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

Changes, page 5 of 52

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 (10273)

Page 5 of 52: changes 801 to 1000 of 10273, newest first. Dates are when the change was recorded.

  1. HeldBy this siteEd Zitron

    Held record wyea-ai-is-really-weird-9bccfa

    e1 is dated 2026, which is not inside the period described (2025-01-01 to 2026-04-08); it can only be context; e1 is decisive for a financial claim but states no metric scope

  2. HeldBy this siteEd Zitron

    Held record wyea-ai-is-really-weird-9427f8

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

  4. HeldBy this siteEd Zitron

    Held record wyea-ai-is-really-weird-6f41a6

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

  6. Assessment publishedBy this siteEd Zitron

    Meta does not track any ROI on token burn; tokenmaxxing there exists only to climb a leaderboard.

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

    Assessment wyea-ai-is-really-weird-4c0f69-a1.

  7. HeldBy this siteEd Zitron

    Held record wyea-ai-is-really-weird-3c4350

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

  8. HeldBy this siteEd Zitron

    Held record wyea-ai-is-really-weird-318f57

    the contract's accounting basis is unknown; a financial claim is scored only on a stated basis; contradicted needs at least one contradicting item of outcome evidence

  9. HeldBy this siteEd Zitron

    Held record wyea-ai-is-really-weird-1d0ff8

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

  11. HeldBy this siteEd Zitron

    Held record wyea-ai-is-already-in-dangerous-hands-9518d4

    a observation claim cannot be assessed on the causal_assessment axis

  12. HeldBy this siteEd Zitron

    Held record wyea-ai-is-already-in-dangerous-hands-5021c1

    The openai review reads the claim differently.

  13. HeldBy this siteEd Zitron

    Held record wyea-ai-is-a-money-trap-74178d

    The two reviews differ: anthropic forecast_outcome/open_horizon/not_adjudicable, openai forecast_outcome/open_horizon/unresolved.

  14. HeldBy this siteEd Zitron

    Held record wyea-ai-is-a-money-trap-4606d0

    The final-text check failed: citations_correct.

  15. HeldBy this siteGary Marcus

    Held record marcus-will-tesla-ever-solve-full-self-driving-d5ca7b

    hedged language is not scored, even on a plain reading

  16. HeldBy this siteGary Marcus

    Held record marcus-why-my-pdoom-has-risen-dramatically-6a0e1a

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

  17. HeldBy this siteGary Marcus

    Held record marcus-why-is-bing-so-reckless-90f9da

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

  19. Assessment publishedBy this siteGary Marcus

    Breaking-news errors in LLMs arise, like hallucinations, because pure LLMs correlate words but lack stable, revisable world models.

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

    Assessment marcus-why-chatgpt-cant-be-trusted-with-a446b0-a1.

  20. Assessment publishedBy this siteGary Marcus

    California's SB-1047 was significantly weakened in last-minute negotiations.

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

    Assessment marcus-why-californias-ai-safety-bill-should-a304ec-a1.

  21. HeldBy this siteGary Marcus

    Held record marcus-when-it-comes-to-security-llms-are-ee710a

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

  22. Assessment publishedBy this siteGary Marcus

    There is little if any government regulation in place to address chatbot guardrail failures.

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

    Assessment marcus-when-it-comes-to-security-llms-are-94566d-a1.

  23. HeldBy this siteGary Marcus

    Held record marcus-when-it-comes-to-security-llms-are-3f9d5d

    The final-text check failed: citations_correct.

  24. HeldBy this siteGary Marcus

    Held record marcus-when-it-comes-to-security-llms-are-2fbe0f

    The anthropic review reads the claim differently.

  25. Assessment publishedBy this siteGary Marcus

    The more people use LLMs, the more trouble we are going to be in.

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

    Assessment marcus-when-it-comes-to-security-llms-are-23d26c-a1.

  26. HeldBy this siteGary Marcus

    Held record marcus-when-it-comes-to-security-llms-are-1e7204

    The openai review does not accept e1 as supports.

  27. Assessment publishedBy this siteGary Marcus

    Despite years of jailbreaks, the tech industry has no robust defense against them.

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

    Assessment marcus-when-it-comes-to-security-llms-are-19b54d-a1.

  28. HeldBy this siteGary Marcus

    Held record marcus-when-it-comes-to-ai-policy-congress-2a3cac-2

    The final-text check failed: citations_correct.

