ClaimsPeople

Gary Marcus

Gary Marcus is an author and AI entrepreneur and professor emeritus at NYU; he has written six books, including Rebooting AI and Taming Silicon Valley, founded Geometric Intelligence (bought by Uber), testified at the May 2023 US Senate AI hearing, and writes a newsletter with over 121,000 subscribers.

712 claims collected, 213 researched, 103 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?

    79 72 85

    High confidence, 64 factual claims resolved. From 18 documents, 2019 to 2026; 19 documents sampled from 37 candidates found.

    Not scored: 32 unresolvable, 4 checkable only against their own records.

  • Predictions

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

    80 72 88

    Medium confidence, 38 predictions resolved. From 7 documents, 2019 to 2025; 19 documents sampled from 37 candidates found.

    Not scored: 60 unresolvable. 13 more fall due later.

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

  • Promises

    Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?

    Insufficient record: 1 promise resolved, 5 needed for a score.

    Not scored: 1 unresolvable.

All types combined: 79 (likely 74 to 84, high confidence), from 102 claims: 64 factual claims, 38 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.

Gary Marcus's claims

712 claims, sorted by newest said. Choose a claim to read it in full.
VerdictKindSaidClaimTopic
Not checkable Not checkable 30 Mar 2023 But I don’t think that legitimate concerns about the hypey-ness of letter in anyway undermines the real need to rein in the systems are… Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Much (not all) of the pushback seemed off base to me Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I personally haven't changed; I still think that LLMs are unreliable, and still think that they are a very poor basis for factuality. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 But that doesn't mean that they don't have the potential to rip apart our social fabric Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 This doesn’t mean we are doomed, but we do need to think hard, and quickly, about what proper measures look like, just as we have done for… Not checkable · 30 Mar 2023
Unresolvable Factual 30 Mar 2023 Gary Marcus wrote a long Twitter thread on the morning of 2023-03-30 responding to criticism of the FLI letter. Factual · 30 Mar 2023 Their own work or organisation
Not researched Factual 30 Mar 2023 The FLI open letter did not call for a ban on GPT-4. Factual · 30 Mar 2023 Their own work or organisation
Not checkable Not checkable 30 Mar 2023 One criticism do I agree with is this Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 I am with Stability AI founder Emad Mostaque in his read on how most people actually working on this stuff see things: Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 Most signers of the FLI open letter, including Gary Marcus and Yoshua Bengio, have no connection to the Future of Life Institute. Factual · 30 Mar 2023 Their own work or organisation
Not checkable Not checkable 30 Mar 2023 The real news here is not that Elon Musk signed it but that so many people who are not natural allies (eg Bengio and l myself, famous for… Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 The letter Marcus was discussing did not call for a ban on most AI research, only for a brief pause on one very specific project involving a technology with known risks and no known solutions. Factual · 30 Mar 2023 Their own work or organisation
Not checkable Not checkable 30 Mar 2023 A lot of the attacks on the letter focused on who sponsored it, not who signed it. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Sooo many of the attacks on the proposed 6 month ban on training super large #LLMs miss the point. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 Please put your own suggestions in the comments below. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 It is *not* fine to do nothing. Not checkable · 30 Mar 2023
Resolved true Factual 30 Mar 2023 The FLI open letter called for more AI research. Factual · 30 Mar 2023 Their own work or organisation
Not checkable Not checkable 30 Mar 2023 The letter should by judged by what it says, not who wrote it. Not checkable · 30 Mar 2023
Not researched Factual 30 Mar 2023 Virtually everyone, including people at OpenAI, has acknowledged serious AI risks, but as of March 2023 few tangible steps to mitigate them have been taken by government or industry. Factual · 30 Mar 2023 AI policy and governance
