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

Changes, page 28 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 28 of 37: changes 5401 to 5600 of 7383, newest first. Dates are when the change was recorded.

  2. Assessment publishedBy this siteGary Marcus

    Research should have been front and center in the Klein–Buchanan episode, at a time when Trump is dismantling American research infrastructure.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-f2426a-a1.

  3. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-ece545

    The anthropic review does not accept e1 as contradicts.

  5. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-dcc9b7

    The anthropic review reads the claim differently.

  7. Assessment publishedBy this siteGary Marcus

    AI will not replace the best humans, such as top scientists, within a few years, though it might replace average people in some tasks soon.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-cfdbec-a1.

  8. Assessment publishedBy this siteGary Marcus

    Because there is no reason to think any of the remaining problems are logically unsolvable in principle, AGI will be achieved someday.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-cfd348-a1.

  9. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-bd234e

    The anthropic screen doubts the attribution.

  10. Assessment publishedBy this siteGary Marcus

    Complex systems built on LLMs, such as deep research agents, will prove massively difficult to debug.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-b1c673-a1.

  11. Assessment publishedBy this siteGary Marcus

    Marcus is not convinced that anything anyone could realistically do with LLMs won't be copied almost immediately, as has been the recent trend.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-ad6bab-a1.

  12. Assessment publishedBy this siteGary Marcus

    Automatically generated bullshit still seems to be the norm, including in systems like Deep Research and Grok 3 Deep Search.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-ab240f-a1.

  13. Assessment publishedBy this siteGary Marcus

    We have to consider the scenario in which the proposition that AGI is imminent is false.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-94e93b-a1.

  14. Assessment publishedBy this siteGary Marcus

    GPT-4.5 received very mixed reviews and is at most a minor, incremental step toward AGI, despite years of work and billions invested.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-9478c3-a1.

  15. Assessment publishedBy this siteGary Marcus

    LLMs including o1-mini fail to generalize on large multi-digit multiplication despite immense training data.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-92e534-a2.

  16. Assessment publishedBy this siteGary Marcus

    The most plausible scenario by far is that getting to AGI will be a slog: some problems get solved in the next 2-3 years and others take longer.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-8ff48a-a1.

  17. Assessment publishedBy this siteGary Marcus

    The hypothesis that AGI is imminent is getting play in high places, and it is absurd to pretend otherwise.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-889674-a1.

  18. Assessment publishedBy this siteGary Marcus

    Most parts of the US still lack functioning driverless car systems after more than a decade of lavish promises.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-87597a-a1.

  19. Assessment publishedBy this siteGary Marcus

    With its academic research institutions being dismantled, the US may be all but surrendering the AGI race to China.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-796301-a1.

  20. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-792f50

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

  21. Assessment publishedBy this siteGary Marcus

    When AGI arrives, it will matter a lot.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-6ac6e3-a1.

  22. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-69c650

    a observation claim cannot be assessed on the causal_assessment axis

  23. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-67efec

    The final-text check failed: rationale_supported_by_evidence; overclaims: It documented a tenfold output-token price premium over Claude 3.7 Sonnet, supporting a very large API price premium during the February 27–March 5 assessment period [e1]. | The tenfold comparison applies to output-token pricing, not input tokens or completed tasks..

  24. Assessment publishedBy this siteGary Marcus

    Future models that keep scaling the same way might be prohibitively expensive, especially since nobody has a moat and price wars are intensifying.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-67efec-2-a1.

  25. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-5c8dbf

    The anthropic review does not accept e3 as supports.

  26. Assessment publishedBy this siteGary Marcus

    An AI system this unreliable on arithmetic will be similarly unreliable on everything else and should not be given society-changing power.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-5c1dd9-a1.

  27. Assessment publishedBy this siteGary Marcus

    Grok 3 was 15 times the size of Grok 2 and was generally viewed as disappointing.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-58ea33-a1.

  28. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-4c0be2

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

  29. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-4445cb

    the horizon on 2026-10-28 is open_horizon, not not_due

  32. Assessment publishedBy this siteGary Marcus

    Claims like Klein's that AGI is imminent are motivating many dodgy bets that may not make sense.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-2ee494-a1.