  29. HeldBy this siteGary Marcus

    Held record marcus-when-it-comes-to-ai-policy-congress-1b268c

    The final-text check failed: citations_correct.

  31. Assessment publishedBy this siteGary Marcus

    The language technology that PaLM-SayCan relies on is well known to be problematic.

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

    Assessment marcus-whats-wrong-with-googles-new-robot-730ea2-a1.

  32. HeldBy this siteGary Marcus

    Held record marcus-whats-wrong-with-googles-new-robot-668692

    The anthropic review does not accept e2 as contradicts.

  33. Assessment publishedBy this siteGary Marcus

    More data alone will not be enough to solve the problems of robotics.

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

    Assessment marcus-whats-wrong-with-googles-new-robot-525fa1-a1.

  34. Assessment publishedBy this siteGary Marcus

    Making driverless cars work reliably has proven far harder than most people realized.

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

    Assessment marcus-whats-wrong-with-googles-new-robot-519bcd-a1.

  35. HeldBy this siteGary Marcus

    Held record marcus-whats-wrong-with-googles-new-robot-2e4ae3

    e1 is dated 2022, which is not inside the period described (2022-01-01 to 2022-08-18); it can only be context

  36. Assessment publishedBy this siteGary Marcus

    Building a robot on top of a language system with so little comprehension of the world cannot be a recipe for success.

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

    Assessment marcus-whats-wrong-with-googles-new-robot-1cbb4b-a1.

  37. Assessment publishedBy this siteGary Marcus

    If citizens don't make clear that AI policy matters, candidates for the House, Senate, and President will not talk about it.

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

    Assessment marcus-what-will-the-candidates-say-about-d9a220-a2.

  38. Assessment publishedBy this siteGary Marcus

    Nonconsensual deepfake porn is a growing problem.

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

    Assessment marcus-what-will-the-candidates-say-about-c2395f-a2.

  39. HeldBy this siteGary Marcus

    Held record marcus-what-will-the-candidates-say-about-9fb955

    "unlikely" is hedged or probabilistic; one outcome does not score it; hedged language is not scored, even on a plain reading

  40. Assessment publishedBy this siteGary Marcus

    AI is contributing to an escalation in fraud and disinformation.

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

    Assessment marcus-what-will-the-candidates-say-about-51ecde-a2.

  41. Assessment publishedBy this siteGary Marcus

    At this point there is no turning back: governments going forward will have to consider which AI deployments are unsafe and block those that are.

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

    Assessment marcus-what-washington-must-do-c0888a-a1.

  43. Assessment publishedBy this siteGary Marcus

    The White House decision will slow down Anthropic, one of the leading US AI labs.

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

    Assessment marcus-what-washington-must-do-2eb682-a1.

  44. HeldBy this siteGary Marcus

    Held record marcus-what-was-60-minutes-thinking-in-that-8f1110

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

  45. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-a7d273

    The final-text check failed: proposition_faithful.

  46. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-1994a4

    The final-text check failed: rationale_supported_by_evidence, citations_correct; overclaims: ORCFLO's consistency score measures repeatability, not users' advance predictions; its 40 English text-only cases exclude several major task types..

  47. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-119f96

    The final-text check failed: proposition_faithful.

  49. Assessment publishedBy this siteGary Marcus

    OpenAI never delivered on its promise to devote 20% of its efforts to AI safety.

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

    Assessment marcus-what-should-we-learn-from-openais-c4e887-a2.

  50. Assessment publishedBy this siteGary Marcus

    Foundation models are deeply unstable, which makes them difficult to engineer into reliable products.

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

    Assessment marcus-what-exactly-are-the-economics-of-62834f-a1.

  51. HeldBy this siteGary Marcus

    Held record marcus-what-does-it-mean-when-an-ai-fails-e19a2c

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

  52. Assessment publishedBy this siteGary Marcus

    AI agents are causing incidents with frightening regularity, and the latest, from Anthropic, is fairly serious.

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

    Assessment marcus-we-must-recall-open-ended-ai-agents-df52ad-a1.

  53. HeldBy this siteGary Marcus

    Held record marcus-unwoven-f9a60a

    The anthropic review does not accept e1 as supports.

  54. Assessment publishedBy this siteGary Marcus

    CoreWeave's IPO valued it at about $23 billion, a major retreat from the roughly $32 billion it was thought to be worth weeks earlier.

