Pundit ClaimsAI claims by Gary Marcus and Ed Zitron, with the evidence

Changes, page 30 of 37

Every substantive change to the register, newest first. Speaker events record what the people covered said or changed. Site events record what this site published, corrected, flagged, or withdrew. The two never share a label.

Substantive changes (7383)

Page 30 of 37: changes 5801 to 6000 of 7383, newest first. Dates are when the change was recorded.

  1. Assessment publishedBy this siteGary Marcus

    People increasingly discouraging American children from learning to code, out of naive enthusiasm about AI, will leave the US flat-footed.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-decoding-and-debunking-hard-forks-3b46ff-a1.

  2. Assessment publishedBy this siteGary Marcus

    For at least another 10 to 20 years, and probably longer, big software projects will require competent human software architects.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-decoding-and-debunking-hard-forks-37519c-a1.

  3. Assessment publishedBy this siteGary Marcus

    Hype about AI coding will increasingly lead people to discourage American children from learning to code.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-decoding-and-debunking-hard-forks-29e0fa-a1.

  5. HeldBy this siteGary Marcus

    Held record marcus-decoding-and-debunking-hard-forks-25ea3a

    The openai screen reads the claim differently.

  6. Assessment publishedBy this siteGary Marcus

    Most or all AI coding systems were probably trained on existing fridge-photo recipe code such as Deep-food.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-decoding-and-debunking-hard-forks-22ec7e-a1.

  7. Assessment publishedBy this siteGary Marcus

    Hype about automated coding may well leave society with too few coders.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-decoding-and-debunking-hard-forks-11c45d-a1.

  10. Assessment publishedBy this siteGary Marcus

    The tribunal judge was right not to accept Air Canada's argument that its chatbot was a separate legal entity responsible for its own actions.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-death-of-a-chatbot-body-blow-to-an-ee0e15-a1.

  11. Assessment publishedBy this siteGary Marcus

    An Air Canada chatbot hallucinated an answer telling a customer that bereavement trips would be refunded.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-death-of-a-chatbot-body-blow-to-an-d61221-a1.

  12. Assessment publishedBy this siteGary Marcus

    The Air Canada chatbot ruling should be the top story in the tech industry.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-death-of-a-chatbot-body-blow-to-an-38965d-a1.

  14. HeldBy this siteGary Marcus

    Held record marcus-death-of-a-chatbot-body-blow-to-an-284f74

    The final-text check failed: citations_correct.

  15. Assessment publishedBy this siteGary Marcus

    The US should pursue AI collaboration with China on cancer and cybersecurity, as with the Soviet Union in the Cold War.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-president-trump-heres-a-deal-fc4e99-a1.

  16. Assessment publishedBy this siteGary Marcus

    If Trump makes a deal with China on the AI collaboration Marcus proposes (such as on cancer and cybersecurity), he really could win a Nobel Prize.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-president-trump-heres-a-deal-d2ebe0-a1.

  17. Assessment publishedBy this siteGary Marcus

    China is eager to collaborate with the US on AI, contrary to claims it would never want to.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-president-trump-heres-a-deal-c0cc6d-a1.

  18. Assessment publishedBy this siteGary Marcus

    Without an AI slowdown, really bad things might happen, and Trump would take much of the blame.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-president-trump-heres-a-deal-829718-a1.

  19. Assessment publishedBy this siteGary Marcus

    A slowdown in AI development might be a good idea, especially if the US does it now while still ahead.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-president-trump-heres-a-deal-57509f-a1.

  20. Assessment publishedBy this siteGary Marcus

    If the US adopted an AI slowdown now while ahead, it could lock in the US lead.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-president-trump-heres-a-deal-57509f-2-a1.

  21. Assessment publishedBy this siteGary Marcus

    Outliers remain a major AI problem with no known robust solution.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-ffdca9-a2.

  22. Assessment publishedBy this siteGary Marcus

    Deep learning lacks tools for verifiable code, and AGI by 2029 would probably require solving that soon.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-fcf450-a1.

  23. Assessment publishedBy this siteGary Marcus

    In 2029, AI will not be able to watch a movie and accurately describe what is going on.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-f025f6-a1.