  33. Assessment publishedBy this siteGary Marcus

    LLMs still lack a reliable ability to reason over abstractions, as their ongoing unpredictable errors show.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-236597-a2.

  34. Assessment publishedBy this siteGary Marcus

    Hallucinations have not gone away, even in the latest models.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-21d48a-a1.

  35. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-2015b9

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

  36. Assessment publishedBy this siteGary Marcus

    The long-standing core AGI problems probably will not all be solved quickly, even with half a trillion dollars of funding.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-1fe025-a1.

  37. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-1eb181

    The anthropic screen doubts the attribution.

  38. Assessment publishedBy this siteGary Marcus

    OpenAI's planned GPT-5, relabeled GPT-4.5, failed despite more than two years of hard and expensive work.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-1dcb61-a1.

  39. HeldBy this siteGary Marcus

    Held record marcus-ezra-kleins-new-take-on-agi-and-why-1d846e

    The openai review reads the claim differently.

  41. Assessment publishedBy this siteGary Marcus

    Hallucinations have not been solved despite constant assurances that they would be.

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

    Assessment marcus-ezra-kleins-new-take-on-agi-and-why-0b5f36-a2.

  42. Assessment publishedBy this siteGary Marcus

    MMLU gains from GPT-4 to GPT-4 Turbo are small compared with earlier generational jumps, tentatively evidencing diminishing returns.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-f2cb8e-a2.

  43. HeldBy this siteGary Marcus

    Held record marcus-evidence-that-llms-are-reaching-a-f1a676

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

  45. Assessment publishedBy this siteGary Marcus

    The AI market may well be in for a correction.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-e4a13f-a1.

  46. Assessment publishedBy this siteGary Marcus

    In recent weeks Inflection AI largely closed, Stability AI is struggling, Ghost closed, and serious questions were raised about Devin.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-d3b9f2-a1.

  47. Assessment publishedBy this siteGary Marcus

    If his graphs are roughly correct, LLM capabilities are no longer doubling every 5 to 14 months.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-c5116b-a1.

  48. Assessment publishedBy this siteGary Marcus

    Any doubling of LLM capabilities has perhaps come to an end.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-a9e671-a1.

  51. Assessment publishedBy this siteGary Marcus

    LLMs' role may be smaller than hoped, and LLMs alone may never live up to 2023's wild expectations.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-8d11a5-a1.

  53. Assessment publishedBy this siteGary Marcus

    Multiple models released since GPT-4 have converged on GPT-4-level performance with none decisively ahead.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-7f3d52-a1.

  54. Assessment publishedBy this siteGary Marcus

    Hallucinations and stupid errors in LLMs have not been solved; GPT-4 Turbo still makes them.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-7cbffe-a1.

  55. Assessment publishedBy this siteGary Marcus

    LLM progress may be approaching the wall Marcus warned about in 2022.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-7c1946-a1.

  56. Assessment publishedBy this siteGary Marcus

    The MMLU plateau is likely a genuine sign of diminishing returns rather than a ceiling effect.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-65b597-a1.

  57. HeldBy this siteGary Marcus

    Held record marcus-evidence-that-llms-are-reaching-a-60cf63

    The anthropic review does not accept e1 as supports.

  58. Assessment publishedBy this siteGary Marcus

    Marcus suspects GPT-4 Turbo was a failed attempt at building GPT-5.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-5f20c2-a1.

  59. Assessment publishedBy this siteGary Marcus

    About 5 to 7 recent models are on par with GPT-4, but none is clearly ahead.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-5d187d-a1.

  60. Assessment publishedBy this siteGary Marcus

    OpenAI's GPT-4 Turbo graph shows only mostly modest improvements over recent months.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-570f6e-a1.

  61. HeldBy this siteGary Marcus

    Held record marcus-evidence-that-llms-are-reaching-a-494548

    The final-text check failed: proposition_faithful, citations_correct; overclaims: LLM reliability problems clearly have not been solved, and this is becoming widely known..

  62. Assessment publishedBy this siteGary Marcus

    The study Mollick cited does not show that LLM capabilities are improving; it only addresses compute efficiency.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-458f77-a1.

  63. HeldBy this siteGary Marcus

    Held record marcus-evidence-that-llms-are-reaching-a-452931

    The final-text check failed: proposition_faithful.