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

    Assessment marcus-unwoven-f15c54-a1.

  55. HeldBy this siteGary Marcus

    Held record marcus-two-years-later-deep-learning-is-dee1ae

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

  56. Assessment publishedBy this siteGary Marcus

    OpenAI's new effort to solve these problems still ends up fabricating authoritative nonsense.

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

    Assessment marcus-two-years-later-deep-learning-is-d4c2a5-a2.

  57. Assessment publishedBy this siteGary Marcus

    Deep learning is a pattern-recognition technique that works best for rough, low-stakes results where perfect results are optional.

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

    Assessment marcus-two-years-later-deep-learning-is-caff83-a1.

  58. Assessment publishedBy this siteGary Marcus

    Scaling has helped but has not solved any of the fundamental deep-learning problems Marcus identified.

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

    Assessment marcus-two-years-later-deep-learning-is-c3a58d-a2.

  59. Assessment publishedBy this siteGary Marcus

    As of March 2024, deep learning still has not replaced a single radiologist.

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

    Assessment marcus-two-years-later-deep-learning-is-8c751c-a1.

  60. Assessment publishedBy this siteGary Marcus

    Current deep-learning systems still often make stupid errors.

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

    Assessment marcus-two-years-later-deep-learning-is-6d3049-a1.

  61. HeldBy this siteGary Marcus

    Held record marcus-two-years-later-deep-learning-is-2b030b

    The anthropic review does not accept e2 as supports.

  62. HeldBy this siteGary Marcus

    Held record marcus-two-years-later-deep-learning-is-105e61

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

  63. Assessment publishedBy this siteGary Marcus

    Microsoft's current AI products are not transparent and in some ways demonstrably not safe.

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

    Assessment marcus-two-models-of-ai-oversight-and-how-8173bb-a2.

  64. Assessment publishedBy this siteGary Marcus

    Scientists lack access to how GPT-4 works and its training data, making it about as far from transparent as possible.

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

    Assessment marcus-two-models-of-ai-oversight-and-how-7164b6-a2.

  65. HeldBy this siteGary Marcus

    Held record marcus-top-5-reasons-teslas-optimus-announcement-d356a6

    The two reviews differ: anthropic forecast_outcome/open_horizon/not_adjudicable, openai forecast_outcome/open_horizon/unresolved.

  66. Assessment publishedBy this siteGary Marcus

    The real problem with robots has never been demos, of which there have been many for decades; it is making robots robust in the real world.

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

    Assessment marcus-top-5-reasons-teslas-optimus-announcement-9426af-a1.

  67. HeldBy this siteGary Marcus

    Held record marcus-top-5-reasons-teslas-optimus-announcement-50989a

    the conclusive event e1 is not outcome evidence in the packet

  69. HeldBy this siteGary Marcus

    Held record marcus-top-5-reasons-openai-is-probably-0fc4e6

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

  70. Assessment publishedBy this siteGary Marcus

    If OpenAI had to retrain only on consented materials, its results would be significantly weakened and retraining would be very costly.

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

    Assessment marcus-top-5-reasons-openai-is-probably-0d189e-a2.

  71. HeldBy this siteGary Marcus

    Held record marcus-top-5-reasons-openai-is-probably-09f4d3

    the contract's accounting basis is unknown; a financial claim is scored only on a stated basis; supported needs at least one supporting item of outcome evidence

  72. HeldBy this siteGary Marcus

    Held record marcus-too-much-benefit-of-the-doubt-eac54d

    e1 is dated 2022, which is not inside the period described (2022-01-01 to 2022-10-16); it can only be context

  73. Assessment publishedBy this siteGary Marcus

    Using AI to augment workers is much easier than replacing them, because augmentation does not require solving the whole job.

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

    Assessment marcus-too-much-benefit-of-the-doubt-cd49a0-a1.

  74. HeldBy this siteGary Marcus

    Held record marcus-too-much-benefit-of-the-doubt-c7d8d8

    e1 is dated 2022, which is not inside the period described (2022-01-01 to 2022-10-16); it can only be context

  75. Assessment publishedBy this siteGary Marcus

    There is no evidence that current scan-reading AI techniques will produce a full radiologist replacement any time soon.

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

    Assessment marcus-too-much-benefit-of-the-doubt-c0ff9d-a1.