  24. Assessment publishedBy this siteGary Marcus

    Tesla will not ship level 5 self-driving cars in 2022 or 2023.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-ed6b53-a1.

  25. Assessment publishedBy this siteGary Marcus

    The long tail problem makes it easy to believe AI problems are much easier than they actually are.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-d56a36-a1.

  28. Assessment publishedBy this siteGary Marcus

    In 2029, AI will not be able to work as a competent cook in an arbitrary kitchen.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-acced4-a1.

  29. Assessment publishedBy this siteGary Marcus

    The long-tail problem of rare events is a very serious issue for the currently popular AI techniques, which rely more on big data than on reasoning.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-a53d89-a1.

  30. Assessment publishedBy this siteGary Marcus

    3D perception and scene understanding remain unsolved for current AI.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-8818ae-a1.

  32. Assessment publishedBy this siteGary Marcus

    Deep learning is decent at object recognition but much worse at planning, reading and language comprehension.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-5dc69d-a1.

  33. HeldBy this siteGary Marcus

    Held record marcus-dear-elon-musk-here-are-five-things-55b9ea

    hedged language is not scored, even on a plain reading

  34. Assessment publishedBy this siteGary Marcus

    AI has no stable or trustworthy solutions for common sense, reasoning, language or analogy, and this has scarcely changed in about five years.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-536f05-a1.

  35. HeldBy this siteGary Marcus

    Held record marcus-dear-elon-musk-here-are-five-things-3a3261

    The openai review found the criteria unfair to the speaker's words.

  39. HeldBy this siteGary Marcus

    Held record marcus-dear-elon-musk-here-are-five-things-20b957

    The anthropic screen reads the claim differently.

  40. Assessment publishedBy this siteGary Marcus

    Unless the field invests in ideas beyond current techniques like deep learning, outliers alone may prevent reaching AGI before the end of the decade.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-1c4f0d-a1.

  44. Assessment publishedBy this siteGary Marcus

    Humanity lacks an adequate methodology for building complex cognitive systems.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dear-elon-musk-here-are-five-things-056704-a1.

  45. Assessment publishedBy this siteGary Marcus

    An amateur Go player, Kellin Pelrine, beat a top Go AI system 14 games to 1 in a 15-game series.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-david-beats-go-liath-b74f7d-a1.

  46. HeldBy this siteGary Marcus

    Held record marcus-david-beats-go-liath-b07505

    The anthropic review found the criteria unfair to the speaker's words.

  47. HeldBy this siteGary Marcus

    Held record marcus-david-beats-go-liath-9ae7e2

    The anthropic screen reads the claim differently.

  48. Assessment publishedBy this siteGary Marcus

    Adversarial attacks have repeatedly shown that deep learning systems have remarkable weaknesses and don't work the way people do.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-david-beats-go-liath-37d4e3-a2.

  49. HeldBy this siteGary Marcus

    Held record marcus-david-beats-go-liath-360fc7

    The openai review reads the claim differently.

  50. Assessment publishedBy this siteGary Marcus

    The past seven years of AI have shown that deep learning is unpredictable and not always human-like.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-david-beats-go-liath-2137d6-a1.

  51. Assessment publishedBy this siteGary Marcus

    However good deep-learning AI trained on immense data looks, we can never be sure it can extend what it knows to novel circumstances.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-david-beats-go-liath-17e697-a1.

  52. Assessment publishedBy this siteGary Marcus

    The AI industry may have spoiled its own prospects faster than any industry in history.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-data-center-madness-eaa8c2-a1.

  53. HeldBy this siteGary Marcus

    Held record marcus-data-center-madness-d4290d

    The anthropic screen reads the claim differently.

  54. Assessment publishedBy this siteGary Marcus

    AI data centers have become politically toxic.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-data-center-madness-d4290d-2-a2.

  55. Assessment publishedBy this siteGary Marcus

    Through a mixture of greed, stupidity, and arrogance, the AI industry figures who were the heroes of 2023 have become the villains of 2026.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-data-center-madness-cb879c-a1.

