ClaimsPeople

Eliezer Yudkowsky

Eliezer Yudkowsky founded the Machine Intelligence Research Institute (MIRI) and LessWrong, co-authored the best-selling book 'If Anyone Builds It, Everyone Dies' (Little, Brown, 2025), and was named to the TIME100 AI 2023 list.

2,016 claims collected, 236 researched, 79 resolved. Report of 4 November 2026, method 0.1.0+c876030c.

Scores by kind of claim

Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.

  • Factual claims

    Statements about the present or the past, checked against the record: did they hold up?

    76 70 82

    High confidence, 71 factual claims resolved. From 16 documents, 2007 to 2026; 18 documents sampled from 23 candidates found.

    Not scored: 50 unresolvable, 8 checkable only against their own records.

  • Predictions

    Statements about events outside the person's control, checked once they fell due: did they come true?

    60 42 78

    Low confidence, 8 predictions resolved. From 4 documents, 2008 to 2024; 18 documents sampled from 23 candidates found.

    Not scored: 97 unresolvable. 1 more fall due later.

    Does not yet allow for how likely each prediction was when made.

All types combined: 74 (likely 68 to 80, high confidence), from 79 claims: 71 factual claims, 8 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.

Eliezer Yudkowsky's claims

2,016 claims, sorted by newest said. Choose a claim to read it in full.
VerdictKindSaidClaimTopic
Not checkable Not checkable 10 Jun 2022 Human beings cannot inspect an AGI’s output to determine whether the consequences will be good. Not checkable · 10 Jun 2022
Mostly true Factual 10 Jun 2022 Human raters make systematic, regular, compactly describable and predictable errors. Factual · 10 Jun 2022 AI safety and alignment
Not checkable Not checkable 10 Jun 2022 This field is not making real progress and does not have a recognition function to distinguish real progress if it took place. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 Corrigibility is anti-natural to consequentialist reasoning Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 This is a very lethal problem, it has to be solved one way or another, it has to be solved at a minimum strength and difficulty level… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 When you explicitly optimize against a detector of unaligned thoughts, you’re partially optimizing for more aligned thoughts, and partially… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 But I’d have more hope – not significant hope, but more hope – in separating the concerns of (a) credibly promising to pay big money… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 You don’t get any cynical old veterans, in this case, because everybody on Earth is dead. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 when it comes to a choice of utility function, there are unbounded degrees of freedom and multiple reflectively coherent fixpoints. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 So far as I’m concerned, if you can get a powerful AGI that carries out some pivotal superhuman engineering task, with a less than fifty… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 It does not appear to me that the field of ‘AI safety’ is currently being remotely productive on tackling its enormous lethal problems. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 Nobody knows what the hell GPT-3 is thinking Not checkable · 10 Jun 2022
Unresolvable Prediction 10 Jun 2022 Systems much smarter than humans would be able to learn from less evidence than humans require. Prediction · 10 Jun 2022 AI capabilities
Not checkable Not checkable 10 Jun 2022 Anticipating all of the really dangerous ones, and then successfully materializing them, in the correct form for early solutions to… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 Human beings can figure out pretty difficult things over time, when they get lots of tries; when a failed guess kills literally everyone… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 If all you need is an object that doesn’t do dangerous things, you could try a sponge; a sponge is very passively safe. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 It’s not like people will yell at themselves about prospective alignment difficulties, they don’t have an internal voice of caution. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 I have several times failed to write up a well-organized list of reasons why AGI will kill you. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 If you perfectly learn and perfectly maximize the referent of rewards assigned by human operators, that kills them. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 Surviving worlds, by this point, and in fact several decades earlier, have a plan for how to survive. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 And become that cynical old veteran right away, before reality whaps you upside the head in the form of everybody dying and you not getting… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 All of these kill you if optimized-over by a sufficiently powerful intelligence Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 The best and easiest-found-by-optimization algorithms for solving problems we want an AI to solve, readily generalize to problems we’d… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 There is no pivotal output of an AGI that is humanly checkable and can be used to safely save the world but only after checking it Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 We’ve got no idea what’s actually going on inside the giant inscrutable matrices and tensors of floating-point numbers. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 The big ask from AGI alignment, the basic challenge I am saying is too difficult, is to obtain by any strategy whatsoever a significant… Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 I do not know how to train it into others. Not checkable · 10 Jun 2022
Unresolvable Prediction 10 Jun 2022 Even if an AGI ends up perfectly inner-aligned on a sensory reward signal, an AGI strongly optimizing on that signal "will kill you", because the sensory reward signal was not a ground truth about alignment as seen by the operators. Prediction · 10 Jun 2022 AI safety and alignment
Not checkable Not checkable 10 Jun 2022 Therefore, capabilities generalize further out-of-distribution than alignment, once they start to generalize at all. Not checkable · 10 Jun 2022
Not checkable Not checkable 10 Jun 2022 This general problem is a fact about the territory, not the map Not checkable · 10 Jun 2022
Unresolvable Prediction 10 Jun 2022 Many anti-corrigible lines of reasoning may first appear only at high levels of AI intelligence. Prediction · 10 Jun 2022 AI safety and alignment
Not checkable Not checkable 10 Jun 2022 They probably do not know where the real difficulties are, they probably do not understand what needs to be done, they cannot tell the… Not checkable · 10 Jun 2022
Mostly true Factual 31 Dec 2021 No electric heater is more energy-efficient than any other electric heater. Factual · 31 Dec 2021 Other
Not checkable Not checkable 25 Dec 2021 English may seem like a terrible language now, but imagine aliens running across it and trying to infer back human society based on the… Not checkable · 25 Dec 2021
Not checkable Not checkable 23 Dec 2021 I really really really hope that they did warn him and Biden is just lying, because the alternative is that everyone in or near the White… Not checkable · 23 Dec 2021
Resolved true Factual 21 Dec 2021 There are 365.2422 days between spring equinoxes, 365.259636 days between Earth's perihelions, and 365.25 days in a light-year. Factual · 21 Dec 2021 Other
Not checkable Not checkable 21 Dec 2021 Yeah, see, the problem is, if you die in that science-fiction movie, you die in real life. Not checkable · 21 Dec 2021
Not checkable Not checkable 16 Dec 2021 look I am not judging you I am comparing your performance unfavorably to that of an ideal agent and that's totally different Not checkable · 16 Dec 2021
Not checkable Not checkable 5 Dec 2021 Directly challenges alignment; doesn't significantly burn the capability commons that I noticed; doesn't overstate the accomplishment and… Not checkable · 5 Dec 2021
Not checkable Not checkable 5 Dec 2021 I'm slightly cheerful about this paper; it caused me to mildly positively update on @AnthropicAI. Not checkable · 5 Dec 2021
Not checkable Not checkable 3 Dec 2021 (Interesting theory on a subject that continues to interest me.) Not checkable · 3 Dec 2021
Not checkable Not checkable 29 Nov 2021 if you don't want to create needless public panic and plunge us all into the overt appearance of a scifi catastrophe flick, don't name… Not checkable · 29 Nov 2021
Mostly true Factual 28 Nov 2021 There is no legal duty for a corporation's board or management to maximize its profits. Factual · 28 Nov 2021 Economics and markets
Not checkable Not checkable 25 Nov 2021 At the rate we're going through the Greek alphabet, we don't have much time left to vaccinate the world before we face the Omega Variant. Not checkable · 25 Nov 2021
Not checkable Not checkable 23 Nov 2021 FDA plans to kill 50,000 Americans over the next month - over 16X the direct casualties from 2001/9/11 - by keeping it illegal for doctors… Not checkable · 23 Nov 2021
Not checkable Not checkable 23 Nov 2021 Lots of small weird stuff still underfunded IMO, and this is where non-billionaires get their chance to shine. Not checkable · 23 Nov 2021
Not checkable Not checkable 22 Nov 2021 If he's representing the state of knowledge accurately, this is the point where I call it for artificial origin of SARS-COV-2. Not checkable · 22 Nov 2021