  76. Assessment publishedBy this siteGary Marcus

    The study's R = 0.5 correlation predicts only 25% of the variance, leaving three-quarters unexplained, not half as the paper implied.

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

    Assessment marcus-too-much-benefit-of-the-doubt-979f4c-a1.

  77. HeldBy this siteGary Marcus

    Held record marcus-too-much-benefit-of-the-doubt-806778

    e2 is dated 2022, which is not inside the period described (2022-01-01 to 2022-10-16); it can only be context

  78. Assessment publishedBy this siteGary Marcus

    The evidence does not support claiming any direct link between deep learning models' inner workings and the human brain.

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

    Assessment marcus-too-much-benefit-of-the-doubt-7a87fd-a1.

  79. Assessment publishedBy this siteGary Marcus

    As of October 2022, AI still lacks comprehension, and very few if any jobs have actually been replaced by AI.

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

    Assessment marcus-too-much-benefit-of-the-doubt-723cae-a1.

  80. Assessment publishedBy this siteGary Marcus

    Having AI read radiology scans is very different from replacing the entire job of a radiologist.

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

    Assessment marcus-too-much-benefit-of-the-doubt-718d37-a1.

  81. Assessment publishedBy this siteGary Marcus

    AI will not fully displace radiologists, as opposed to augmenting them, before 2033.

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

    Assessment marcus-too-much-benefit-of-the-doubt-59b79b-a1.

  82. Assessment publishedBy this siteGary Marcus

    No current AI system can learn new things from tiny amounts of data the way human children constantly do.

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

    Assessment marcus-too-much-benefit-of-the-doubt-4cae0d-a1.

  84. HeldBy this siteGary Marcus

    Held record marcus-too-much-benefit-of-the-doubt-25ebc5

    The openai review reads the claim differently.

  85. Assessment publishedBy this siteGary Marcus

    There is currently no AI system that is a viable competitor to human general intelligence.

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

    Assessment marcus-too-much-benefit-of-the-doubt-1b6e98-a1.

  86. Assessment publishedBy this siteGary Marcus

    Humanoid domestic robots require general intelligence, and there are no serious candidates for such intelligence yet.

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

    Assessment marcus-too-much-benefit-of-the-doubt-114c2e-a1.

  87. Assessment publishedBy this siteGary Marcus

    We do not currently know how to build machines that can read and understand tax codes well enough to replace accountants.

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

    Assessment marcus-too-much-benefit-of-the-doubt-0b495a-a1.

  88. HeldBy this siteGary Marcus

    Held record marcus-too-big-to-fail-b703ff

    The final-text check failed: citations_correct.

  89. Assessment publishedBy this siteGary Marcus

    Gary Marcus has long been worried that OpenAI and the LLM industry were getting too big to fail.

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

    Assessment marcus-too-big-to-fail-8fb273-a1.

  90. Assessment publishedBy this siteGary Marcus

    Because of unreliability and immense costs from inefficiency, ChatGPT has failed so far to make profits for companies like OpenAI.

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

    Assessment marcus-three-years-on-chatgpt-still-isnt-d4be08-a2.

  91. Assessment publishedBy this siteGary Marcus

    If the trial is framed as a referendum on Musk rather than OpenAI's founders, Musk will lose the lawsuit against OpenAI.

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

    Assessment marcus-three-thoughts-on-the-musk-openai-351df3-a2.

  92. Assessment publishedBy this siteGary Marcus

    Debugging is hard and a big part of what coders do, and it is not about to be replaced by AI.

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

    Assessment marcus-those-claiming-were-mere-months-away-d66a7a-a1.

  93. Assessment publishedBy this siteGary Marcus

    Current machine learning techniques handle outliers poorly and so struggle with driverless cars and high-level reasoning in unusual circumstances.

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

    Assessment marcus-this-one-important-fact-about-current-f263c7-a1.

  94. HeldBy this siteGary Marcus

    Held record marcus-this-one-important-fact-about-current-e72ed7

    The anthropic review reads the claim differently.

  95. HeldBy this siteGary Marcus

    Held record marcus-this-one-important-fact-about-current-8c35c2

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

  96. HeldBy this siteGary Marcus

    Held record marcus-this-one-important-fact-about-current-7ad1f3

    The final-text check failed: citations_correct.