  56. HeldBy this siteGary Marcus

    Held record marcus-data-center-madness-942312

    The anthropic review found the criteria unfair to the speaker's words.

  57. HeldBy this siteGary Marcus

    Held record marcus-data-center-madness-712310

    The two reviews differ: anthropic historical_observation/not_applicable/supported, openai historical_observation/not_applicable/unresolved.

  58. Assessment publishedBy this siteGary Marcus

    There is essentially no public appetite for the AI data center buildout.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-data-center-madness-1a2f47-a2.

  59. HeldBy this siteGary Marcus

    Held record marcus-data-center-madness-1a2f47-2

    The openai screen reads the claim differently.

  60. Assessment publishedBy this siteGary Marcus

    The economics of AI data center capex relative to AI revenue do not add up.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-data-center-madness-02bfc2-a1.

  61. HeldBy this siteGary Marcus

    Held record marcus-dario-takes-it-on-the-chin-f8f19a

    The anthropic screen reads the claim differently.

  62. Assessment publishedBy this siteGary Marcus

    As bad as Sam Altman is, Dario Amodei is not all that different from him.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-takes-it-on-the-chin-dea9b2-a1.

  63. Assessment publishedBy this siteGary Marcus

    Most of the goodwill Amodei gained by standing up to Hegseth appears to have dissipated, as most goodwill toward Altman did.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-takes-it-on-the-chin-d82b9a-a1.

  64. HeldBy this siteGary Marcus

    Held record marcus-dario-takes-it-on-the-chin-d31692

    The two reviews differ: anthropic historical_observation/not_applicable/mixed, openai historical_observation/not_applicable/unresolved.

  65. HeldBy this siteGary Marcus

    Held record marcus-dario-takes-it-on-the-chin-c0dc51

    The anthropic screen doubts the attribution.

  66. Assessment publishedBy this siteGary Marcus

    It is far from obvious that either Altman or Amodei will regain the public's trust.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-takes-it-on-the-chin-ae65a4-a1.

  67. Assessment publishedBy this siteGary Marcus

    Because AI could be very damaging, AI needs leaders who can be fully trusted.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-takes-it-on-the-chin-9b1505-a1.

  68. Assessment publishedBy this siteGary Marcus

    Given how damaging AI could ultimately be, we need people we can fully trust, and Marcus does not think we have that.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-takes-it-on-the-chin-9b1505-2-a1.

  70. Assessment publishedBy this siteGary Marcus

    Many people now share the view that Amodei is not much different from Altman.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-takes-it-on-the-chin-0826bc-a1.

  71. Assessment publishedBy this siteGary Marcus

    We should not trust coding agents.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-fbb769-a1.

  72. HeldBy this siteGary Marcus

    Held record marcus-dario-amodei-hype-ai-safety-and-the-fa612d

    The final-text check failed; overclaims: e2 explanation: "Anthropic disclosed unintended actions, including form submissions despite instructions to stop before submission." The captured source says only that Claude submitted a sensitive form "when it should not have". It does not mention any instruction to stop before submission. The added detail makes the incident look more like a breach of explicit rules, which matters to this claim..

  73. HeldBy this siteGary Marcus

    Held record marcus-dario-amodei-hype-ai-safety-and-the-ec138a

    The anthropic screen reads the claim differently.

  74. HeldBy this siteGary Marcus

    Held record marcus-dario-amodei-hype-ai-safety-and-the-e86ab7

    e1 is dated 2026-04, which is not inside the period described (2026-04-01 to 2026-04-27); it can only be context

  75. Assessment publishedBy this siteGary Marcus

    Many more catastrophic vibe-coding failures involving privacy, security and data loss will occur soon.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-e162e5-a1.

  76. HeldBy this siteGary Marcus

    Held record marcus-dario-amodei-hype-ai-safety-and-the-da4084

    e1 is dated 2026-04, which is not inside the period described (2026-04-01 to 2026-04-27); it can only be context

  77. HeldBy this siteGary Marcus

    Held record marcus-dario-amodei-hype-ai-safety-and-the-ac6439

    The anthropic screen finds the claim checkable, so it must be graded.

