 AI policy and governance
Not checkable Nassim Taleb Not checkable 28 Mar 2024 Most people knew him as a thinker. I knew him as a person. Nassim Taleb · Not checkable · 28 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 27 Mar 2024 The current setup is simply not competent to do that safely or validly. Eliezer Yudkowsky · Not checkable · 27 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 27 Mar 2024 I think that given the vast distance between the modern-day bureaucracy trying to handle this, and the ways of thinking required to… Eliezer Yudkowsky · Not checkable · 27 Mar 2024
Unresolvable Eliezer Yudkowsky Prediction 27 Mar 2024 Civilization must back off from building artificial superintelligence or it will die; there is no other outcome. Eliezer Yudkowsky · Prediction · 27 Mar 2024 AI policy and governance
Not checkable Yann LeCun Not checkable 27 Mar 2024 Glad to confirm that Vitalik is not an AI doomer. Yann LeCun · Not checkable · 27 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 27 Mar 2024 There are only people who know enough of alignment to say, "If we don't back off, we will die." Eliezer Yudkowsky · Not checkable · 27 Mar 2024
Unresolvable Eliezer Yudkowsky Prediction 27 Mar 2024 No institution on Earth can solve AI alignment in time, and extremely likely none can solve it at all. Eliezer Yudkowsky · Prediction · 27 Mar 2024 AI safety and alignment
Not checkable Eliezer Yudkowsky Not checkable 27 Mar 2024 Because Science the way we know it does need empiricism. Eliezer Yudkowsky · Not checkable · 27 Mar 2024
Unresolvable Eliezer Yudkowsky Prediction 27 Mar 2024 If today's civilization is allowed to build superintelligence, everyone will die. Eliezer Yudkowsky · Prediction · 27 Mar 2024 AI safety and alignment
Not checkable Nassim Taleb Not checkable 27 Mar 2024 Converging over time is healthy, diverging is not. Nassim Taleb · Not checkable · 27 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 27 Mar 2024 Humans augmented far enough to stop being idiots, smart enough that they never put faith in a step that will not hold, as do not exist in… Eliezer Yudkowsky · Not checkable · 27 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 27 Mar 2024 Only by then, of course, everyone is dead. Eliezer Yudkowsky · Not checkable · 27 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 27 Mar 2024 A gap like that cannot be crossed by appointing a prestigious committee of the same scientists who told you 10 years ago that there was no… Eliezer Yudkowsky · Not checkable · 27 Mar 2024
Not checkable Nassim Taleb Not checkable 26 Mar 2024 Golfers are only 18%. Nassim Taleb · Not checkable · 26 Mar 2024
Not researched Yann LeCun Factual 26 Mar 2024 Crypto money is secretly funding AI doomerism and lobbying for AI regulations, effectively working against open source AI platforms. Yann LeCun · Factual · 26 Mar 2024 AI policy and governance
Not checkable Yann LeCun Not checkable 26 Mar 2024 Canine planning in the presence of uncertainty. Yann LeCun · Not checkable · 26 Mar 2024
Unresolvable Eliezer Yudkowsky Prediction 26 Mar 2024 If a government's privileged-access AI turns into an artificial superintelligence, humanity will be killed. Eliezer Yudkowsky · Prediction · 26 Mar 2024 AI safety and alignment
Not checkable Eliezer Yudkowsky Not checkable 26 Mar 2024 What we want is a universal ban on anything that might turn into a superintelligence. Eliezer Yudkowsky · Not checkable · 26 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 25 Mar 2024 The reasoning is locally valid as a matter of economics, but you need a rather different "regulation" to prevent ASIs from just illegally… Eliezer Yudkowsky · Not checkable · 25 Mar 2024
Not checkable Nassim Taleb Not checkable 25 Mar 2024 If all Bedouins have 2 eyes, not all those who have 2 eyes are Bedouin. Nassim Taleb · Not checkable · 25 Mar 2024
Not checkable Nassim Taleb Not checkable 25 Mar 2024 Trump is playing it in a clever way, avoiding to allienate both Muslim and Jewish voters upset with Genocied Joe. Nassim Taleb · Not checkable · 25 Mar 2024
Mostly true Eliezer Yudkowsky Factual 25 Mar 2024 Yudkowsky was unpleasantly surprised by how well transformers scaled, having believed the human brain is not homogeneous. Eliezer Yudkowsky · Factual · 25 Mar 2024 AI capabilities