  64. Assessment publishedBy this siteGary Marcus

    In the most extreme case, OpenAI's $86 billion valuation could in hindsight look like a WeWork moment for AI.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-1954b3-a1.

  65. Assessment publishedBy this siteGary Marcus

    Reliable, trustworthy AI is surely achievable, but we may need to go back to the drawing board to get there.

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

    Assessment marcus-evidence-that-llms-are-reaching-a-022b91-a1.

  66. Assessment publishedBy this siteGary Marcus

    Unless citizens speak up very loudly, AI will not come to work for the public.

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

    Assessment marcus-everything-i-warned-about-in-taming-fe5bcc-a1.

  67. Assessment publishedBy this siteGary Marcus

    At present, AI is not working for the public.

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

    Assessment marcus-everything-i-warned-about-in-taming-faebb3-a1.

  68. HeldBy this siteGary Marcus

    Held record marcus-everything-i-warned-about-in-taming-d75e63

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

  69. Assessment publishedBy this siteGary Marcus

    The new administration will do everything in its power to protect businesses, and nothing to protect individuals.

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

    Assessment marcus-everything-i-warned-about-in-taming-c7ab0d-a2.

  70. Assessment publishedBy this siteGary Marcus

    We should worry about authoritarian uses of AI.

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

    Assessment marcus-everything-i-warned-about-in-taming-b4f830-a1.

  71. HeldBy this siteGary Marcus

    Held record marcus-everything-i-warned-about-in-taming-924c28

    The final-text check failed: proposition_faithful; overclaims: e2 explanation: "Vance also promised worker participation, criticized incumbent companies seeking advantageous regulation, and retained a role for safety." The e2 excerpt and its context support only the worker promises. They do not show criticism of incumbents seeking advantageous regulation or a retained role for safety..

  72. Assessment publishedBy this siteGary Marcus

    US possession of massive LLMs will do nothing to stop China from using its LLMs for authoritarian purposes.

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

    Assessment marcus-everything-i-warned-about-in-taming-8ea8f6-a1.

  74. Assessment publishedBy this siteGary Marcus

    If AI has significant negative externalities upon the world, we the citizens are screwed.

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

    Assessment marcus-everything-i-warned-about-in-taming-07de7e-a1.

  76. HeldBy this siteGary Marcus

    Held record marcus-everyone-is-cheating-their-way-through-db39cb

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

  77. Assessment publishedBy this siteGary Marcus

    The creators and distributors of GenAI have brought this huge problem, GenAI undermining college education and evaluation, to society.

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

    Assessment marcus-everyone-is-cheating-their-way-through-c4b1dd-a1.

  78. Assessment publishedBy this siteGary Marcus

    The current university education model no longer works now that GenAI exists.

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

    Assessment marcus-everyone-is-cheating-their-way-through-abdf42-a1.

  79. HeldBy this siteGary Marcus

    Held record marcus-everyone-is-cheating-their-way-through-97442e

    The final-text check failed: citations_correct.

  80. HeldBy this siteGary Marcus

    Held record marcus-everyone-is-cheating-their-way-through-93f29c

    The anthropic screen reads the claim differently.

  81. Assessment publishedBy this siteGary Marcus

    University curricula need to be redesigned because of GenAI.

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

    Assessment marcus-everyone-is-cheating-their-way-through-9255ec-a1.

  82. Assessment publishedBy this siteGary Marcus

    GenAI has destroyed term papers' role in giving writing practice, prompting reflection and evaluating students at scale.

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

    Assessment marcus-everyone-is-cheating-their-way-through-5ff51f-a1.

  84. Assessment publishedBy this siteGary Marcus

    With university budgets being cut, few universities will have enough money to replace term papers with small TA-led sections.

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

    Assessment marcus-everyone-is-cheating-their-way-through-4add9d-a1.

  85. Assessment publishedBy this siteGary Marcus

    Education will need a redesign more thorough than one-off assignments like having students find hallucinations in LLM-generated papers.

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

    Assessment marcus-everyone-is-cheating-their-way-through-21e6a3-a1.

  86. Assessment publishedBy this siteGary Marcus

    Negotiations (over Sam Altman's return to OpenAI) broke down altogether, which Marcus says is perhaps in keeping with his surmise that Altman's bluff was called.

