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 2008 Anthropomorphic bias can be classed as insidious: it takes place with no deliberate intent, without conscious realization, and in the face… Not checkable · 2008
Not checkable Not checkable 2008 To improve the truth of our beliefs we must change our beliefs. Not checkable · 2008
Resolved true Factual 2008 Biases implicit in the availability heuristic affect people's estimates of risk. Factual · 2008 Science and research methods
Not checkable Not checkable 2008 A success on human intelligence enhancement would make Friendly AI easier, and also help on other technologies. Not checkable · 2008
Not checkable Not checkable 2008 If you throw too much money at a problem that a field is not prepared to solve, the excess money is more likely to produce anti-science… Not checkable · 2008
Not checkable Not checkable 2008 Molecular nanotechnology fits this criterion, first because its elementary operations are fast, and second because there exists a ready… Not checkable · 2008
Not checkable Not checkable 2008 The prospect of AI interacts with global risk in more complex ways than that; if AI were a pure liability, matters would be simple. Not checkable · 2008
Not checkable Not checkable 2008 All that vast potential hinges on our survival here, now, in the days when the realm of humankind is a single planet orbiting a single star. Not checkable · 2008
Not checkable Not checkable 2008 EP is a powerful, still maturing technique that is intrinsically unsuited to the demands of Friendly AI. Not checkable · 2008
Not checkable Not checkable 2008 An optimization process steers the future into particular regions of the possible. Not checkable · 2008
Not checkable Not checkable 2008 But if people are liable to confidently believe that the button does something quite different from its actual consequence, that is cause… Not checkable · 2008
Not checkable Not checkable 2008 I don’t think this exact scenario is plausible, mostly because I don’t expect the curve of recursive self-improvement to move at a linear… Not checkable · 2008
Not checkable Not checkable 2008 There is a taboo against talking about humanlevel capabilities. Not checkable · 2008
Not checkable Not checkable 2008 Nature does not match her challenges to your skill, or your resources, or how much free time you have to think about the problem. Not checkable · 2008
Not checkable Not checkable 2008 Storytellers obey strict rules of narrative unrelated to reality. Not checkable · 2008
Not checkable Not checkable 2008 Any technology which increases available computing power decreases the minimum theoretical sophistication necessary to develop Artificial… Not checkable · 2008
Unresolvable Prediction 2008 Yudkowsky said he thinks Artificial Intelligence is the road we will end up taking into "that challenge", which the surrounding text describes as the way out of the high-mortality phase of the life cycle and probably the single most dangerous risk we face. Prediction · 2008 AI timelines and forecasting
Not checkable Not checkable 2008 Of all possible configurations for the atoms making up the Corolla, only an infinitesimal fraction qualify as a useful working car. Not checkable · 2008
Not researched Citation 2008 Latane and Darley's 1969 experiments found the bystander effect: larger groups are less likely to act in emergencies, individually and collectively. Citation · 2008 Science and research methods
Not checkable Not checkable 2008 Yet for a futurist, disjunctions make for an awkward and unpoetic-sounding prophecy. Not checkable · 2008
Not checkable Not checkable 2008 The danger of confusing general intelligence with g-factor is that it leads to tremendously underestimating the potential impact of… Not checkable · 2008
Unresolvable Prediction 2008 When technology first becomes capable of uploading, it will have overwhelmingly more computing power and probably far better cognitive science than building an AI requires. Prediction · 2008 AI timelines and forecasting
Not checkable Not checkable 2008 The subjects in these experiments don't make mistakes on purpose; they make mistakes because they don't know how to do better. Not checkable · 2008
Not checkable Not checkable 2008 To build an AI that discovers the orbits of the planets, the programmers need not know the math of Newtonian mechanics, only the math of… Not checkable · 2008
Not checkable Not checkable 2008 The above methodology for expert elicitation looks perfectly reasonable Not checkable · 2008
Not checkable Not checkable 2008 This may have been one contributing factor to the embarrassing history of AI, which stems not from the difficulty of AI as such, but from the… Not checkable · 2008
Not checkable Not checkable 2008 The scenario of humanity going extinct in the next century is a disjunctive event. Not checkable · 2008
Not checkable Not checkable 2008 Since the given description of Linda was chosen to be similar to a feminist and dissimilar to a bank teller, (8) is more representative of… Not checkable · 2008
Not researched Factual 2008 Discussions of existential risk not uncommonly draw their categories, choices, consequences and strategies from movies, books and television shows. Factual · 2008 Other global risks
Not researched Factual 2008 Later research has generalized and elaborated the phenomena in the Kahneman, Slovic and Tversky volume, and its basic results still stand. Factual · 2008 Science and research methods
Not researched Factual 2008 Extreme confidence in wrong answers is common in the heuristics and biases literature, where it is known as overconfidence. Factual · 2008 Science and research methods
Not checkable Not checkable 2008 g-factor metaphors seem most common in popular futurism: when people think of “intelligence” they think of human geniuses instead of humans. Not checkable · 2008
