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

Changes, page 12 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 12 of 37: changes 2201 to 2400 of 7383, newest first. Dates are when the change was recorded.

  1. HeldBy this siteGary Marcus

    Held record marcus-what-was-60-minutes-thinking-in-that-624b13

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

  2. HeldBy this siteGary Marcus

    Held record marcus-what-was-60-minutes-thinking-in-that-565dc4

    The openai screen reads the claim differently.

  3. Assessment publishedBy this siteGary Marcus

    Current AI machines are not conscious.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-51b25c-a1.

  4. HeldBy this siteGary Marcus

    Held record marcus-what-was-60-minutes-thinking-in-that-4f7956

    The final-text check failed: proposition_faithful.

  5. Assessment publishedBy this siteGary Marcus

    Humanity is rushing forward with AI, a technology it does not fully understand.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-4d1386-a1.

  6. Assessment publishedBy this siteGary Marcus

    Bard's 'baby shoes' story was parroting of an existing story, and such parroting and IP theft is characteristic of large language models.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-478ede-a1.

  7. Assessment publishedBy this siteGary Marcus

    Wholesale, automated fake news will be used to manipulate both markets and elections, and it might undermine democracy.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-4516df-a1.

  8. Assessment publishedBy this siteGary Marcus

    We cannot guarantee that AI systems will not fabricate or write flawed code.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-41beff-a1.

  9. Assessment publishedBy this siteGary Marcus

    Good next-word prediction by LLMs does not necessarily imply understanding.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-40e9ec-a1.

  10. HeldBy this siteGary Marcus

    Held record marcus-what-was-60-minutes-thinking-in-that-3942bb

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

  11. Assessment publishedBy this siteGary Marcus

    AI is at a turning point of enormous uncertainty, and society needs to make the right choices now.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-2e953b-a1.

  13. Assessment publishedBy this siteGary Marcus

    Generative AI works better than most people expected.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-1d627a-a1.

  14. Assessment publishedBy this siteGary Marcus

    AI will eventually have a big impact on drug design, possibly but not necessarily with current techniques.

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

    Assessment marcus-what-was-60-minutes-thinking-in-that-06078f-a1.

  15. Assessment publishedBy this siteGary Marcus

    GPT-4 will be reckless and hard to control like its predecessors.

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

    Assessment marcus-what-to-expect-when-youre-expecting-fe6503-a2.

  16. Assessment publishedBy this siteGary Marcus

    Marcus hopes that, if all seven of his predictions prove correct, the field will finally realize that it is time to move on.

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

    Assessment marcus-what-to-expect-when-youre-expecting-ea72de-a1.

  17. Assessment publishedBy this siteGary Marcus

    ChatGPT is easily confused.

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

    Assessment marcus-what-to-expect-when-youre-expecting-e374e9-a1.

  18. Assessment publishedBy this siteGary Marcus

    On medical queries GPT-4 will either refuse (with aggressive guardrails) or occasionally produce plausible but dangerous nonsense.

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

    Assessment marcus-what-to-expect-when-youre-expecting-d59bcf-a2.

  19. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-d47c7c

    The final-text check failed: citations_correct.

  20. Assessment publishedBy this siteGary Marcus

    When AGI comes, large language models like GPT-4 may be seen in hindsight as part of the eventual solution, but only as part of it.

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

    Assessment marcus-what-to-expect-when-youre-expecting-d427cc-a1.

  21. Assessment publishedBy this siteGary Marcus

    Some of GPT-4's advice will be very bad.

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

    Assessment marcus-what-to-expect-when-youre-expecting-c9e809-a1.

  22. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-c4ae6c

    The final-text check failed: citations_correct.

  23. Assessment publishedBy this siteGary Marcus

    ChatGPT cannot currently be trusted.

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

    Assessment marcus-what-to-expect-when-youre-expecting-c24f5f-a1.

  24. Assessment publishedBy this siteGary Marcus

    Scaling alone will prove useful only up to a point.

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

    Assessment marcus-what-to-expect-when-youre-expecting-ba0b73-a2.

  25. Assessment publishedBy this siteGary Marcus

    GPT-4 will likely generate immense buzz followed by scientific scrutiny and recognition that many problems remain.

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

    Assessment marcus-what-to-expect-when-youre-expecting-b885c5-a1.

  26. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-a70743

    The two reviews differ: anthropic forecast_outcome/elapsed/supported, openai forecast_outcome/conclusive_event/supported.

  27. Assessment publishedBy this siteGary Marcus

    ChatGPT regularly makes things up and a fair amount of its output is false.

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

    Assessment marcus-what-to-expect-when-youre-expecting-8696cb-a1.

  28. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-79c713

    the conclusive event's date differs from its evidence

  29. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-797225

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

  30. Assessment publishedBy this siteGary Marcus

    GPT-4 will make a significant number of hard-to-predict stupid errors.

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

    Assessment marcus-what-to-expect-when-youre-expecting-6a17e4-a1.

  31. Assessment publishedBy this siteGary Marcus

    Because scale improves output in many ways, GPT-4 will be very powerful.