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

    Assessment marcus-everybodys-out-a65-7e5da1-a1.

  87. Assessment publishedBy this siteGary Marcus

    A lot of OpenAI employees will probably leave following the breakdown of negotiations to reinstate Sam Altman.

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

    Assessment marcus-everybodys-out-a65-4fe2f8-a1.

  88. Assessment publishedBy this siteGary Marcus

    The secondary share sale of OpenAI to Thrive at an $86B valuation is presumably unlikely to close.

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

    Assessment marcus-everybodys-out-a65-3b211a-a1.

  89. HeldBy this siteGary Marcus

    Held record marcus-everybodys-out-a65-33401f

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

  90. Assessment publishedBy this siteGary Marcus

    Claude Code is not AGI.

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

    Assessment marcus-even-more-good-news-for-the-future-f80f95-a1.

  93. Assessment publishedBy this siteGary Marcus

    The Tufts paper strongly vindicates the Apple reasoning paper's findings.

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

    Assessment marcus-even-more-good-news-for-the-future-d1a541-a1.

  94. Assessment publishedBy this siteGary Marcus

    The neurosymbolic hybrid used nearly two orders of magnitude less energy than the VLAs in the Tufts study.

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

    Assessment marcus-even-more-good-news-for-the-future-bd8edb-a1.

  95. HeldBy this siteGary Marcus

    Held record marcus-even-more-good-news-for-the-future-b5a2ee

    The final-text check failed: citations_correct.

  97. Assessment publishedBy this siteGary Marcus

    LLMs are efficient for imperfect pattern recognition but inefficient for reasoning out a plan.

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

    Assessment marcus-even-more-good-news-for-the-future-931695-a1.

  98. Assessment publishedBy this siteGary Marcus

    Evidence increasingly suggests neurosymbolic AI is a stronger starting point than pure LLMs, with potential for better generalization and efficiency.

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

    Assessment marcus-even-more-good-news-for-the-future-6cd8dd-a1.

  99. Assessment publishedBy this siteGary Marcus

    LLMs have persistent weaknesses in planning, reasoning and generalization.

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

    Assessment marcus-even-more-good-news-for-the-future-698865-a1.

  100. HeldBy this siteGary Marcus

    Held record marcus-even-more-good-news-for-the-future-5a5db5

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

  101. Assessment publishedBy this siteGary Marcus

    A neurosymbolic hybrid with a symbolic planner generalizes far better than VLAs on the Tower of Hanoi task.

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

    Assessment marcus-even-more-good-news-for-the-future-46baef-a1.

  102. Assessment publishedBy this siteGary Marcus

    The replies to the Apple reasoning paper were weak.

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

    Assessment marcus-even-more-good-news-for-the-future-2df921-a1.

  103. HeldBy this siteGary Marcus

    Held record marcus-even-more-good-news-for-the-future-055f7c

    The anthropic review reads the claim differently.

  105. Assessment publishedBy this siteGary Marcus

    Specialized, domain-specific AI might help with climate change, but it is not the kind getting massive funding or causing massive harm.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-d4690a-a1.

  106. Assessment publishedBy this siteGary Marcus

    Massive bets on AI's environmental costs should not be made without transparency and accountability.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-c82c3a-a1.

  107. Assessment publishedBy this siteGary Marcus

    Independent scientists should assess the technologies, odds of success, and costs of major AI bets before they are made.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-c15414-a1.

  109. HeldBy this siteGary Marcus

    Held record marcus-eric-schmidts-risky-bet-on-ai-and-ae0621

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

  110. Assessment publishedBy this siteGary Marcus

    Marcus hopes we will think extremely carefully before going literally all in with the Earth on an unproven technology (AI).

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-a18547-a1.

  111. Assessment publishedBy this siteGary Marcus

    The case that AI will do serious harm to the environment if we continue on the current path is much stronger than the case for AI's putative benefits.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-8d49b5-a1.

  112. HeldBy this siteGary Marcus

    Held record marcus-eric-schmidts-risky-bet-on-ai-and-825771

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

  113. Assessment publishedBy this siteGary Marcus

    Decisions on AI's environmental bet should not be left to billionaires or profit-driven companies; society must have a vote.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-62983a-a1.