  78. HeldBy this siteGary Marcus

    Held record marcus-dario-amodei-hype-ai-safety-and-the-94a2a8

    The openai review found the criteria unfair to the speaker's words.

  79. HeldBy this siteGary Marcus

    Held record marcus-dario-amodei-hype-ai-safety-and-the-7c8513

    e1 is dated 2026-04, which is not inside the period described (2026-04-01 to 2026-04-27); it can only be context

  80. Assessment publishedBy this siteGary Marcus

    AI coding tools are genuinely revolutionizing the software industry.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-687527-a1.

  81. Assessment publishedBy this siteGary Marcus

    People maybe shouldn't use AI coding tools without significant prior experience.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-687527-2-a1.

  82. Assessment publishedBy this siteGary Marcus

    Maintainability of AI-generated code is a serious issue and may in the long run be the most serious one.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-399a58-a1.

  83. Assessment publishedBy this siteGary Marcus

    Software engineers are still needed, and Amodei's claim that they will disappear is absurd.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-3682a0-a1.

  84. Assessment publishedBy this siteGary Marcus

    The viral vibe-coding data-loss failure was predictable because the system mimics data without truly understanding what was asked.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-34e7f5-a1.

  85. HeldBy this siteGary Marcus

    Held record marcus-dario-amodei-hype-ai-safety-and-the-153ca0

    e1 is dated 2026-04, which is not inside the period described (2026-04-01 to 2026-04-27); it can only be context

  86. Assessment publishedBy this siteGary Marcus

    Because coding agents can't be counted on for expert knowledge, software engineers need to be kept in the loop.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-1421b5-a1.

  87. Assessment publishedBy this siteGary Marcus

    AI agents are wildly premature technology that is being rolled out way too fast.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dario-amodei-hype-ai-safety-and-the-0ae5fb-a1.

  88. Assessment publishedBy this siteGary Marcus

    Copyright litigation over generative AI image and text outputs will be wild.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dall-es-new-guardrails-fast-furious-dfa54e-a1.

  89. HeldBy this siteGary Marcus

    Held record marcus-dall-es-new-guardrails-fast-furious-ddb599

    The anthropic screen reads the claim differently.

  90. Assessment publishedBy this siteGary Marcus

    Guardrails sometimes end up too permissive, and this is still happening over a year into the chatbot revolution.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dall-es-new-guardrails-fast-furious-b04cd9-a1.

  91. HeldBy this siteGary Marcus

    Held record marcus-dall-es-new-guardrails-fast-furious-acc522

    The openai screen doubts the attribution.

  92. Assessment publishedBy this siteGary Marcus

    The UK will not grant OpenAI the exemption from copyright responsibility it requested.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dall-es-new-guardrails-fast-furious-9c4bb1-a1.

  93. Assessment publishedBy this siteGary Marcus

    Even after the new filters, some prompts to Bing's DALL-E 3 still produced apparent copyright infringements.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dall-es-new-guardrails-fast-furious-9b2ca8-a1.

  94. Assessment publishedBy this siteGary Marcus

    Guardrail builders generally end up patching individual failures one by one instead of solving the problem in general.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dall-es-new-guardrails-fast-furious-8dabb1-a1.

  96. HeldBy this siteGary Marcus

    Held record marcus-dall-es-new-guardrails-fast-furious-6241ce

    The final-text check failed: rationale_supported_by_evidence; overclaims: The post's screenshots are images the reviewer could not read, so their content was not checked..

  97. HeldBy this siteGary Marcus

    Held record marcus-dall-es-new-guardrails-fast-furious-3f7689

    hedged language is not scored, even on a plain reading

  98. HeldBy this siteGary Marcus

    Held record marcus-dall-es-new-guardrails-fast-furious-37170d

    The openai review found the criteria unfair to the speaker's words.

  99. Assessment publishedBy this siteGary Marcus

    No AI guardrail so far has effectively balanced being neither too restrictive nor too permissive.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dall-es-new-guardrails-fast-furious-2b243e-a1.

