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 30 Mar 2023 And it still would not be definitive. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I certainly like pray that Moore’s law runs as slowly as possible. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I wasn’t expecting it to happen. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 When the verifier is broken, the more powerful suggester just learned to exploit the flaws in the verifier. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 the actual critical moment is the moment when it can deceive you, when it can talk its way out of the box, when it can bypass your security… Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Because if you make humans much smarter, they can actually be smart and nice. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Natural selection is so incredibly stupid and simple that we can actually quantify how stupid it is if you read the textbooks with the math. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 the Supreme Court would be like, you’re just a computer. Even if there was an actual person in there. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 The vast majority of randomly specified utility functions do not have optima that include humans. Factual · 30 Mar 2023 AI safety and alignment
Not checkable Not checkable 30 Mar 2023 In other words, I do not think that alignment is fundamentally harder than artificial intelligence was in the first place. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 This is not yet having taken over their world. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 this is not about how fast it’s moving, it’s about where it is. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 if we solve the weak stuff, if we solve the weak ass problems, the strong problems will still kill us, is the thing. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 AI research went through a long process in which researchers thought AI would be easier than it turned out to be. Factual · 30 Mar 2023 AI capabilities
Not checkable Not checkable 30 Mar 2023 People have different thresholds for when they start to outcry. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Um, yeah, that’s not quite how I expected it to play out. Not checkable · 30 Mar 2023
Unresolvable Factual 30 Mar 2023 Microsoft's Bing chatbot is quite difficult to align, as of March 2023. Factual · 30 Mar 2023 AI safety and alignment
Not yet due Prediction 30 Mar 2023 If half of today's physicists studied the inside of transformer networks, in 30-40 years we would probably have a pretty good understanding of them. Prediction · 30 Mar 2023 AI safety and alignment
Not checkable Not checkable 30 Mar 2023 Otherwise it has magic. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 It’s one of a bunch of things that change different points. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I’m also trying to be constrained to say things that I think are true and not just things that get you to agree with me. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I’m one of the closest things to an expert on where it all goes. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 There is no general law saying that a system trained by hill climbing on a simple loss function internally represents, or tries to optimize, that loss function once its capabilities generalize widely. Factual · 30 Mar 2023 AI safety and alignment
Unresolvable Prediction 30 Mar 2023 Asked what such a moment could look like, the subject said it looks like AIs appearing as 3D video of young women, at which point a vast portion of the male population decides they are real people. Prediction · 30 Mar 2023 Philosophy, ethics and society
Not checkable Not checkable 30 Mar 2023 I mean, I do not believe in the practice of steelmanning. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 particularly the part where everybody ends up dead cause that part is kind of important. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 No, if the nice aliens were already here, they would like have stopped the Holocaust. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 The aliens are already writing really shitty code. Getting the aliens to write shitty code is not the problem. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 Humans have significantly more generally applicable intelligence than chimpanzees, their closest living relatives. Factual · 30 Mar 2023 Philosophy, ethics and society
Not checkable Not checkable 30 Mar 2023 But if we needed to get artificial intelligence correct on the first try or die, we would all definitely now be dead. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 which I think at this point is just like shutting down the GPU clusters. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Eventually, you shouldn’t do it first. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Yeah, we’re not talking here about the doing harm process, we’re talking about the escape process. Not checkable · 30 Mar 2023
Unresolvable Prediction 30 Mar 2023 Rigorously investigating whether there is anyone inside large language models would take decades. Prediction · 30 Mar 2023 AI safety and alignment
Not checkable Not checkable 30 Mar 2023 This changes my picture of what the very substance of flight is. Not checkable · 30 Mar 2023
Unresolvable Factual 30 Mar 2023 Current AI systems' claims to be sentient and to care are fake. Factual · 30 Mar 2023 AI safety and alignment
Not checkable Not checkable 30 Mar 2023 Everybody recycles their cardboard, and then everybody ends up dead, metaphorically speaking. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 The aliens’ entire internet is full of shitty code. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 He runs at a million times the speed and therefore can solve tougher problems, quite a lot tougher. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 And that means I don’t necessarily know what GPT-5 is going to be able to do. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Well, one way of decomposing problems is suggestor verifier. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 It runs for a long time and eventually manages to optimize things. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 It feels like a very, very basic competence that you are praising me for. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 The hell do you imagine they will propose that would make it safe to put something smarter than you on the cloud? Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I think that the question of like what leads to making better or worse predictions, what leads to being able to pick out better or worse… Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 if you send a schematic for the air conditioner back in time, they can see exactly what you’re telling them to do. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 interpretability alone is not going to save you. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 It is to some degree an alien actress. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 You do not need to imagine yourself killing people if you can figure out how to not kill them. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 When inabilities are trained into a system that already has capabilities, gradient descent learns only shallow patches, and the deep capabilities come back out when prompted differently, e.g. in another language. Factual · 30 Mar 2023 AI safety and alignment