  114. Assessment publishedBy this siteGary Marcus

    We should not risk the planet for chatbots and video synthesis systems.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-558f36-a1.

  115. Assessment publishedBy this siteGary Marcus

    Schmidt's large stake in AI companies creates potential conflicts of interest that should be taken seriously when weighing his views on AI and climate.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-526269-a1.

  116. Assessment publishedBy this siteGary Marcus

    All the major potential benefits of AI are quite speculative.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-3bdc57-a2.

  117. Assessment publishedBy this siteGary Marcus

    For now, Schmidt's assumption that AI will eventually be worth essentially arbitrary environmental costs seems very premature.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-32a70a-a1.

  119. Assessment publishedBy this siteGary Marcus

    Almost every claimed benefit of AI, other than enriching a few people, has so far been only a promise.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-11de58-a1.

  121. Assessment publishedBy this siteGary Marcus

    It might make more sense to divert funding from chatbots and video synthesis to special-purpose AI addressing climate change.

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

    Assessment marcus-eric-schmidts-risky-bet-on-ai-and-035fac-a1.

  122. Assessment publishedBy this siteGary Marcus

    Bubeck's Erdős claim, which would have been extraordinary, should have been closely vetted.

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

    Assessment marcus-erdosgate-ceb515-a1.

  123. Assessment publishedBy this siteGary Marcus

    In 2019 OpenAI's claim that its robot 'solved' the Rubik's cube was misleading, and OpenAI refused to correct it.

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

    Assessment marcus-erdosgate-a7e202-a1.

  124. Assessment publishedBy this siteGary Marcus

    GPT-5 did not solve open Erdős problems; it essentially located existing solutions to already-solved problems on the web.

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

    Assessment marcus-erdosgate-a597e8-a1.

  125. HeldBy this siteGary Marcus

    Held record marcus-erdosgate-6dcda6

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

  126. Assessment publishedBy this siteGary Marcus

    'Some things never change': a closing quip linking the Erdős episode to OpenAI's 2019 Rubik's cube 'solved' claim, which Marcus found misleading.

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

    Assessment marcus-erdosgate-5ef844-a1.

  127. Assessment publishedBy this siteGary Marcus

    Quick acceptance of Bubeck's claim was likely driven by people believing what they wanted to believe.

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

    Assessment marcus-erdosgate-2d013a-a1.

  128. HeldBy this siteGary Marcus

    Held record marcus-erdosgate-201eba

    The final-text check failed: citations_correct; overclaims: He also says the earlier nested post explicitly described literature search, which limits any inference of deliberate concealment..

  130. HeldBy this siteGary Marcus

    Held record marcus-elon-musks-terrifying-vision-for-addaec

    hedged language is not scored, even on a plain reading

  131. Assessment publishedBy this siteGary Marcus

    Soon such unreliable AI, with butt-covering warning messages that most people will ignore, will be everywhere.

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

    Assessment marcus-elon-musks-terrifying-vision-for-9e4331-a1.

  132. HeldBy this siteGary Marcus

    Held record marcus-elon-musks-terrifying-vision-for-8ac6c6

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

  133. Assessment publishedBy this siteGary Marcus

    Musk wants to deploy biased and unreliable AI in government immediately.

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

    Assessment marcus-elon-musks-terrifying-vision-for-6d8096-a1.

  134. HeldBy this siteGary Marcus

    Held record marcus-elon-musks-terrifying-vision-for-61f856

    The openai screen reads the claim differently.

  135. Assessment publishedBy this siteGary Marcus

    Grok's biased outputs are not free speech but a massive manipulation of public opinion by Musk.

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

    Assessment marcus-elon-musks-terrifying-vision-for-6024e5-a1.

  136. Assessment publishedBy this siteGary Marcus

    The problem with Grok is not just its blatant propaganda but also more subtle content that people may not even notice.

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

    Assessment marcus-elon-musks-terrifying-vision-for-35aa38-a1.

  137. Assessment publishedBy this siteGary Marcus

    The US should not be deploying biased, unreliable AI in government under Musk's influence.

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

    Assessment marcus-elon-musks-terrifying-vision-for-2f7e5f-a1.