  100. Assessment publishedBy this siteGary Marcus

    Bing falsely said its apparently infringing images were original.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-dall-es-new-guardrails-fast-furious-059e57-a2.

  101. Assessment publishedBy this siteGary Marcus

    There is a substantial risk that AI will cause chaos.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-fdad08-a1.

  102. Assessment publishedBy this siteGary Marcus

    Marcus says he is not all that worried about humans going extinct anytime very soon.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-f6c48c-a1.

  103. Assessment publishedBy this siteGary Marcus

    Documented risks of current AI such as bias, defamation, cybercrime and disinformation may escalate quickly and could cause geopolitical instability within the next few years.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-f2c25c-a1.

  104. Assessment publishedBy this siteGary Marcus

    There is no solid argument that AI doom risk is zero or even infinitesimal, though some arguments by Yudkowsky and Bostrom may be weak.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-ecc4a5-a1.

  105. Assessment publishedBy this siteGary Marcus

    Yudkowsky's claim that AGI will kill everyone by default is far from proven; it is possible but not shown to be the default.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-ea059a-a1.

  107. Assessment publishedBy this siteGary Marcus

    Not being able to predict technology precisely does not excuse us from the moral responsibility to take precautions against its risks.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-d4c380-a1.

  108. Assessment publishedBy this siteGary Marcus

    Imprecise risk estimates do not justify failing to act on AI risk.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-c57e84-a1.

  109. HeldBy this siteGary Marcus

    Held record marcus-d28-bc06c4

    The openai screen doubts the attribution.

  111. Assessment publishedBy this siteGary Marcus

    AI can bring catastrophe without mass death, and society is heading in that direction.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-b225ae-a1.

  112. Assessment publishedBy this siteGary Marcus

    LeCun has given no real argument that humans will always set the agenda over AI, though that conclusion might be true.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-a90faa-a1.

  114. Assessment publishedBy this siteGary Marcus

    A large fraction of the AI and AI-adjacent community does not want to discuss AI doom or catastrophe risk at all.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-9dbc35-a2.

  115. Assessment publishedBy this siteGary Marcus

    Perhaps extinction is vanishingly unlikely, but the chance of an incident that kills, say, one percent or more of the population seems a lot higher.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-77175d-a1.

  116. Assessment publishedBy this siteGary Marcus

    Whatever one thinks about doom from superintelligence, people should worry about the probability of catastrophe given widespread AI adoption.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-757255-a1.

  117. Assessment publishedBy this siteGary Marcus

    No one will be able to calculate p(doom) precisely, to three decimal places.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-702ab0-a1.

  118. Assessment publishedBy this siteGary Marcus

    We are not close to super-smart machines.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-639a26-a1.

  119. Assessment publishedBy this siteGary Marcus

    It is worth getting a rough estimate of p(doom) even if there is no consensus on the exact number.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-4692ed-a1.

  120. Assessment publishedBy this siteGary Marcus

    AI doom probabilities are not high or imminent enough to justify bombing data centers any time soon, and probably never will be.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-41b59e-a1.

  121. Assessment publishedBy this siteGary Marcus

    Marcus says that, honestly, we haven't a clue whether super-smart machines, if and when we get there, will be dominating or non-dominating.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-3d7576-a1.

  122. Assessment publishedBy this siteGary Marcus

    Meta is poor at making LLMs that do not hallucinate.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-267622-a1.

  123. Assessment publishedBy this siteGary Marcus

    Marcus says Gebru and Birhane are very likely right that there is a good chance any immediate effects of AI will disproportionately affect the dispossessed.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-249663-a1.

  125. Assessment publishedBy this siteGary Marcus

    To the extent AI carries any real risk of doom or catastrophe, society should not wait too long to address it, unlike with climate change.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-d28-209eb4-a1.

  126. HeldBy this siteGary Marcus

    Held record marcus-d28-076e39

    e2 is dated 2023, which is not inside the period described (2023-01-01 to 2023-08-27); it can only be context

  127. Assessment publishedBy this siteGary Marcus

    Open-source AI might be used to design bioweapons.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-critical-national-security-questions-e7bfa9-a1.