Not checkable Not checkable 30 Mar 2023 I don't think they are close to AGI. Not checkable · 30 Mar 2023
Delivered on time Promise 30 Mar 2023 Gary Marcus will speak at TED about short-term and long-term AI risk and what to do about it, in about two weeks from 2023-03-30. Promise · 30 Mar 2023 Their own work or organisation
Resolved true Factual 30 Mar 2023 The FLI open letter did not call for a ban on AI or for a permanent ban. Factual · 30 Mar 2023 Their own work or organisation
Not checkable Not checkable 30 Mar 2023 It is perfectly fine to propose a different alternative, but most of the critiques of the letter have not. Not checkable · 30 Mar 2023
Mostly true Factual 30 Mar 2023 As of March 2023, none of the top AI labs is remotely transparent, AI governance basically doesn't exist, and there are basically no safeguards in place. Factual · 30 Mar 2023 AI policy and governance
Not checkable Not checkable 30 Mar 2023 As I said yesterday to The New York Times, “We have a perfect storm of corporate irresponsibility, widespread adoption, lack of regulation… Not checkable · 30 Mar 2023
Unresolvable Factual 30 Mar 2023 Gary Marcus signed the Future of Life Institute open letter and promoted it publicly. Factual · 30 Mar 2023 Their own work or organisation
Not checkable Not checkable 30 Mar 2023 Not everyone who signed the letter is principally concerned with long-term risk; many of us who signed are worried at least as much about… Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 The idea (which I have actually heard expressed) that AI should be exempt from regulation is absurd. Not checkable · 30 Mar 2023
Not checkable Not checkable 30 Mar 2023 To my mind, doing nothing is truly the most foolish option. Not checkable · 30 Mar 2023
Resolved true Factual 24 Mar 2023 There have already been reports of contamination in GPT-4's training data. Factual · 24 Mar 2023 AI capabilities
Not checkable Not checkable 24 Mar 2023 The strengths and weaknesses of GPT-4 are qualitatively the same as before. Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 They reify the practice of substituting press releases for science and the practice of discussing models with entirely undisclosed… Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 I don’t expect Coca Cola to present its secret formula. Not checkable · 24 Mar 2023
Mostly true Factual 24 Mar 2023 The OpenAI and Microsoft GPT-4 papers reveal nothing about the model's architecture or training set. Factual · 24 Mar 2023 AI capabilities
Not checkable Not checkable 24 Mar 2023 It’s a silly claim, given that it is entirely open to interpretation Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 Perhaps all this would all be fine if the companies weren’t pretending to be contributors to science, formatting their work as science with… Not checkable · 24 Mar 2023
Mostly true Factual 24 Mar 2023 Citing Ernie Davis, Marcus said OpenAI had begun to incorporate user experiments into its models' training corpus. Factual · 24 Mar 2023 AI capabilities
Resolved true Factual 24 Mar 2023 GPT-4 has not solved hallucinations, reliability or planning on complex tasks, as the paper's authors acknowledge for planning. Factual · 24 Mar 2023 AI capabilities
Not checkable Not checkable 24 Mar 2023 Microsoft and OpenAI are producing and deploying products with potentially enormous risks at mass scale, and making extravagant, unreviewed… Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 Microsoft and OpenAI are rolling out extraordinarily powerful yet unreliable systems with multiple disclosed risks and no clear measure… Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 But nor do I plan to give them scientific credibility for alleged advances that we know nothing about. Not checkable · 24 Mar 2023
Mostly true Factual 24 Mar 2023 Gary Marcus said that on 2023-03-23 Microsoft had put out what he called a press release masquerading as science, which claimed that GPT-4 was an early (yet still incomplete) version of an AGI system. Factual · 24 Mar 2023 AI capabilities
Not checkable Not checkable 24 Mar 2023 if Coca Cola wants to keep secrets, that’s fine; it’s not particularly in the public interest to know the exact formula. Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 By excluding the scientific community from any serious insight into the design and function of these models, Microsoft and OpenAI are… Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 if we don’t get it, we must contemplate shutting these projects down. Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 This cannot and should not stand. Not checkable · 24 Mar 2023