  139. Assessment publishedBy this siteGary Marcus

    When Marcus asked Grok 2 over the last couple of days to draw images identifying and tagging parts of images, the results were consistently awful.

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

    Assessment marcus-elon-musks-terrifying-vision-for-120653-a1.

  140. Assessment publishedBy this siteGary Marcus

    About a year ago the City of New York tried this kind of thing (deploying problematic, unreliable AI to the public), with less than spectacular results.

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

    Assessment marcus-elon-musks-terrifying-vision-for-09136c-a2.

  141. Assessment publishedBy this siteGary Marcus

    Elon Musk's timeline predictions are notoriously and, in Marcus's view, obviously bad, yet all too often reported as gospel.

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

    Assessment marcus-elon-musks-preposterous-and-possibly-e4ff36-a1.

  142. HeldBy this siteGary Marcus

    Held record marcus-elon-musks-preposterous-and-possibly-cdd9c1

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

  143. Assessment publishedBy this siteGary Marcus

    Musk will later re-issue his robot-surgery prediction with a new, later date.

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

    Assessment marcus-elon-musks-preposterous-and-possibly-b55c1f-a1.

  144. Assessment publishedBy this siteGary Marcus

    Musk's prediction risks deterring potential surgeons, as Hinton's wrong 2016 radiology claim probably deterred radiologists.

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

    Assessment marcus-elon-musks-preposterous-and-possibly-8de7b5-a1.

  145. HeldBy this siteGary Marcus

    Held record marcus-elon-musks-preposterous-and-possibly-701f64

    The anthropic screen reads the claim differently.

  146. Assessment publishedBy this siteGary Marcus

    It would be nice if these tech CEOs were more accountable for what they said.

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

    Assessment marcus-elon-musks-preposterous-and-possibly-6e46e2-a1.

  147. Assessment publishedBy this siteGary Marcus

    Elon Musk, like a lot of LLMs, is frequently wrong and never in doubt.

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

    Assessment marcus-elon-musks-preposterous-and-possibly-6a5673-a1.

  148. Assessment publishedBy this siteGary Marcus

    Robots surpassing human surgeons will probably take at least a decade longer than Musk predicts, possibly more.

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

    Assessment marcus-elon-musks-preposterous-and-possibly-4eef22-a1.

  149. Assessment publishedBy this siteGary Marcus

    Robots will not be the best surgeons on earth anytime soon; the idea is absurd.

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

    Assessment marcus-elon-musks-preposterous-and-possibly-3b86d1-a2.

  150. HeldBy this siteGary Marcus

    Held record marcus-elon-musks-preposterous-and-possibly-0f0b7e

    The anthropic screen reads the claim differently.

  151. Assessment publishedBy this siteGary Marcus

    Optimus will not be better than the best human surgeons by the end of the decade (Marcus bets $1M against it).

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

    Assessment marcus-elon-musks-preposterous-and-possibly-00040f-a1.

  153. Assessment publishedBy this siteGary Marcus

    Musk's alignment with Trump has undermined the cause of AI safety, as Marcus correctly warned in July 2024.

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

    Assessment marcus-elon-musks-inability-to-listen-to-172d37-a2.

  154. Assessment publishedBy this siteGary Marcus

    Elon Musk did not get the OpenAI he paid for and worked for.

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

    Assessment marcus-elon-musk-sues-openai-d2f81f-a2.

  155. HeldBy this siteGary Marcus

    Held record marcus-elon-musk-sues-openai-c72334

    The anthropic review reads the claim differently.

  156. HeldBy this siteGary Marcus

    Held record marcus-elon-musk-sues-openai-89201d

    The anthropic review does not accept e1 as supports.

  157. Assessment publishedBy this siteGary Marcus

    The facts set out in Musk's lawsuit against OpenAI are largely accurate.

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

    Assessment marcus-elon-musk-sues-openai-84745e-a1.

  158. Assessment publishedBy this siteGary Marcus

    Elon Musk has a point in his lawsuit against OpenAI, Altman and Brockman, though Marcus declines to predict who will win.

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

    Assessment marcus-elon-musk-sues-openai-3e12ae-a1.