  129. Assessment publishedBy this siteGary Marcus

    Whether to open-source powerful AI should be decided through careful public discussion rather than by a few tech leaders.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-critical-national-security-questions-ab67b8-a1.

  130. HeldBy this siteGary Marcus

    Held record marcus-critical-national-security-questions-9d30fe

    The openai screen reads the claim differently.

  131. Assessment publishedBy this siteGary Marcus

    Decisions about open-releasing powerful AI should not be entirely up to Meta's leadership, though currently they are.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-critical-national-security-questions-83e831-a1.

  133. HeldBy this siteGary Marcus

    Held record marcus-critical-national-security-questions-564368

    The final-text check failed: proposition_faithful.

  134. Assessment publishedBy this siteGary Marcus

    We are already partway toward a technopolar world where Big Tech companies rather than nations run the world.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-critical-national-security-questions-50d62c-a1.

  136. Assessment publishedBy this siteGary Marcus

    The draft EU AI Act is in serious jeopardy as of November 2023.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-crisis-point-in-ai-56514c-a2.

  137. HeldBy this siteGary Marcus

    Held record marcus-crisis-point-in-ai-56514c-3

    The anthropic review found the criteria unfair to the speaker's words.

  138. Assessment publishedBy this siteGary Marcus

    The jeopardy facing the draft EU AI Act may be a direct result of regulatory capture, with governments doing the bidding of tech companies.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-crisis-point-in-ai-56514c-2-a1.

  139. Assessment publishedBy this siteGary Marcus

    LLM companies should recall their systems until they find an adequate solution to covert racism.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-covert-racism-in-llms-fb7f13-a1.

  140. HeldBy this siteGary Marcus

    Held record marcus-covert-racism-in-llms-ebba76

    hedged language is not scored, even on a plain reading

  141. HeldBy this siteGary Marcus

    Held record marcus-covert-racism-in-llms-dddc08

    The anthropic review found the criteria unfair to the speaker's words.

  142. HeldBy this siteGary Marcus

    Held record marcus-covert-racism-in-llms-c924a0

    The anthropic review found the criteria unfair to the speaker's words.

  143. Assessment publishedBy this siteGary Marcus

    LLMs tested looked fine on overt racism measures but did very badly on covert racism measures when prompted with African American English.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-covert-racism-in-llms-b36446-a2.

  144. Assessment publishedBy this siteGary Marcus

    Because LLMs are so widely adopted, the potential consequences of their covert racism for the world are massive.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-covert-racism-in-llms-abf119-a1.

  145. Assessment publishedBy this siteGary Marcus

    Congress, the EEOC, HUD, the FTC and others should investigate covert racism in LLMs and demand user logs and interaction data from the major LLM makers.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-covert-racism-in-llms-9807cf-a1.

  146. Assessment publishedBy this siteGary Marcus

    Google's recent debacle shows that building guardrails for LLMs is never easy.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-covert-racism-in-llms-8f13f0-a1.

  147. Assessment publishedBy this siteGary Marcus

    The covert racism documented in LLMs may already be having real-world impact.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-covert-racism-in-llms-636e14-a1.

  148. Assessment publishedBy this siteGary Marcus

    LLMs cannot understand concepts like people and race, and their reliance on superficial statistics causes racist stereotyping.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-covert-racism-in-llms-39ce7c-a1.

  149. HeldBy this siteGary Marcus

    Held record marcus-covert-racism-in-llms-30aa57

    The two reviews differ: anthropic causal_assessment/not_applicable/supported, openai causal_assessment/not_applicable/unresolved.

  150. Assessment publishedBy this siteGary Marcus

    Regulators outside the US should consider investigating LLM makers and demanding user logs.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-covert-racism-in-llms-0befd3-a1.

  152. Assessment publishedBy this siteGary Marcus

    If neither AI nor AI companies are controlled, society's chances of avoiding serious consequences are poor.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-we-all-be-doomed-without-ranked-af4ab1-a1.

  153. Assessment publishedBy this siteGary Marcus

    Marcus thinks ranked-choice voting, or some similar alternative, is very much worth serious consideration.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-we-all-be-doomed-without-ranked-9d0b71-a1.