Not checkable Nassim Taleb Not checkable 24 Mar 2024 If all Frenchmen have 2 eyes, having 2 eyes does not imply being French. Nassim Taleb · Not checkable · 24 Mar 2024
Not checkable Nassim Taleb Not checkable 24 Mar 2024 Classical generator behind Fooled by Randomness: affirming the consequent. Nassim Taleb · Not checkable · 24 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 24 Mar 2024 "Doomers" are the anti-doom faction Eliezer Yudkowsky · Not checkable · 24 Mar 2024
Not checkable Nassim Taleb Not checkable 24 Mar 2024 Noticed a certain number of propaganda accounts that don't follow you but jump on your thread when you mention "Palestinian"; the hasbara… Nassim Taleb · Not checkable · 24 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 23 Mar 2024 Please just sign up for cryonics instead of killing everyone on Earth in a mad quest for immortality. Eliezer Yudkowsky · Not checkable · 23 Mar 2024
Not checkable Yann LeCun Not checkable 23 Mar 2024 AI saves lives today, and will save even more tomorrow. Yann LeCun · Not checkable · 23 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 23 Mar 2024 I think a major problem with explaining what superintelligences can (very probably) do is that people think that literally this here video… Eliezer Yudkowsky · Not checkable · 23 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 23 Mar 2024 This wasn't even an original argument for the world needing a cryonics option, because nobody who founded Alcor imagined that ANYONE WOULD… Eliezer Yudkowsky · Not checkable · 23 Mar 2024
Not checkable Nassim Taleb Not checkable 23 Mar 2024 Why AI bothers some people: most humans love the soothing rythms of the routine, prefer to expend physical rather than intellectual energy. Nassim Taleb · Not checkable · 23 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 23 Mar 2024 It being hard to stop them, does not mean that an omnicide machine won't kill everyone. Eliezer Yudkowsky · Not checkable · 23 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 23 Mar 2024 I wonder if it's the case that there's been negligible overlap between chess Masters (and above), and people who ask, "But how could an… Eliezer Yudkowsky · Not checkable · 23 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 23 Mar 2024 I don't like it when things I say hurt people. Eliezer Yudkowsky · Not checkable · 23 Mar 2024
Not checkable Nassim Taleb Not checkable 23 Mar 2024 Figured out chatting with Syrian workers why some Bedouins can cooperate with the Israelis. Nassim Taleb · Not checkable · 23 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 23 Mar 2024 Regardless of whether it's true, it's not a valid counterargument. Eliezer Yudkowsky · Not checkable · 23 Mar 2024
Not checkable Nassim Taleb Not checkable 23 Mar 2024 By definition, being nomads, they have no attachment to land; their allegiance is to the tribe, unlike the dispossessed Palestinians who… Nassim Taleb · Not checkable · 23 Mar 2024
Not checkable Nassim Taleb Not checkable 22 Mar 2024 Predictably, virtue positioning is steadily shifting from pro-Israel to pro-Palestinian. Nassim Taleb · Not checkable · 22 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 22 Mar 2024 Would you buy a bridge from an engineering firm that dismissed all warnings about bridges falling down as the work of "doomers"? Eliezer Yudkowsky · Not checkable · 22 Mar 2024
Not checkable Nassim Taleb Not checkable 22 Mar 2024 Just reduce the number of things you read that you wouldn't reread a decade (or more) from now. Nassim Taleb · Not checkable · 22 Mar 2024
Not checkable Eliezer Yudkowsky Not checkable 22 Mar 2024 At best, "safety testing" asks OpenAI to stop, and then Altman else Microsoft else Google else Meta destroys Earth anyways. Eliezer Yudkowsky · Not checkable · 22 Mar 2024
Not checkable Yann LeCun Not checkable 22 Mar 2024 The civilized world uses 00:00 to 23:59 Yann LeCun · Not checkable · 22 Mar 2024
Not checkable Nassim Taleb Not checkable 22 Mar 2024 Not by actively studying boring or nonboring material. Nassim Taleb · Not checkable · 22 Mar 2024
Back to the list