Not checkable Not checkable 2008 The second threat scenario is less vivid and alarming, but the defense is more useful because it is more vague. Not checkable · 2008
Mostly true Factual 2008 As of 2008 there is no shield that can deflect a nuclear explosion. Factual · 2008 Other global risks
Not checkable Not checkable 2008 Natural selection isn’t friendly, nor does it hate you, nor will it leave you alone. Not checkable · 2008
Not checkable Not checkable 2008 Often Nature poses requirements that are grossly unfair, even on tests where the penalty for failure is death. Not checkable · 2008
Unresolvable Prediction 2008 Yudkowsky said he feared regulatory agencies would not approve human experiments, in the context of augmenting human intelligence through biology. Prediction · 2008 AI policy and governance
Not checkable Not checkable 2008 But human beings are not trustworthy to peer over a purported proof of ten billion steps; we have too high a chance of missing an error. Not checkable · 2008
Not researched Factual 2008 Taking Piracetam or drinking caffeine will not make a person substantially smarter than Einstein. Factual · 2008 Science and research methods
Not researched Citation 2008 By the halting problem (Rice 1953), it is impossible to predict whether an arbitrary computational system implements any given input-output function, including simple multiplication. Citation · 2008 Science and research methods
Not checkable Not checkable 2008 We must execute the creation of Artificial Intelligence as the exact application of an exact art. Not checkable · 2008
Not checkable Not checkable 2008 People may be overconfident and over-optimistic. Not checkable · 2008
Not checkable Not checkable 2008 Awareness of human fallibility is a dangerous knowledge, if you remind yourself of the fallibility of those who disagree with you. Not checkable · 2008
Not checkable Not checkable 2008 The first moral is that confusing the speed of AI research with the speed of a real AI once built is like confusing the speed of physics… Not checkable · 2008
Not checkable Not checkable 2008 A majoritarian strategy only works if it is not possible for a single defector to cause global catastrophic damage. Not checkable · 2008
Not checkable Not checkable 2008 And when you run into a lethal challenge too difficult for you, you die. Not checkable · 2008
Not checkable Not checkable 2008 Every true idea which discomforts you will seem to match the pattern of at least one psychological error. Not checkable · 2008
Not checkable Not checkable 2008 The main advice the metaphor gives us is that we had better get Friendly AI right, which is good advice in any case. Not checkable · 2008
Not checkable Not checkable 2008 But in forecasting, every added detail is an extra burden! Not checkable · 2008
Unresolvable Prediction 2008 A superintelligent AI whose goal system operates in an unforeseen fashion may actively resist human attempts to correct it. Prediction · 2008 AI safety and alignment
Unresolvable Factual 2008 It seems that people do not extrapolate from experienced small hazards to the possibility of large risks. Factual · 2008 Philosophy, ethics and society
Unresolvable Factual 2008 The obvious experimental manipulations of heuristics-and-biases studies have already been tried, and the results were found to be robust. Factual · 2008 Science and research methods
Not researched Citation 2008 In the Buehler et al. study, students took on average 55 days to complete their thesis, 22 days longer than anticipated and 7 days longer than their worst-case predictions. Citation · 2008 Science and research methods
Not checkable Not checkable 2008 The elapsed turnaround time would be, imaginably, on the order of a week from when the fast intelligence first became able to solve the… Not checkable · 2008
Not checkable Not checkable 2008 Familiarity with heuristics and biases can enable us to detect a wide variety of logical flaws that might otherwise evade our inspection. Not checkable · 2008
Not checkable Not checkable 2008 A majoritarian strategy is sometimes workable, if you have decades in which to do your work. Not checkable · 2008
Not researched Factual 2008 The two Gilovich, Griffin and Kahneman edited volumes overview the heuristics and biases field in its current form. Factual · 2008 Science and research methods
Not checkable Not checkable 2008 As logic, the above argument contains holes. Not checkable · 2008
Not checkable Not checkable 2008 Few people would push a button that they clearly knew would cause a global catastrophe. Not checkable · 2008
Not checkable Not checkable 2008 I introduce the concept of an optimization process Not checkable · 2008
Not researched Factual 2008 Yudkowsky wrote that, to the best of his ability to discern, there were no existing major works on the topic to cite as of January 2006. Factual · 2008 AI safety and alignment
Not checkable Not checkable 2008 Because of hindsight bias, we learn overly specific lessons. Not checkable · 2008
Not checkable Not checkable 2008 We would expect a true AI, an Artificial General Intelligence, to be capable of changing its empirical beliefs Not checkable · 2008
Not checkable Not checkable 2008 And the stakes of existential risk extend beyond even the six billion humans alive today, to all the stars in all the galaxies that… Not checkable · 2008
Not checkable Not checkable 2008 But if they had said might, I would have no objection. Not checkable · 2008
Not checkable Not checkable 2008 The raw materials for AI are already everywhere. Not checkable · 2008
Resolved true Prediction 2008 Yudkowsky doubted that a thousandfold increase in AI hardware would buy anything like a thousandfold increase in effective intelligence, while noting that such doubt is not reliable without an analytical calculation. Prediction · 2008 AI compute, hardware and data