Not researched Factual 24 Mar 2023 OpenAI CEO Sam Altman has recently publicly expressed fears about where AI development is going. Factual · 24 Mar 2023 AI safety and alignment
Not checkable Not checkable 24 Mar 2023 That claim would never survive serious scientific peer review. Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 You would archive the message without reading further. Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 literally cannot be tested with serious scrutiny, because the scientific community has no access to the training data. Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 We must demand transparency Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 As I have said before, I don’t really think GPT-4 has much to do with AGI. Not checkable · 24 Mar 2023
Not checkable Not checkable 24 Mar 2023 At some point, we would want public hearings. Not checkable · 24 Mar 2023
Not researched Factual Feb 2023 Marcus's February 2023 warning that generative AI would corrupt the internet increasingly appears correct. Factual · Feb 2023 AI safety and alignment
Not checkable Not checkable 25 Dec 2022 I have never seen anything like this much buzz. Not checkable · 25 Dec 2022
Mostly true Prediction 25 Dec 2022 GPT-4 will still be reckless and hard to control, making a significant number of stupid errors. Prediction · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 GPT-4's natural language output still won't be reliably usable as input to downstream programs such as databases or virtual assistants. Prediction · 25 Dec 2022 AI capabilities
Mostly true Factual 25 Dec 2022 ChatGPT regularly makes things up, and a fair bit of what it says is untrue. Factual · 25 Dec 2022 AI capabilities
Not checkable Not checkable 25 Dec 2022 But will it solve the problems we have seen before? I am not so sure. Not checkable · 25 Dec 2022
Unresolvable Prediction 25 Dec 2022 Trustworthy AI will need genuinely new architectures incorporating explicit knowledge and world models at their core. Prediction · 25 Dec 2022 AI capabilities
Unresolvable Factual 25 Dec 2022 ChatGPT is frequently retuned based on feedback, but weeks later Marcus still keeps being sent tweets showing its blunders. Factual · 25 Dec 2022 AI capabilities
Unresolvable Prediction 25 Dec 2022 When GPT-4 comes out it will totally eclipse ChatGPT, and it is a safe bet that even more people will be talking about it. Prediction · 25 Dec 2022 AI capabilities
Resolved true Prediction 25 Dec 2022 GPT-4 will definitely seem smarter than its predecessors, but its internal architecture will remain problematic. Prediction · 25 Dec 2022 AI capabilities
Not checkable Not checkable 25 Dec 2022 If all seven predictions prove correct, I hope that the field will finally realize that it is time to move on. Not checkable · 25 Dec 2022
Mostly true Prediction 25 Dec 2022 GPT-4 will not be trustworthy and complete enough to give reliable medical advice. Prediction · 25 Dec 2022 AI capabilities
Mostly true Factual 25 Dec 2022 ChatGPT might sound like the Star Trek computer, but for now it still cannot be trusted. Factual · 25 Dec 2022 AI capabilities
Unresolvable Prediction 25 Dec 2022 Marcus fully expected GPT-4 to be the shiniest model so far, but said that this does not mean it is a critical step on the optimal path to AI that can be trusted. Prediction · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 GPT-4 will be a very large, powerful system (a 'monster'). Prediction · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 With GPT-4, fluent hallucinations will still be common and easily induced, escalating the risk of LLMs being used to create misinformation. Prediction · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 GPT-4 may be better at faking term papers, but if it follows the same playbook as its predecessors, it still won't really understand the world, and its flaws will eventually show. Prediction · 25 Dec 2022 AI capabilities
Not yet due Prediction 25 Dec 2022 Within a decade, maybe much less, AI's focus will shift from pure scaling of large language models to integrating them with other techniques. Prediction · 25 Dec 2022 AI capabilities
Unresolvable Factual 25 Dec 2022 GPT-4 is architecturally essentially the same as GPT-3, according to rumors. Factual · 25 Dec 2022 AI capabilities
Unresolvable Prediction 25 Dec 2022 GPT-4 by itself won't be a general-purpose AGI; without external aids it won't beat Cicero in Diplomacy, drive a car reliably, or reliably guide a robot like Optimus. Prediction · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 GPT-4 will blow people's minds when released. Prediction · 25 Dec 2022 AI capabilities