Not checkable Not checkable 22 Nov 2021 Matt Ridley is on the very short list of journalists I would answer if they emailed me. Not checkable · 22 Nov 2021
Not checkable Not checkable 19 Nov 2021 Having built up dath ilan a bit, I'm surprised in retrospect that I can't think of any other science fiction set in a World of Nerds. Not checkable · 19 Nov 2021
Not checkable Not checkable 17 Nov 2021 I am being sooo very good right now. Not checkable · 17 Nov 2021
Not checkable Not checkable 17 Nov 2021 Today there was an antisemitic tweet mentioning me that has over 1K Likes, and I'm doing the right thing and *not* screenshotting it to… Not checkable · 17 Nov 2021
Not checkable Not checkable 16 Nov 2021 I notice that I am confused about how any human being discovered smelting ore for the first time. Not checkable · 16 Nov 2021
Not checkable Not checkable 12 Nov 2021 remember, anything over 70% cacao is not technically "chocolate" but a kind of edible dirt from which all the fun has been extracted Not checkable · 12 Nov 2021
Not checkable Not checkable 12 Nov 2021 The dath ilani equivalent of the 4th of July is the annual Oops It's Time To Overthrow The Government Festival, in which citizens rehearse… Not checkable · 12 Nov 2021
Not checkable Not checkable 12 Nov 2021 The stereotype I have for venture capital says that they will tend to fund 7 companies that all claim they'll have 20% of the sprocket… Not checkable · 12 Nov 2021
Not checkable Not checkable 10 Nov 2021 This is not a universal fact, but a circumstantial one. Not checkable · 10 Nov 2021
Not checkable Not checkable 10 Nov 2021 Let me now be clear, in case anyone who matters happens to be listening, that the Fed should tighten more, right away. Not checkable · 10 Nov 2021
Mostly true Factual 10 Nov 2021 Eliezer Yudkowsky has often criticized the Federal Reserve for not creating enough money fast enough. Factual · 10 Nov 2021 Economics and markets
Not checkable Not checkable 9 Nov 2021 If you're going to cheat at all, cheat recursively. Not checkable · 9 Nov 2021
Not checkable Not checkable 8 Nov 2021 Okay, honestly I'm a bit happy about living in a world where the richest openly-rich person on the planet just renamed himself to "Lorde… Not checkable · 8 Nov 2021
Not checkable Not checkable 7 Nov 2021 Just imagine if each Bitcoin had been one-trillionth of the maximum total supply, by a universal convention which all coins adopted… Not checkable · 7 Nov 2021
Not checkable Not checkable 7 Nov 2021 That this never happened says something basic about humanity. Not checkable · 7 Nov 2021
Unresolvable Prediction 6 Nov 2021 As a worst-case ('pessimal') prediction, Yudkowsky said the Fed would stay wrong about rising NGDP flow being the real cause of inflation until a Republican won the 2024 presidential election, after which it would decide interest rates needed to go way up. Prediction · 6 Nov 2021 Economics and markets
Not checkable Not checkable 6 Nov 2021 Keep the "fall back" where we set the clock an hour earlier, but ditch the "spring forward" when we set the clock an hour later. Not checkable · 6 Nov 2021
Not checkable Not checkable 6 Nov 2021 Be suspicious when you witness a debate where the pessimists have lots of concrete examples and scenarios and the optimists are speaking… Not checkable · 6 Nov 2021
Not checkable Not checkable 3 Nov 2021 If EMH wasn't true about almost-all prices relative to you, you could trade arbitrary assets and go up 10% per day. Not checkable · 3 Nov 2021
Not checkable Not checkable 3 Nov 2021 The only serious debate is whether the Efficient Markets Hypothesis is true about *all* asset prices *all* of the time for you, or merely… Not checkable · 3 Nov 2021
Not checkable Not checkable 31 Dec 2020 _Screwtape Letters_ is close enough to being Good and True that I'm having trouble reading it Not checkable · 31 Dec 2020
Not checkable Not checkable 31 Dec 2020 Savor the last day of 2020 while you can. Not checkable · 31 Dec 2020
Not checkable Not checkable 30 Dec 2020 This is practically an AU Sequence. Not checkable · 30 Dec 2020
Not checkable Not checkable 28 Dec 2020 White chocolate is best chocolate Not checkable · 28 Dec 2020
Not checkable Not checkable 28 Dec 2020 they have chosen to professionally specialize in fact-insensitive Frowning In Concern At Things. Not checkable · 28 Dec 2020
Mostly true Factual 26 Dec 2020 E. T. Jaynes said many times that probability is ignorance and that ignorance is a mental rather than physical property, and Yudkowsky repeated this often. Factual · 26 Dec 2020 Science and research methods
Not checkable Not checkable 22 Dec 2020 are just Trans-Exclusionary Radical Humanists. Not checkable · 22 Dec 2020
Not checkable Not checkable 20 Dec 2020 why would I want to become a vampire and see in the dark when I could, instead, just cover everything around me in glow-in-the-dark stickers Not checkable · 20 Dec 2020