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

    Assessment marcus-what-to-expect-when-youre-expecting-651690-a1.

  32. Assessment publishedBy this siteGary Marcus

    GPT-4 will solve many prior benchmark items but still fail, especially on longer, complex scenarios.

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

    Assessment marcus-what-to-expect-when-youre-expecting-6392ec-a1.

  33. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-eff808

    The anthropic review does not accept e4 as supports.

  34. Assessment publishedBy this siteGary Marcus

    If his seven predictions prove correct, the AI field should move on from the LLM approach.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-ea72de-a1.

  35. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-c9e809

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

  36. Assessment publishedBy this siteGary Marcus

    Predictions over several years that GPT-5 would arrive imminently were repeatedly wrong.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-c599c4-a1.

  37. Assessment publishedBy this siteGary Marcus

    Scaling alone will prove useful only up to a point.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-ba0b73-a1.

  38. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-b2e16c

    The openai screen reads the claim differently.

  39. Assessment publishedBy this siteGary Marcus

    Alignment between human wants and machine behavior will remain a critical unsolved problem.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-a9525d-a1.

  40. Assessment publishedBy this siteGary Marcus

    Examples of concealed bias in GPT-5 will be discovered within days or months of release.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-a70743-a1.

  41. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-6fc5cf

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

  42. Assessment publishedBy this siteGary Marcus

    Marcus's December 2022 predictions about GPT-4 were accurate.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-6f365c-a2.

  43. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-6a17e4

    The anthropic review does not accept e1 as supports.

  44. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-488add

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

  46. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-3a4fdc

    The final-text check failed: citations_correct.

  47. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-3868af

    The openai review does not accept e2 as contradicts.

  50. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-2b6f70

    The final-text check failed: citations_correct.

  51. Assessment publishedBy this siteGary Marcus

    When AGI comes, large language models like GPT-5 may be seen in hindsight as part of the eventual solution, but only as part of it.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-275fba-a1.

  54. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-173b9e

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

  55. Assessment publishedBy this siteGary Marcus

    GPT-5, like its predecessors, will be reckless and hard to control.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-172c71-a2.

  56. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-62e-1335b1

    The final-text check failed: rationale_supported_by_evidence, citations_correct; overclaims: METR's July 2025 evaluation, published August 7, found more bugs and fewer completed features on a small set of naturally specified product tasks; it did not separate misunderstood intent from execution failures [e2]..

  57. Assessment publishedBy this siteGary Marcus

    Trustworthy AI will require genuinely new architectures with explicit knowledge and world models at their core.

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

    Assessment marcus-what-to-expect-when-youre-expecting-62e-075246-a1.

  59. Assessment publishedBy this siteGary Marcus

    GPT-4 will seem smarter than predecessors but its architecture remains problematic.

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

    Assessment marcus-what-to-expect-when-youre-expecting-629194-a1.

  60. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-4d83b4

    The anthropic review reads the claim differently.

  62. Assessment publishedBy this siteGary Marcus

    ChatGPT is not reliable, does not understand the physical or psychological world, and hallucinates.

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

    Assessment marcus-what-to-expect-when-youre-expecting-40b365-a1.

  63. Assessment publishedBy this siteGary Marcus

    GPT-4 will not reliably follow human values of helpfulness, harmlessness and truthfulness.

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

    Assessment marcus-what-to-expect-when-youre-expecting-3a4fdc-a1.

  64. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-3868af

    the conclusive event's date differs from its evidence

  65. Assessment publishedBy this siteGary Marcus

    If GPT-4 follows its predecessors' approach, it will not truly understand the world and its flaws will eventually show.

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

    Assessment marcus-what-to-expect-when-youre-expecting-28af80-a1.

  66. Assessment publishedBy this siteGary Marcus

    GPT-4 will be trained on considerably more data than GPT-3, a significant fraction of the internet.

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

    Assessment marcus-what-to-expect-when-youre-expecting-19aaa4-a1.

  67. Assessment publishedBy this siteGary Marcus

    It will remain difficult to anticipate in advance when GPT-4 will or won't do what users want.

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

    Assessment marcus-what-to-expect-when-youre-expecting-1994a4-a1.

  68. HeldBy this siteGary Marcus

    Held record marcus-what-to-expect-when-youre-expecting-171d30

    the conclusive event's date differs from its evidence

  69. Assessment publishedBy this siteGary Marcus

    GPT-4 will impress people enormously when released.

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

    Assessment marcus-what-to-expect-when-youre-expecting-128994-a1.

  70. HeldBy this siteGary Marcus

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

    The anthropic review reads the claim differently.

  71. Assessment publishedBy this siteGary Marcus

    Anthropic's cybersecurity consortium is likely a good thing but should work in conjunction with law enforcement and regulation.

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

    Assessment marcus-what-should-we-take-from-anthropics-f0441a-a1.

  73. HeldBy this siteGary Marcus

    Held record marcus-what-should-we-take-from-anthropics-cc054d

    The anthropic review does not accept e9 as supports.

  74. Assessment publishedBy this siteGary Marcus

    Time to build safeguards and regulation against AI risks is running out.