  97. HeldBy this siteGary Marcus

    Held record marcus-this-one-important-fact-about-current-6e4fe6

    hedged language is not scored, even on a plain reading

  98. Assessment publishedBy this siteGary Marcus

    People in AI who do not understand that machine learning handles outliers poorly have wasted a tremendous amount of money.

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

    Assessment marcus-this-one-important-fact-about-current-667be0-a1.

  99. HeldBy this siteGary Marcus

    Held record marcus-this-one-important-fact-about-current-59b46c

    e2 is dated 2024, which is not inside the period described (1990-01-01 to 2024-08-01); it can only be context

  100. Assessment publishedBy this siteGary Marcus

    The industry built on generative AI will collapse because people don't understand the outlier problem.

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

    Assessment marcus-this-one-important-fact-about-current-54019a-a1.

  101. Assessment publishedBy this siteGary Marcus

    For 30 years, neural networks have struggled with outliers while symbolic systems handled them well.

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

    Assessment marcus-this-one-important-fact-about-current-4b8735-a2.

  102. Assessment publishedBy this siteGary Marcus

    More than a quarter century after his 1998 work, handling outliers (distribution shift) is still the Achilles' heel of neural networks.

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

    Assessment marcus-this-one-important-fact-about-current-0d51f7-a1.

  103. HeldBy this siteGary Marcus

    Held record marcus-this-one-important-fact-about-current-08bd7a

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

  104. HeldBy this siteGary Marcus

    Held record marcus-this-one-important-fact-about-current-0059de

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

  105. Assessment publishedBy this siteGary Marcus

    After GPT-5's disappointment, OpenAI's lead and Altman's reputation have dwindled, though the lead is not gone.

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

    Assessment marcus-things-are-so-desperate-at-openai-db9bad-a1.

  106. HeldBy this siteGary Marcus

    Held record marcus-things-are-so-desperate-at-openai-d668ca

    The final-text check failed: citations_correct.

  107. HeldBy this siteGary Marcus

    Held record marcus-things-are-so-desperate-at-openai-c77d62

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

  108. Assessment publishedBy this siteGary Marcus

    No form of deep learning has yet solved reasoning, hallucinations, stupid errors, or factuality.

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

    Assessment marcus-things-are-so-desperate-at-openai-670510-a1.

  109. HeldBy this siteGary Marcus

    Held record marcus-things-are-so-desperate-at-openai-3e39d1

    The final-text check failed: citations_correct.

  110. HeldBy this siteGary Marcus

    Held record marcus-things-are-about-to-get-a-lot-worse-fe969c

    The final-text check failed: proposition_faithful.

  111. Assessment publishedBy this siteGary Marcus

    Generative AI systems do not inform users when they produce infringing material.

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

    Assessment marcus-things-are-about-to-get-a-lot-worse-fe79c3-a2.

  112. Assessment publishedBy this siteGary Marcus

    Current generative AI systems do not give users a manifest of sources.

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

    Assessment marcus-things-are-about-to-get-a-lot-worse-679273-2-a1.

  113. Assessment publishedBy this siteGary Marcus

    Unless a new architecture that reliably tracks provenance is invented, generative AI infringement, often unrequested by users, will continue.

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

    Assessment marcus-things-are-about-to-get-a-lot-worse-395723-a2.

  114. Assessment publishedBy this siteGary Marcus

    OpenAI has not been transparent about its training data.

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

    Assessment marcus-things-are-about-to-get-a-lot-worse-1c64d7-a1.

  115. Assessment publishedBy this siteGary Marcus

    Generative AI systems can produce copyright-infringing materials.

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

    Assessment marcus-things-are-about-to-get-a-lot-worse-0adbbc-a1.

  119. HeldBy this siteGary Marcus

    Held record marcus-the-secret-behind-metas-muse-bb66a1

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

  120. Assessment publishedBy this siteGary Marcus

    Facebook M was canceled in January 2018, less than three years after launch.

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

    Assessment marcus-the-secret-behind-metas-muse-328161-a2.

  121. HeldBy this siteGary Marcus

    Held record marcus-the-secret-behind-metas-muse-2a2175

    The anthropic review reads the claim differently.

  122. Assessment publishedBy this siteGary Marcus

    Facebook's Project M agent never really took off.

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

    Assessment marcus-the-secret-behind-metas-muse-08363b-a2.

  124. Assessment publishedBy this siteGary Marcus

    OpenAI has shown no evident moat beyond its initial customers, who could move.