Not checkable Not checkable 20 Dec 2020 I'm pretty sure they've already been given some info, and that info is all wrong Not checkable · 20 Dec 2020
Not checkable Not checkable 19 Dec 2020 A person who needs to feel ironically superior to something, and somehow manages to pick fucking probability theory Not checkable · 19 Dec 2020
Not checkable Not checkable 17 Dec 2020 Lawful good: Base 10 Not checkable · 17 Dec 2020
Unresolvable Prediction 16 Dec 2020 By about 2022, there may be empirical data on how much average men value female intelligence, e.g. what share would pay $5/month for a girlfriend as smart as GPT-3 rather than GPT-2. Prediction · 16 Dec 2020 AI, the economy and work
Not checkable Not checkable 16 Dec 2020 I kept expecting this to happen and wondering why it wasn't already happening Not checkable · 16 Dec 2020
Resolved true Factual 12 Dec 2020 AlphaFold2 was apparently a trendline-breaking sudden jump in capability on protein structure prediction. Factual · 12 Dec 2020 AI capabilities
Not checkable Not checkable 12 Dec 2020 I worry about dividing the United States because I currently estimate that if you take the center of all blue US states, plus the center of… Not checkable · 12 Dec 2020
Not checkable Not checkable 11 Dec 2020 Vote for good epistemics. Not checkable · 11 Dec 2020
Not checkable Not checkable 7 Dec 2020 How strange it is that modern US politics has become completely devoid of both liberalism and conservatism Not checkable · 7 Dec 2020
Not checkable Not checkable 2 Dec 2020 Looks like all our tweeting has had an effect Not checkable · 2 Dec 2020
Not checkable Not checkable 30 Nov 2020 It was silly that people tried to argue with that massively overdetermined step, but they did. Not checkable · 30 Nov 2020
Not checkable Not checkable 25 Nov 2020 2000 was the most consequential election if you were alive then. Not checkable · 25 Nov 2020
Not checkable Not checkable 25 Nov 2020 We need a Wikipedia-like group that can duly investigate such allegations of journalistic malfeasance Not checkable · 25 Nov 2020
Not checkable Not checkable 20 Nov 2020 Those effect sizes seem large for people who were already selected to be salespeople in the first place. Not checkable · 20 Nov 2020
Not checkable Not checkable 20 Nov 2020 It's disconcerting how much online advertising seems to be centered around an absolute denial of agency to the viewers. Not checkable · 20 Nov 2020
Resolved true Factual 14 Nov 2020 Yudkowsky said that all the sources he could find in November 2020 said that scaling up Pfizer vaccine production to more than 1e8 doses takes months. Factual · 14 Nov 2020 Technology
Not checkable Not checkable 13 Nov 2020 Maybe the best way to fight fake news is to create so much of it that people give up and check Wikipedia. Not checkable · 13 Nov 2020
Unresolvable Factual 11 Nov 2020 People who heard the slogan "defund the police" did not believe it meant something other than what it literally said. Factual · 11 Nov 2020 Philosophy, ethics and society
Not checkable Not checkable 9 Nov 2020 Though technically I'd say this is Optimizer's Curse, not Winner's Curse. Not checkable · 9 Nov 2020
Unresolvable Factual 9 Nov 2020 The stock market rise after Biden's 2020 election victory seems to imply that the primary stock markets assigned odds to a Trump victory more like the betting markets' 35% than 538's 10%. Factual · 9 Nov 2020 Economics and markets
Not checkable Not checkable 7 Nov 2020 I don't think that's how the system should work Not checkable · 7 Nov 2020
Not checkable Not checkable 6 Nov 2020 So many memes, so few that are truly important. Not checkable · 6 Nov 2020
Not checkable Not checkable 4 Nov 2020 I can't ever remember seeing a big-money betting market doing anything like this. Not checkable · 4 Nov 2020
Not checkable Not checkable 4 Nov 2020 I think the timelines are literally diverging right now! Not checkable · 4 Nov 2020
Not checkable Not checkable 3 Nov 2020 There's so many different sad jokes I could make about this that I'm going to leave them as a reader exercise. Not checkable · 3 Nov 2020
Back to the list

Not checkable

“I'm slightly cheerful about this paper; it caused me to mildly positively update on @AnthropicAI.”

Eliezer Yudkowsky 5 December 2021 Post by @ESYudkowsky on X, 2021-12-05 (X)

I'm slightly cheerful about this paper; it caused me to mildly positively update on @AnthropicAI.

Not checkable An opinion, value judgement or statement with nothing to check.

Cite this claim

Eliezer Yudkowsky, “I'm slightly cheerful about this paper; it caused me to mildly positively update on @AnthropicAI.” (said 5 December 2021). Verdict: not checkable. Reliability API, claim c_d21a21d54c8e, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_d21a21d54c8e

Claims without a verdict

Is a verdict here wrong, or do you speak for Eliezer Yudkowsky? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.