  154. Assessment publishedBy this siteGary Marcus

    By default, AI companies will write the rules for AI regulation and governments will go along.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-we-all-be-doomed-without-ranked-8b5ae0-a1.

  155. Assessment publishedBy this siteGary Marcus

    US AI policy so far has mostly been voluntary guidelines the companies are comfortable with.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-we-all-be-doomed-without-ranked-60d406-a1.

  156. Assessment publishedBy this siteGary Marcus

    AI companies aim at returns to shareholders rather than benefit to humanity.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-we-all-be-doomed-without-ranked-56d8eb-a1.

  157. Assessment publishedBy this siteGary Marcus

    An AI-induced extinction event that kills everyone is unlikely, though not ruled out.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-we-all-be-doomed-without-ranked-558347-a1.

  158. Assessment publishedBy this siteGary Marcus

    There has been no tangible progress in getting big AI companies to commit to disclosing their training data.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-we-all-be-doomed-without-ranked-3eea03-a1.

  159. Assessment publishedBy this siteGary Marcus

    Government must step up to achieve an AI regulatory regime that protects consumers without stifling innovation.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-we-all-be-doomed-without-ranked-33dc44-a1.

  160. HeldBy this siteGary Marcus

    Held record marcus-could-we-all-be-doomed-without-ranked-28dffb

    The anthropic screen doubts the attribution.

  162. Assessment publishedBy this siteGary Marcus

    ChatGPT is notorious for making illegal moves in chess.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-gpt-5-revolutionize-military-f7c3fb-a1.

  163. HeldBy this siteGary Marcus

    Held record marcus-could-gpt-5-revolutionize-military-ed7862

    e2 is dated 2023-03, which is not inside the period described (2023-03-14 to 2024-02-07); it can only be context

  164. HeldBy this siteGary Marcus

    Held record marcus-could-gpt-5-revolutionize-military-ac61f1

    The anthropic review reads the claim differently.

  165. Assessment publishedBy this siteGary Marcus

    Generative AI systems should not be allowed to run a war; doing so would be malpractice.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-gpt-5-revolutionize-military-a6856c-a1.

  166. Assessment publishedBy this siteGary Marcus

    Developers and military personnel should not let chatbots take on the role of military generals.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-gpt-5-revolutionize-military-971235-a1.

  167. Assessment publishedBy this siteGary Marcus

    GPT-4 plays chess at about the level of top chess programs in 1976, far from the current state of the art.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-gpt-5-revolutionize-military-846b37-a2.

  169. HeldBy this siteGary Marcus

    Held record marcus-could-gpt-5-revolutionize-military-3f1d9b

    The anthropic screen reads the claim differently.

  170. Assessment publishedBy this siteGary Marcus

    Generative AI systems continue to be unstable in virtually everything they do.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-gpt-5-revolutionize-military-3723d1-a1.

  172. Assessment publishedBy this siteGary Marcus

    GPT-5 will not convey an immense advantage in military strategy.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-gpt-5-revolutionize-military-1d0963-a1.

  173. Assessment publishedBy this siteGary Marcus

    If the public unites, it might be possible to steer AI to a better place.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-generative-ai-could-turn-out-eb575b-a1.

  174. Assessment publishedBy this siteGary Marcus

    Marcus thinks the multitrillion-dollar investment in AI could turn out to be another epic, arrogance-fueled mistake like the Vietnam war.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-generative-ai-could-turn-out-e82bc9-a1.

  175. Assessment publishedBy this siteGary Marcus

    Most AI pilot studies show no return on investment, and AI has ongoing reliability problems.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-generative-ai-could-turn-out-e82bc9-2-a2.

  176. HeldBy this siteGary Marcus

    Held record marcus-could-generative-ai-could-turn-out-d03693

    the result turns on our interpretation rather than the speaker's words, so it cannot be scored without a stated plain reading

  177. Assessment publishedBy this siteGary Marcus

    If public backlash against AI is strong enough, Trump may move further away from his 2025 anti-regulatory AI policies.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-generative-ai-could-turn-out-c9c436-a1.