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

    Assessment marcus-the-sam-altman-playbook-e78f08-a2.

  125. Assessment publishedBy this siteGary Marcus

    OpenAI does not pay most creators for training materials, and many creators are losing gigs.

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

    Assessment marcus-the-sam-altman-playbook-c1e7e9-a1.

  126. HeldBy this siteGary Marcus

    Held record marcus-the-sam-altman-playbook-9aa24b

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

  128. Assessment publishedBy this siteGary Marcus

    OpenAI mostly sells chatbots, and charitable work is at best a small part of its operation.

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

    Assessment marcus-the-sam-altman-playbook-567e9e-a1.

  129. Assessment publishedBy this siteGary Marcus

    People who buy Altman's narrative will continue to invest.

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

    Assessment marcus-the-sam-altman-playbook-5242b7-a1.

  130. Assessment publishedBy this siteGary Marcus

    OpenAI has had no net profits so far, and roughly half of any future profits must be shared with Microsoft for a long time.

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

    Assessment marcus-the-sam-altman-playbook-476421-a1.

  132. Assessment publishedBy this siteGary Marcus

    ChatGPT usership has more or less stabilized without significant growth since May 2023.

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

    Assessment marcus-the-sam-altman-playbook-17dd17-a1.

  133. HeldBy this siteGary Marcus

    Held record marcus-the-rise-and-fall-of-microsofts-new-dc1c99

    The anthropic review does not accept e2 as supports.

  134. Assessment publishedBy this siteGary Marcus

    Financial challenges facing ChatGPT/OpenAI were starting to mount as of August 2023.

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

    Assessment marcus-the-rise-and-fall-of-chatgpt-b6c05d-a1.

  136. Assessment publishedBy this siteGary Marcus

    A backlash against generative AI and ChatGPT had begun by August 2023.

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

    Assessment marcus-the-rise-and-fall-of-chatgpt-a80dd0-a1.

  137. Assessment publishedBy this siteGary Marcus

    Not every enterprise-scale generative AI project will succeed.

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

    Assessment marcus-the-rise-and-fall-of-chatgpt-822283-a2.

  138. HeldBy this siteGary Marcus

    Held record marcus-the-rise-and-fall-of-chatgpt-5a340a

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

  139. HeldBy this siteGary Marcus

    Held record marcus-the-one-issue-unifying-the-american-bbe03a

    The final-text check failed: citations_correct.

  140. HeldBy this siteGary Marcus

    Held record marcus-the-one-issue-unifying-the-american-b2269b

    The final-text check failed: rationale_supported_by_evidence, citations_correct; overclaims: The release identifies June 16–July 19 as the survey period and cautions that subgroup uncertainty exceeds the full-sample margin of error..

  141. Assessment publishedBy this siteGary Marcus

    Marcus has repeatedly stressed that a model could be a good model for engineering without being a good model of humans, or vice versa.

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

    Assessment marcus-the-new-sport-of-misrepresenting-854a4c-a1.

  143. Assessment publishedBy this siteGary Marcus

    Marcus's 2012 concerns about deep learning still stand, and the problem for the neural network community is not that he has changed his goal posts.

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

    Assessment marcus-the-new-sport-of-misrepresenting-50b5c8-a1.

  144. Assessment publishedBy this siteGary Marcus

    Current neural networks lack the cognitive depth and robustness needed for reliable Level 5 self-driving from vision alone.

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

    Assessment marcus-the-new-sport-of-misrepresenting-0c607b-a2.

  146. Assessment publishedBy this siteGary Marcus

    DALL-E 2 is fairly prone to uncomprehending errors such as failing to render 'a red cube on top of a blue cube'.

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

    Assessment marcus-the-new-science-of-alt-intelligence-d724f8-a1.

  147. Assessment publishedBy this siteGary Marcus

    Scaling is currently the predominant strand of work in AI that substitutes data for intelligence.

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

    Assessment marcus-the-new-science-of-alt-intelligence-d2de9c-a1.

  148. Assessment publishedBy this siteGary Marcus

    DeepMind's Gato achieves novel cross-modal feats but remains unreliable, mixing brilliance with total discomprehension.

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

    Assessment marcus-the-new-science-of-alt-intelligence-ca7f59-a1.