Not checkable Not checkable 2008 If the AI tribe wins in military conflict and wipes out the humans, that is an existential catastrophe of the Bang variety Not checkable · 2008
Not checkable Not checkable 2008 Substantially larger numbers, such as 500 million deaths, and especially qualitatively different scenarios such as the extinction of the… Not checkable · 2008
Not checkable Not checkable 2008 I do not assign strong confidence to the assertion that Friendly AI is easier than human augmentation, or that it is safer. Not checkable · 2008
Not researched Factual 2008 Building dams reduces flood frequency but damage per flood becomes so much greater that average yearly damage increases. Factual · 2008 Economics and markets
Not checkable Not checkable 2008 It only takes a single counterexample to falsify a universal quantifier. Not checkable · 2008
Not researched Factual 2008 It would be physically possible to build a brain computing a million times faster than a human brain without shrinking it, cooling it, or using reversible or quantum computing. Factual · 2008 Science and research methods
Not checkable Not checkable 2008 They are somewhat less accessible to a general audience. Not checkable · 2008
Not checkable Not checkable 2008 Otherwise you become slower to react to incoming evidence. Not checkable · 2008
Not checkable Not checkable 2008 But one should not so quickly write off the challenge, especially considering the stakes. Not checkable · 2008
Not checkable Not checkable 2008 Intelligence is the foundation of human power, the strength that fuels our other arts. Not checkable · 2008
Not checkable Not checkable 2008 The AI does not hate you, nor does it love you, but you are made out of atoms which it can use for something else. Not checkable · 2008
Not checkable Not checkable 2008 When we talk about “AIs” we are really talking about minds-in-general, or optimization processes in general. Not checkable · 2008
Not checkable Not checkable 2008 A dead planet, lifelessly orbiting its star, is also stable. Not checkable · 2008
Not checkable Not checkable 2008 But the stakes are much, much higher. Not checkable · 2008
Not checkable Not checkable 2008 Absolute power is a conservative engineering assumption in Friendly AI, exposing broken designs. Not checkable · 2008
Not checkable Not checkable 2008 I cannot see how nanotechnology would make it easier to develop Friendly AI. Not checkable · 2008
Not checkable Not checkable 2008 There is no way to obtain more truth for a proposition by bribery, flattery, or the most passionate argument - you can make more people… Not checkable · 2008
Not checkable Not checkable 2008 Friendly AI is not about brute-forcing the problem. Not checkable · 2008
Not checkable Not checkable 2008 Any two AI designs might be less similar to one another than you are to a petunia. Not checkable · 2008
Not checkable Not checkable 2008 The AI might transcend the factual error of its programmers, provided that the programmers understood inference rather better than they… Not checkable · 2008
Not checkable Not checkable 2008 A technical failure occurs when the code does not do what you think it does, though it faithfully executes as you programmed it. Not checkable · 2008
Not researched Factual 2008 Overconfidence applies strongly to planning, where it is known as the planning fallacy. Factual · 2008 Science and research methods
Not researched Factual 2008 Spearman's g is a statistical abstraction from individual differences in intelligence among humans, a species far more intelligent than lizards. Factual · 2008 Science and research methods
Unresolvable Prediction 2008 Yudkowsky said he thinks it will prove easier, in the end, to build a 747 from scratch than to scale up an existing bird or graft on jet engines, an analogy he used for building Artificial Intelligence rather than taking another road into the challenge of smarter-than-human intelligence. Prediction · 2008 AI timelines and forecasting
Unresolvable Prediction 2008 With molecular nanotechnology, an AI could potentially rewrite the solar system unopposed. Prediction · 2008 Other global risks
Not checkable Not checkable 2008 To live even a million years, as an unaging human in a world as risky as our own, you must somehow drive your annual probability of… Not checkable · 2008
Not checkable Not checkable 2008 Someone should know that when an expert says the probability is "a million to one" without using actuarial data or calculations from a… Not checkable · 2008
Not checkable Not checkable 2008 The first AI built cannot by itself do catastrophic damage. Not checkable · 2008
Own records only Factual 2008 When he first began thinking about these matters, Yudkowsky expected the largest benefits from Artificial Intelligence yet felt no anxiety about it. Factual · 2008 Their own work or organisation
Not checkable Not checkable 2008 It once occurred to me that modern civilization occupies an unstable state. Not checkable · 2008
Not checkable Not checkable 2008 Heuristics and biases are widespread in human reasoning. Not checkable · 2008
Not checkable Not checkable 2008 Friendly AI programmers should assume that the AI has total access to its own source code. Not checkable · 2008
Not checkable Not checkable 2008 Do-What-I-Mean is a major, nontrivial technical challenge of Friendly AI. Not checkable · 2008
Back to the list

Not checkable

“People may be overconfident and over-optimistic.”

Eliezer Yudkowsky 2008 Cognitive biases potentially affecting judgment of global risks (UC Berkeley (hosted))

People may be overconfident and over-optimistic.

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

Cite this claim

Eliezer Yudkowsky, “People may be overconfident and over-optimistic.” (said 2008). Verdict: not checkable. Reliability API, claim c_739506c2da3a, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_739506c2da3a

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.