  159. Assessment publishedBy this siteGary Marcus

    The OpenAI that Altman and Brockman built has little to do with what was originally promised.

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

    Assessment marcus-elon-musk-sues-openai-3dd9e4-a1.

  160. Assessment publishedBy this siteGary Marcus

    Anyone making deals with Sam Altman should read Musk's lawsuit.

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

    Assessment marcus-elon-musk-sues-openai-0c8761-a1.

  161. Assessment publishedBy this siteGary Marcus

    Right now AI has a lot of engineering and alchemy but not enough basic principles and formal understanding.

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

    Assessment marcus-elon-musk-1-yann-lecun-1-humanity-d9ed3c-a1.

  162. Assessment publishedBy this siteGary Marcus

    Musk's xAI is just building bigger LLMs, and Meta is mostly doing the same.

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

    Assessment marcus-elon-musk-1-yann-lecun-1-humanity-be82b3-a1.

  163. HeldBy this siteGary Marcus

    Held record marcus-elon-musk-1-yann-lecun-1-humanity-b61db4

    e1 is dated 2024-05, which is not inside the period described (2024-05-01 to 2024-05-29); it can only be context

  164. HeldBy this siteGary Marcus

    Held record marcus-elon-musk-1-yann-lecun-1-humanity-972f92

    The anthropic screen reads the claim differently.

  165. Assessment publishedBy this siteGary Marcus

    AI needs more science as well as engineering.

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

    Assessment marcus-elon-musk-1-yann-lecun-1-humanity-924565-a1.

  166. Assessment publishedBy this siteGary Marcus

    Reaching AGI requires more basic science.

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

    Assessment marcus-elon-musk-1-yann-lecun-1-humanity-8a72e8-a1.

  167. Assessment publishedBy this siteGary Marcus

    Marcus is not convinced that either Musk's xAI or Meta is making real AI progress despite massive budgets.

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

    Assessment marcus-elon-musk-1-yann-lecun-1-humanity-854b6b-a1.

  168. Assessment publishedBy this siteGary Marcus

    What AI really needs is a breakthrough, not just money and publicity.

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

    Assessment marcus-elon-musk-1-yann-lecun-1-humanity-812976-a1.

  169. HeldBy this siteGary Marcus

    Held record marcus-elon-musk-1-yann-lecun-1-humanity-7b655c

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

  170. Assessment publishedBy this siteGary Marcus

    Grok has a shaky basic understanding of time.

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

    Assessment marcus-elon-musk-1-yann-lecun-1-humanity-79b561-a1.

  172. HeldBy this siteGary Marcus

    Held record marcus-elegant-and-powerful-new-result-that-f5a79a

    The anthropic screen reads the claim differently.

  173. HeldBy this siteGary Marcus

    Held record marcus-elegant-and-powerful-new-result-that-f58c6b

    hedged language is not scored, even on a plain reading

  174. HeldBy this siteGary Marcus

    Held record marcus-elegant-and-powerful-new-result-that-d301bb

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

  175. HeldBy this siteGary Marcus

    Held record marcus-elegant-and-powerful-new-result-that-d301bb-2

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

  176. Assessment publishedBy this siteGary Marcus

    The generalization problems Marcus raised in 2001 about neural networks have not been adequately addressed since.

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

    Assessment marcus-elegant-and-powerful-new-result-that-bedcd1-a1.

  177. Assessment publishedBy this siteGary Marcus

    The will to believe in neural networks is frequently so strong that counterevidence is often dismissed or ignored for much too long.

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

    Assessment marcus-elegant-and-powerful-new-result-that-afed8e-a1.

  178. Assessment publishedBy this siteGary Marcus

    LLMs trained on 'A is B' cannot automatically infer that 'B is A'.

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

    Assessment marcus-elegant-and-powerful-new-result-that-99694d-a2.

  179. Assessment publishedBy this siteGary Marcus

    Computers should be expected to make optimal use of the information given, regardless of human limitations.

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

    Assessment marcus-elegant-and-powerful-new-result-that-8de031-a1.

  180. Assessment publishedBy this siteGary Marcus

    Popular neural networks still struggle with systematic relational reasoning and are not as systematic as reasoning machines should be.

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

    Assessment marcus-elegant-and-powerful-new-result-that-6cd679-a2.