  149. HeldBy this siteGary Marcus

    Held record marcus-the-new-science-of-alt-intelligence-a08647

    e1 is dated 2022, which is not inside the period described (2022-01-01 to 2022-05-14); it can only be context

  150. Assessment publishedBy this siteGary Marcus

    So far, in most but not all domains, Big Data approaches have beaten careful knowledge engineering.

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

    Assessment marcus-the-new-science-of-alt-intelligence-99f3a2-a1.

  151. Assessment publishedBy this siteGary Marcus

    Virtually all of the world's software still depends on knowledge engineering.

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

    Assessment marcus-the-new-science-of-alt-intelligence-9051c6-a1.

  154. HeldBy this siteGary Marcus

    Held record marcus-the-new-science-of-alt-intelligence-759653

    The final-text check failed: citations_correct.

  155. HeldBy this siteGary Marcus

    Held record marcus-the-new-science-of-alt-intelligence-421790

    The final-text check failed: citations_correct.

  156. HeldBy this siteGary Marcus

    Held record marcus-the-new-science-of-alt-intelligence-1728ef

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

  158. HeldBy this siteGary Marcus

    Held record marcus-the-need-for-a-president-that-speaks-e1119d

    e1 is dated 2024-06, which is not inside the period described (2024-01-01 to 2024-06-28); it can only be context

  159. HeldBy this siteGary Marcus

    Held record marcus-the-need-for-a-president-that-speaks-c383e4

    hedged language is not scored, even on a plain reading

  161. Assessment publishedBy this siteGary Marcus

    Returns from scaling generative AI are starting to diminish.

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

    Assessment marcus-the-need-for-a-president-that-speaks-6091ac-a1.

  162. HeldBy this siteGary Marcus

    Held record marcus-the-need-for-a-president-that-speaks-3df855

    a observation claim cannot be assessed on the causal_assessment axis

  163. Assessment publishedBy this siteGary Marcus

    The next US President's AI policy decisions on corporate liability and government oversight will have lasting impact.

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

    Assessment marcus-the-need-for-a-president-that-speaks-0292f9-a1.

  165. Assessment publishedBy this siteGary Marcus

    OpenAI strayed from its original mission.

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

    Assessment marcus-the-irony-and-the-agony-of-elon-musks-cfd930-a1.

  166. Assessment publishedBy this siteGary Marcus

    OpenAI, Altman and Brockman have changed OpenAI's mission since Musk gave them money, time and reputation.

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

    Assessment marcus-the-irony-and-the-agony-of-elon-musks-909dd8-a1.

  167. HeldBy this siteGary Marcus

    Held record marcus-the-hottest-new-programming-language-a7b060

    The final-text check failed: citations_correct.

  168. HeldBy this siteGary Marcus

    Held record marcus-the-hottest-new-programming-language-8231fa

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

  169. HeldBy this siteGary Marcus

    Held record marcus-the-grand-agi-delusion-2aaae2-2

    The final-text check failed: citations_correct.

  170. HeldBy this siteGary Marcus

    Held record marcus-the-generative-ai-fizzle-eb59aa

    hedged language is not scored, even on a plain reading

  171. Assessment publishedBy this siteGary Marcus

    Most of the generative-AI-related stocks he highlighted are down significantly for the month.

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

    Assessment marcus-the-generative-ai-fizzle-da37fe-a1.

  172. Assessment publishedBy this siteGary Marcus

    Many things in AI are overvalued, though that does not mean they will all deflate at once.

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

    Assessment marcus-the-generative-ai-fizzle-b97943-a1.

  173. HeldBy this siteGary Marcus

    Held record marcus-the-generative-ai-fizzle-2f0902

    The final-text check failed: proposition_faithful, citations_correct.

  174. HeldBy this siteGary Marcus

    Held record marcus-the-five-stages-of-agi-grief-f662a8

    The openai review does not accept e1 as supports.

  175. HeldBy this siteGary Marcus

    Held record marcus-the-five-stages-of-agi-grief-a3c4ec

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

  176. HeldBy this siteGary Marcus

    Held record marcus-the-five-stages-of-agi-grief-9d153a

    The openai review does not accept e2 as supports.

  178. Assessment publishedBy this siteGary Marcus

    Yann LeCun originated none of the five ideas he is publicly known for, and he rarely if ever credits the people who did.

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

    Assessment marcus-the-false-glorification-of-yann-lecun-f07e39-a1.