Own records only Factual 25 Dec 2022 Marcus knows several people who have tried GPT-4, and all were impressed. Factual · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 Marcus suspected that GPT-4's release would follow a pattern of immense buzz, then scientific scrutiny, then recognition that many problems remain. Prediction · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 If GPT-4 is architecturally essentially the same as GPT-3, it can be expected to be unable to construct internal models of how the world works and, in consequence, unable to understand things at an abstract level. Prediction · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 AI alignment between human wants and machine behavior will remain a critical unsolved problem, with GPT-4 unable to reliably follow shared human values. Prediction · 25 Dec 2022 AI safety and alignment
Mostly true Factual 25 Dec 2022 ChatGPT can write about anything but is easily confused. Factual · 25 Dec 2022 AI capabilities
Unresolvable Prediction 25 Dec 2022 GPT-4's reasoning about the physical, psychological and mathematical world will still be unreliable. Prediction · 25 Dec 2022 AI capabilities
Mostly true Prediction 25 Dec 2022 GPT-4 will remain a pastiche generator and useful for brainstorming and first drafts, but not trustworthy general intelligence. Prediction · 25 Dec 2022 AI capabilities
Mostly true Factual Dec 2022 AI-generated misinformation became increasingly widespread, as Marcus warned in December 2022, but its effects have so far been less than he feared. Factual · Dec 2022 AI safety and alignment
Not researched Factual Dec 2022 Marcus's December 2022 prediction of chatbot-associated deaths was confirmed in 2023, with another apparent incident in 2024. Factual · Dec 2022 AI safety and alignment
Mostly true Factual Mar 2022 Gary Marcus's March 2022 article argued pure scaling would not solve hallucinations or abstraction and concluded there are serious holes in the scaling argument. Factual · Mar 2022 Their own work or organisation
Resolved true Factual Mar 2022 In March 2022 Gary Marcus published an article arguing that deep learning is hitting a wall. Factual · Mar 2022 Their own work or organisation
Not researched Factual 2022 Marcus's predictions since 2022 that pure LLM scaling would run out appear to have been confirmed in November and December 2024 by comments from industry insiders. Factual · 2022 AI timelines and forecasting
Not checkable Not checkable 29 Dec 2021 No current AI is remotely close to understanding the everyday physical or psychological world Not checkable · 29 Dec 2021
Not checkable Not checkable 29 Dec 2021 clueless statistical approximators. Not checkable · 29 Dec 2021
Not checkable Not checkable 29 Dec 2021 these systems are long way from understanding basic linguistic like negation Not checkable · 29 Dec 2021
Not checkable Not checkable 29 Dec 2021 Game on! Not checkable · 29 Dec 2021
Unresolvable Prediction 29 Dec 2021 Current AI, being an approximation to intelligence rather than the real thing, will never really be trustworthy. Prediction · 29 Dec 2021 AI safety and alignment
Not checkable Not checkable 28 Dec 2021 seriously lacking in AI Not checkable · 28 Dec 2021
Not checkable Not checkable 26 Dec 2021 This is not "awesome"; it is deeply confused. Not checkable · 26 Dec 2021
Not checkable Not checkable 23 Dec 2021 Good thread (and review) on neurosymbolic programming, very much in line with The Algebraic Mind, which I hope will get the attention it… Not checkable · 23 Dec 2021
Not checkable Not checkable 22 Dec 2021 we are calling for a respect for what nature has to teach us, and for humility in accepting what we have not yet discovered Not checkable · 22 Dec 2021
Not checkable Not checkable 22 Dec 2021 For those who have forgotten their biology Not checkable · 22 Dec 2021
Not checkable Not checkable 20 Dec 2021 Trainable deep networks accumulate statistics between features; they don’t induce operations over variables or structured representations… Not checkable · 20 Dec 2021
Unresolvable Factual 20 Dec 2021 Gary Marcus wrote the 2001 book The Algebraic Mind, and its predictions have so far held true. Factual · 20 Dec 2021 Their own work or organisation
Not checkable Not checkable 19 Dec 2021 It's not all about the data. Not checkable · 19 Dec 2021
Not checkable Not checkable 19 Dec 2021 the deeper problem, per The Algebraic Mind, remains lack of adequate substrate for structured representations, individuals & abstract… Not checkable · 19 Dec 2021
Back to the list