  181. Assessment publishedBy this siteGary Marcus

    In the Berglund et al. experiments, the model scored zero percent on reversed made-up facts and performed dismally on reversed celebrity facts.

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

    Assessment marcus-elegant-and-powerful-new-result-that-6044cd-a1.

  182. Assessment publishedBy this siteGary Marcus

    The symmetry failure is a big, persistent error in neural networks that has endured for decades.

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

    Assessment marcus-elegant-and-powerful-new-result-that-59f0e0-a1.

  183. HeldBy this siteGary Marcus

    Held record marcus-elegant-and-powerful-new-result-that-0ff197

    The anthropic screen doubts the attribution.

  185. HeldBy this siteGary Marcus

    Held record marcus-elegant-and-powerful-new-result-that-02e257

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

  186. HeldBy this siteGary Marcus

    Held record marcus-doug-lenat-1950-2023-efc67a

    The openai review reads the claim differently.

  187. HeldBy this siteGary Marcus

    Held record marcus-doug-lenat-1950-2023-eed19a

    e1 is dated 2023, which is not inside the period described (2023-01-01 to 2023-09-01); it can only be context

  188. Assessment publishedBy this siteGary Marcus

    From an academic perspective, Cyc's biggest problem is that it is proprietary.

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

    Assessment marcus-doug-lenat-1950-2023-dca95b-a1.

  189. HeldBy this siteGary Marcus

    Held record marcus-doug-lenat-1950-2023-cb29ff

    e1 is dated 2023, which is not inside the period described (2023-01-01 to 2023-09-01); it can only be context

  190. Assessment publishedBy this siteGary Marcus

    Even if this specific LLM error is patched, other errors of a similar flavor will inevitably appear.

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

    Assessment marcus-doug-lenat-1950-2023-8238f4-a1.

  191. Assessment publishedBy this siteGary Marcus

    Cyc will seem more and more important in hindsight as real progress toward artificial general intelligence is eventually made.

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

    Assessment marcus-doug-lenat-1950-2023-75e6ec-a1.

  192. Assessment publishedBy this siteGary Marcus

    Getting machines to represent and reason about common sense still has to be done, though Cyc itself will not serve as a drop-in replacement for LLMs.

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

    Assessment marcus-doug-lenat-1950-2023-6f3a48-a1.

  193. Assessment publishedBy this siteGary Marcus

    Every young AI researcher should know something about Cyc.

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

    Assessment marcus-doug-lenat-1950-2023-693401-a1.

  194. Assessment publishedBy this siteGary Marcus

    Capturing common sense is the critical problem AI still has to solve, but it will take a different path from Cyc's.

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

    Assessment marcus-doug-lenat-1950-2023-57b89a-a1.

  195. HeldBy this siteGary Marcus

    Held record marcus-doug-lenat-1950-2023-4d9fba

    e1 is dated 2023, which is not inside the period described (2023-01-01 to 2023-09-01); it can only be context

  196. Assessment publishedBy this siteGary Marcus

    In Marcus's own view, Cyc has been neither a success nor a failure but somewhere in between: a ground-breaking, clarion experiment that never fully gelled.

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

    Assessment marcus-doug-lenat-1950-2023-3f52dc-a1.

  197. Assessment publishedBy this siteGary Marcus

    Generative AI, a system that people shouldn't trust with their taxes, is being used by certain parts of Washington to blow people up.

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

    Assessment marcus-dont-trust-generative-ai-to-do-your-e6f13a-a1.

  198. Assessment publishedBy this siteGary Marcus

    Generative AI should not be trusted with people's lives, given that it cannot even be trusted to do taxes.

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

    Assessment marcus-dont-trust-generative-ai-to-do-your-0c5563-a1.

  199. Assessment publishedBy this siteGary Marcus

    ChatGPT with DALL-E 3 cannot reliably produce images with a non-standard number of body parts.

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

    Assessment marcus-dont-ride-this-bike-generative-ais-ec6b8b-a2.

  200. HeldBy this siteGary Marcus

    Held record marcus-dont-ride-this-bike-generative-ais-ebe0e9

    e1 is dated 2024, which is not inside the period described (2024-01-01 to 2024-12-08); it can only be context