  179. Assessment publishedBy this siteGary Marcus

    The idea of world models in AI is not new; it goes back to the 1950s and Herb Simon's General Problem Solver.

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

    Assessment marcus-the-false-glorification-of-yann-lecun-dfd44f-a1.

  180. Assessment publishedBy this siteGary Marcus

    LeCun has never publicly acknowledged that Marcus was on target with his conjecture that scaling LLMs would not lead to AGI.

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

    Assessment marcus-the-false-glorification-of-yann-lecun-d60b6e-a1.

  182. HeldBy this siteGary Marcus

    Held record marcus-the-false-glorification-of-yann-lecun-5567ed

    The anthropic review does not accept e2 as contradicts.

  183. HeldBy this siteGary Marcus

    Held record marcus-the-false-glorification-of-yann-lecun-29c961

    hedged language is not scored, even on a plain reading

  184. Assessment publishedBy this siteGary Marcus

    The myths about the originality of LeCun's thought are not true.

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

    Assessment marcus-the-false-glorification-of-yann-lecun-06734f-a1.

  185. Assessment publishedBy this siteGary Marcus

    LeCun only fully changed his mind about LLMs a few weeks after November 2022, after Meta's Galactica tanked and ChatGPT outcompeted Meta.

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

    Assessment marcus-the-false-glorification-of-yann-lecun-024a9f-a1.

  186. Assessment publishedBy this siteGary Marcus

    Reviewers and journals will not be able to keep up with the flood of LLM-generated submissions.

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

    Assessment marcus-the-exponential-enshittification-aa15a9-a2.

  187. HeldBy this siteGary Marcus

    Held record marcus-the-exponential-enshittification-a80466

    the conclusive event's date differs from its evidence

  188. Assessment publishedBy this siteGary Marcus

    LLM makers are about to destroy the scientific journal review system.

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

    Assessment marcus-the-exponential-enshittification-910ba6-2-a1.

  189. Assessment publishedBy this siteGary Marcus

    The problem of LLM-generated content in scientific publishing is getting worse quickly.

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

    Assessment marcus-the-exponential-enshittification-48f0fb-a1.

  190. Assessment publishedBy this siteGary Marcus

    People's private data will be used to feed data-hungry AI systems, whether or not they like it.

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

    Assessment marcus-the-dirty-underbelly-of-ai-5d0270-a2.

  191. Assessment publishedBy this siteGary Marcus

    Generative image programs are prone to the same kind of regurgitation of copyrighted material as text models.

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

    Assessment marcus-the-desperate-race-to-save-generative-e83129-a1.

  192. Assessment publishedBy this siteGary Marcus

    If copyright rules are rewritten for GenAI, eventually no one will create good new content and the internet will degrade into recycled AI output.

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

    Assessment marcus-the-desperate-race-to-save-generative-67c50f-a2.

  193. Assessment publishedBy this siteGary Marcus

    We are about to enter an era in which AI is ubiquitous but still fallible.

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

    Assessment marcus-the-cnet-fake-news-fiasco-autopilot-f0c71e-2-a1.

  194. Assessment publishedBy this siteGary Marcus

    Mackworth was right that humans can't pay attention 100% of the time.

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

    Assessment marcus-the-cnet-fake-news-fiasco-autopilot-b7b0a5-a1.

  195. HeldBy this siteGary Marcus

    Held record marcus-the-cnet-fake-news-fiasco-autopilot-b74698

    The anthropic review does not accept e9 as supports.

  196. HeldBy this siteGary Marcus

    Held record marcus-the-cnet-fake-news-fiasco-autopilot-b6c681

    The anthropic review does not accept e3 as supports.

  197. HeldBy this siteGary Marcus

    Held record marcus-the-cnet-fake-news-fiasco-autopilot-5528cc

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

  198. HeldBy this siteGary Marcus

    Held record marcus-the-cnet-fake-news-fiasco-autopilot-4a9185

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

  199. Assessment publishedBy this siteGary Marcus

    As automated systems approach but do not reach perfection, humans overseeing them become more likely to stop paying attention.

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

    Assessment marcus-the-cnet-fake-news-fiasco-autopilot-3a4f96-a1.

  200. Assessment publishedBy this siteGary Marcus

    OpenAI, one of the most visible AI companies, had its governance change literally overnight in November 2023.

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

    Assessment marcus-the-biggest-immediate-lesson-from-fe5d16-2-a1.