Gary Marcus

Open its page

Factual Their own work or organisation

“I wrote an entire book on this, in 2001 (The Algebraic Mind); so far its predictions have held true.”

Gary Marcus, author of the post 20 December 2021 Post by @GaryMarcus on X, 2021-12-20 (X)

Gary Marcus wrote the 2001 book The Algebraic Mind, and its predictions have so far held true.

Unresolvable The public record could not settle it.

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

What would count as true or false

Two parts. (a) Authorship: Gary Marcus wrote The Algebraic Mind, published in 2001 (MIT Press). (b) The book's predictions had held true as of December 2021, for example its predictions that multilayer perceptron or neural-network models lacking explicit symbol manipulation would fail to generalise free abstract rules outside the training space. True: (a) holds, and assessments up to 2021 (research literature, reviews) show the book's main stated predictions were borne out, with no clear counterexample. Mostly true: (a) holds and the main predictions were largely borne out, but at least one is contested or only partly borne out. Mixed: (a) holds, but the record on the predictions is roughly evenly split or seriously contested in the literature. Mostly false: (a) holds, but the predictions had substantially failed by 2021 or were too vague to have been borne out as claimed. False: Marcus did not write the book or it was not published around 2001, or its predictions were clearly refuted. An exact year off by one (for example a 2001 vs 2002 printing) counts as a minor detail.

Evidence

  1. Pre-Wiring and Pre-Training: What Does a Neural Network Need to Learn Truly General Identity Rules? Journal of Artificial Intelligence Research · April 2018
    “We therefore confirm that the SRN does not reproduce the empirical findings.”
  2. Pre-Wiring and Pre-Training: What Does a Neural Network Need to Learn Truly General Identity Rules? Journal of Artificial Intelligence Research · April 2018
    “Our study suggests that both the initial optimism about the generalization abilities of infants and the pessimism about neural networks were overstated.”
  3. Analysing Mathematical Reasoning Abilities of Neural Models arXiv (ar5iv rendering)
    “Marcus (2003) Gary F Marcus. The algebraic mind: Integrating connectionism and cognitive science. MIT press, 2003.”
  4. Analysing Mathematical Reasoning Abilities of Neural Models arXiv (ar5iv rendering)
    “Modules on which good extrapolation performance was obtained include rounding larger numbers than seen during training, comparing more numbers, and adding and subtracting larger numbers.”

Reasoning

E3 attributes The Algebraic Mind to Marcus but dates it 2003, two years after the stated 2001, and no item explains the gap. E1, E2 and E4 discuss what neural networks can and cannot do, but none of them refers to the book or assesses its predictions, so the evidence does not settle whether those predictions held true as of 2021.

Cite this claim

Gary Marcus, “Gary Marcus wrote the 2001 book The Algebraic Mind, and its predictions have so far held true.” (said 20 December 2021). Verdict: unresolvable. Reliability API, claim c_a386a09214ab, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_a386a09214ab

Claims without a verdict

Is a verdict here wrong, or do you speak for Gary Marcus? 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.