Factual AI capabilities

“I was a passionate noticer of the human brain not being homogenous, and was therefore unpleasantly surprised by transformers scaling as well as they did.”

Eliezer Yudkowsky, writing on X 25 March 2024 Post by @ESYudkowsky on X, 2024-03-25 (X)

Yudkowsky was unpleasantly surprised by how well transformers scaled, having believed the human brain is not homogeneous.

Mostly true True in substance, wrong in a detail.

The second judge would not resolve this claim; the verdict shown is the first judge's.

What would count as true or false

This is a claim about Yudkowsky's own past beliefs and reaction. True: his earlier writings show he stressed that the brain is not homogeneous (e.g. it has specialised, complex architecture) and, before or around the scaling of transformers, expected that simple uniform architectures would not scale as well as they did; and his record from that period (or a credible account of it) is consistent with him being surprised. Mostly true: the earlier writings show the non-homogeneity view, but there is little from that period to show the surprise beyond his 2024 statement, and nothing contradicts it. Mixed: the record is ambiguous, with some earlier statements anticipating scaling success and others not. Mostly false: his earlier statements mostly predicted that large homogeneous architectures would scale well, so the claimed surprise misdescribes his past views, although he did stress the brain's non-homogeneity. False: contemporaneous record shows he expected transformer scaling to succeed roughly as it did, or that he did not hold the stated view about the brain.

Evidence

  1. Dialogue with Eliezer Yudkowsky on Neural Networks — LessWrong LessWrong (linkpost from aashishreddy.substack.com) Own record
    “Still no general intelligence. But, somehow, everyone outside the field knows that neural networks are the Dominant-Paradigm-Overthrowing New Idea, ever since backpropagation was invented in the 1970s.”
  2. Analogies and General Priors on Intelligence — AI Alignment Forum AI Alignment Forum · 20 August 2021
    “The idea that you could create human-level intelligence by just feeding raw data into the right math-inspired architecture is pure fantasy.”
  3. Analogies and General Priors on Intelligence — AI Alignment Forum AI Alignment Forum · 20 August 2021
    “By contrast, I think Yudkowsky was saying that AGI source code would be relatively simple and coherent and plausibly written by one team, and then that relatively simple source code would give rise to lots of different skills via learning.”
  4. Dialogue with Eliezer Yudkowsky on Neural Networks — LessWrong LessWrong (post by aashish, with replies by Eliezer Yudkowsky) · 2 October 2026 Own record
    “One said: If you wire enough neurons together, it will emergently start thinking just like a human brain does. They usually did not use gradient descent, since gradient descent would never work on that many “neurons”.”
  5. Dialogue with Eliezer Yudkowsky on Neural Networks — LessWrong LessWrong (post by aashish, with replies by Eliezer Yudkowsky) · 2 October 2026 Own record
    “There is to be clear still a big damn postmortem from my perspective on two actual Bad Predictions made for Bad Reasons, which are from my perspective something like:”
  6. Dialogue with Eliezer Yudkowsky on Neural Networks — LessWrong LessWrong (post by aashish, with replies by Eliezer Yudkowsky) · 2 October 2026 Own record
    “This cached thought has been around for three decades. Still no general intelligence.”
  7. Analogies and General Priors on Intelligence — AI Alignment Forum AI Alignment Forum (riceissa, Sammy Martin; comment by Steven Byrnes) · 20 August 2021
    “By contrast, I think Yudkowsky was saying that AGI source code would be relatively simple and coherent and plausibly written by one team, and then that relatively simple source code would give rise to lots of different skills via learning.”
  8. Eliezer Yudkowsky: Track Record LongtermWiki · 1 February 2026
    “Yudkowsky's own assessment: "Reality was far to the Eliezer side of Eliezer on the Eliezer-Robin axis, and things like GPT-3 were built with less architectural complexity and generalized more than I was arguing to Robin that complex architectures should generalize over domains."”
  9. Eliezer Yudkowsky: Track Record LongtermWiki · 1 February 2026
    “| Scaling | "Stack more layers" won't work | Updated that it went further than expected | GPT-4 |”
  10. Eliezer Yudkowsky: Track Record LongtermWiki · 1 February 2026
    “LLMs are "a bit smarter than I thought this technology was going to scale to"”
  11. How Intelligible is Intelligence? Machine Intelligence Research Institute (Anna Salamon, Stephen Rayhawk, János Kramár) · 2010 Own record
    “We discuss theoretical and empirical results bearing on the intelligibility of intelligence, and argue that they modestly favor a relatively high estimate.”

Reasoning

An independent track-record page quotes Yudkowsky saying GPT-3 was built with less architectural complexity, and generalized further, than he had argued to Robin Hanson that complex architectures would (E9). The same page lists his earlier scaling position as "'Stack more layers' won't work" (E10) and quotes him saying LLMs went further than he expected (E11). This fits him having favoured complex, non-uniform architectures and being surprised. However, the surprise itself is shown mostly by his own later statements, and no quote from that period directly shows him stressing that the brain is not homogeneous. One commenter reads his foom-debate position as expecting relatively simple AGI code (E4/E8), but alongside E9 that shows a difference of degree rather than a contradiction.

Cite this claim

Eliezer Yudkowsky, “Yudkowsky was unpleasantly surprised by how well transformers scaled, having believed the human brain is not homogeneous.” (said 25 March 2024). Verdict: mostly true. Reliability API, claim c_e12fd543dc03, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_e12fd543dc03

Verdicts are made automatically; how verdicts and scores are made. The same data is on Hugging Face under CC BY 4.0.