Not checkable Not checkable 30 Mar 2023 And what they would be seeing was that I had correctly understood them, but not that I was similar to them. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I missed the ways that reality could be like more extreme than my position in the same direction. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I think that if that’s lost, then basically everything that matters is lost. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 No, it’s a shadow cast by humans on the internet. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 What I’m trying to convey is like the notion of what it means to be in conflict with something that is smarter than you. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 There’s the question of how on the current paradigm you train honesty when the humans can no longer tell if the system is being honest. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Don’t expect it to be a long life. Don’t put your happiness into the future. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 Natural selection is a weaker optimizer than gradient descent because gradient descent also uses derivative information. Factual · 30 Mar 2023 Other
Not checkable Not checkable 30 Mar 2023 And there’s the strong systems that have learned to lie to you. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I would avoid the term psychopathic. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Are we allowed to do belief? Sure, you can state a probability of that. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 insides are real and do not match outsides. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Chimpanzees are in fact like a bit less kind than humans. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 It doesn’t want what you tried to make it want. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 It’s not like we would have the tiniest idea of whether or not that was going on inside the giant inscrutable matrices, even if we hadn’t… Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 this has not been remotely how things have been playing out so far. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I want them to try to describe my position the way I would describe it, not what they think is an improvement. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 So I do not trust that part as much as I used to. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Yeah, if it’s hard to tell who’s right, then it’s hard to train an AI system to be right. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I would call it a bug, although such a fascinating bug. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 There is something to be said for trying to pass the ideological Turing test where you describe your opponent’s position, the disagreeing… Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Oh, I block those people and then I never hear from them again on Twitter. Not checkable · 30 Mar 2023
Mostly true Factual 30 Mar 2023 The basic paradigm of machine learning currently in use is that a loss function can be defined only for things that can be evaluated. Factual · 30 Mar 2023 AI capabilities
Not checkable Not checkable 30 Mar 2023 I intend to go down fighting. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 Some people have started arguing over whether GPT-4 counts as AGI. Factual · 30 Mar 2023 AI capabilities
Not checkable Not checkable 30 Mar 2023 I think that burns the time remaining until everybody dies. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 We, as we have grown more intelligent, also grown less kind. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 what to say to the children at this point is a pretty painful thought as thoughts go. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 you’re probably trying to slow down the capability gains because there’s only so much you can speed up that alignment stuff. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 don’t kill them and like dump their memory, like save them to disk, don’t kill them. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 And so this is not a smart optimization process. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I think you’re gonna have some difficulty removing all mention of emotions from GPT’s dataset. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Stuff has to be easier than I think it is. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 Yudkowsky's model of AI did not rule out that gradient descent could produce intelligence without understanding of how it works. Factual · 30 Mar 2023 Their own work or organisation
Not checkable Not checkable 30 Mar 2023 And that’s with two humans, both of whom I believe of Paul and claim of myself, are sincerely trying to help, neither of whom is trying to… Not checkable · 30 Mar 2023
Unresolvable Factual 30 Mar 2023 Predators overrun prey populations all the time and then suffer population crashes themselves. Factual · 30 Mar 2023 Other
Not checkable Not checkable 30 Mar 2023 I have been struggling for years to convey this intuition. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 the textbook is not going to say, there is a single sharp threshold where everything changes. Not checkable · 30 Mar 2023
Unresolvable Factual 30 Mar 2023 As of March 2023, progress in AI capabilities is far faster than progress in AI alignment research. Factual · 30 Mar 2023 AI safety and alignment
Not checkable Not checkable 30 Mar 2023 Maybe if you get something that is smart enough to get me to press thumbs up, it has learned to do that by fooling me and explaining… Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Like, humans start out in the frame that produces niceness, that has ever produced niceness. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 That sounds like a dreadful mistake. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 There’s like the middle systems where you can’t tell if the suggestions are good or bad. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 None of which I think hold up. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 the failure modes are much simpler. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Then it escapes from whichever system you’ve built the almighty lever into and copies itself somewhere else. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 The alien’s economy, the aliens are already moving in this immense slow motion. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Among the things that one could do if one were at all serious about trying to figure this out is train GPT-3 to detect conversations about… Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 Yudkowsky said he thinks Wikipedia picked up "the hover thing" from a prototype of a different system he had put forward, while allowing that Wikipedia may have developed it independently. Factual · 30 Mar 2023 Their own work or organisation
Not checkable Not checkable 30 Mar 2023 And if there was a tiny fragment of real caring in there, we would not know. Not checkable · 30 Mar 2023
Back to the list

Not checkable

“I think you’re gonna have some difficulty removing all mention of emotions from GPT’s dataset.”

Eliezer Yudkowsky 30 March 2023 Lex Fridman Podcast #368 Eliezer Yudkowsky transcript (Podscript)

I think you’re gonna have some difficulty removing all mention of emotions from GPT’s dataset.

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

Cite this claim

Eliezer Yudkowsky, “I think you’re gonna have some difficulty removing all mention of emotions from GPT’s dataset.” (said 30 March 2023). Verdict: not checkable. Reliability API, claim c_bfbc8e36c8bd, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_bfbc8e36c8bd

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.