  178. HeldBy this siteGary Marcus

    Held record marcus-could-generative-ai-could-turn-out-ac95fa

    The anthropic screen reads the claim differently.

  179. Assessment publishedBy this siteGary Marcus

    Public backlash against AI is growing fast as of May 2026.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-generative-ai-could-turn-out-abf6e0-a2.

  180. Assessment publishedBy this siteGary Marcus

    Trump's forecast repositioning away from pro-AI-industry policy may have begun by May 2026.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-generative-ai-could-turn-out-8b5b70-a1.

  181. Assessment publishedBy this siteGary Marcus

    By the end of 2026, President Trump will have begun distancing himself from the aggressively pro-AI-industry policies of his 2025 AI strategy.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-generative-ai-could-turn-out-5daad5-a1.

  182. HeldBy this siteGary Marcus

    Held record marcus-could-generative-ai-could-turn-out-0523d4

    The anthropic review found the criteria unfair to the speaker's words.

  183. HeldBy this siteGary Marcus

    Held record marcus-could-cruise-be-the-theranos-of-ai-c851bb

    The openai review found the criteria unfair to the speaker's words.

  184. HeldBy this siteGary Marcus

    Held record marcus-could-cruise-be-the-theranos-of-ai-c6077c

    The anthropic review found the criteria unfair to the speaker's words.

  185. Assessment publishedBy this siteGary Marcus

    Certainly no commercial product could ever work at a profit if it needed remote operators anything like as often as Cruise reportedly does.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-cruise-be-the-theranos-of-ai-8d7fd0-a2.

  186. Assessment publishedBy this siteGary Marcus

    California should immediately demand that driverless car companies disclose their remote operator use and make the answers public.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-cruise-be-the-theranos-of-ai-690d99-a1.

  187. Assessment publishedBy this siteGary Marcus

    Driverless car companies in general have disclosed almost nothing publicly about their use of remote operators.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-cruise-be-the-theranos-of-ai-200bf8-a1.

  188. Assessment publishedBy this siteGary Marcus

    Shareholders will sue over Cruise.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-cruise-be-the-theranos-of-ai-187c3b-a1.

  189. Assessment publishedBy this siteGary Marcus

    If the situation is as bad as it looks, GM probably will not continue the Cruise driverless project.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-cruise-be-the-theranos-of-ai-187c3b-2-a1.

  190. Assessment publishedBy this siteGary Marcus

    Cruise's CEO essentially confirmed that Cruise's driverless cars require very regular human intervention.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-cruise-be-the-theranos-of-ai-0d3a17-a1.

  191. Assessment publishedBy this siteGary Marcus

    It increasingly appears that Cruise cars would perform disastrously if left to operate without remote human help.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-cruise-be-the-theranos-of-ai-08b318-a1.

  192. Assessment publishedBy this siteGary Marcus

    There is no robust solution to LLM hallucinations.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-china-devastate-the-us-without-e898be-a1.

  193. Assessment publishedBy this siteGary Marcus

    The AI industry has entered a perhaps unprecedented era of extreme financial interdependence and circular investment.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-china-devastate-the-us-without-d15aa1-a1.

  194. Assessment publishedBy this siteGary Marcus

    GPT-5 was disappointing and was not released in 2024.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-china-devastate-the-us-without-ca05dd-a1.

  195. HeldBy this siteGary Marcus

    Held record marcus-could-china-devastate-the-us-without-b3a50c

    The anthropic review reads the claim differently.

  197. Assessment publishedBy this siteGary Marcus

    Most LLM developers are losing money.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-china-devastate-the-us-without-ae344c-a1.

  198. HeldBy this siteGary Marcus

    Held record marcus-could-china-devastate-the-us-without-a43222

    The anthropic screen reads the claim differently.

  199. Assessment publishedBy this siteGary Marcus

    Investments in the single LLM/transformer approach continue to grow.

    First assessment published after both reviews and the final-text check passed.

    Assessment marcus-could-china-devastate-the-us-without-6efa26-a2.