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

    Assessment marcus-what-should-we-take-from-anthropics-cc054d-2-a1.

  75. HeldBy this siteGary Marcus

    Held record marcus-what-should-we-take-from-anthropics-cb332c

    The final-text check failed: citations_correct; overclaims: The investigation attributes the fatal Mountain View crash partly to Autopilot steering outside the travel lane because of vision-processing limitations, alongside driver distraction and ineffective engagement monitoring..

  76. Assessment publishedBy this siteGary Marcus

    ChatGPT cannot reliably run a timer, yet it has been implicated in delusions, suicides, cognitive surrender and mass disinformation.

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

    Assessment marcus-what-should-we-take-from-anthropics-c3a38b-a1.

  77. Assessment publishedBy this siteGary Marcus

    Anthropic showed some admirable restraint in not publicly releasing a potentially dangerous technology (Mythos).

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

    Assessment marcus-what-should-we-take-from-anthropics-87db42-a1.

  78. Assessment publishedBy this siteGary Marcus

    Self-regulation (of AI companies) is too little, too late.

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

    Assessment marcus-what-should-we-take-from-anthropics-6c1ba1-a1.

  79. Assessment publishedBy this siteGary Marcus

    An international AI agency and treaty are needed.

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

    Assessment marcus-what-should-we-take-from-anthropics-662f57-a1.

  80. Assessment publishedBy this siteGary Marcus

    Mythos may wreak havoc even if it hallucinates and is unreliable outside domains like coding and math.

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

    Assessment marcus-what-should-we-take-from-anthropics-6120d4-a1.

  81. Assessment publishedBy this siteGary Marcus

    Without government oversight of AI releases, the public is entirely at the mercy of individual AI CEOs, some of whom have not earned trust.

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

    Assessment marcus-what-should-we-take-from-anthropics-56f0f5-a1.

  82. Assessment publishedBy this siteGary Marcus

    Mythos is probably not AGI, as suggested by the report saying little about overall capabilities.

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

    Assessment marcus-what-should-we-take-from-anthropics-3feb6f-a1.

  83. Assessment publishedBy this siteGary Marcus

    Mythos's cybersecurity danger is probably not as bad as claimed.

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

    Assessment marcus-what-should-we-take-from-anthropics-35cf74-a1.

  84. Assessment publishedBy this siteGary Marcus

    Competitors such as OpenAI and xAI might well not show restraint and could publicly release potentially dangerous technology.

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

    Assessment marcus-what-should-we-take-from-anthropics-2559e5-a1.

  85. Assessment publishedBy this siteGary Marcus

    The US government has repeatedly been complicit, repeatedly treating Altman as royalty.

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

    Assessment marcus-what-should-we-learn-from-openais-ff467c-a1.

  86. Assessment publishedBy this siteGary Marcus

    The only real hope for governing AI is citizens speaking up.

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

    Assessment marcus-what-should-we-learn-from-openais-fbe796-a1.

  88. Assessment publishedBy this siteGary Marcus

    OpenAI should not be fully trusted.

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

    Assessment marcus-what-should-we-learn-from-openais-d25978-a1.

  89. Assessment publishedBy this siteGary Marcus

    OpenAI has likely taken millions, perhaps billions, of documents without permission.

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

    Assessment marcus-what-should-we-learn-from-openais-a72903-a1.

  90. HeldBy this siteGary Marcus

    Held record marcus-what-should-we-learn-from-openais-8db22b

    The openai screen reads the claim differently.

  91. Assessment publishedBy this siteGary Marcus

    OpenAI has massive power and could possibly put humanity at risk.

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

    Assessment marcus-what-should-we-learn-from-openais-8c37d2-a1.

  92. Assessment publishedBy this siteGary Marcus

    Without independent scientists in the loop with a real voice, we are lost.

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

    Assessment marcus-what-should-we-learn-from-openais-8541eb-a1.

  93. Assessment publishedBy this siteGary Marcus

    The public should take everything OpenAI says with a grain of salt.

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

    Assessment marcus-what-should-we-learn-from-openais-6f84fb-a1.

  94. HeldBy this siteGary Marcus

    Held record marcus-what-scaling-does-and-doesnt-buy-fc009e

    The anthropic screen reads the claim differently.

  96. HeldBy this siteGary Marcus

    Held record marcus-what-scaling-does-and-doesnt-buy-80b302

    The openai screen reads the claim differently.

  97. HeldBy this siteGary Marcus

    Held record marcus-what-scaling-does-and-doesnt-buy-7092b4

    The anthropic screen doubts the attribution.

  98. HeldBy this siteGary Marcus

    Held record marcus-what-scaling-does-and-doesnt-buy-6e9cb7

    The openai screen reads the claim differently.

  99. Assessment publishedBy this siteGary Marcus

    Thanks to scaling (and some other tricks), Google's Nano Banana has terrific graphics, and its ability to edit uploaded photos is genuinely cool.

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

    Assessment marcus-what-scaling-does-and-doesnt-buy-2ae75c-a1.

  100. HeldBy this siteGary Marcus

    Held record marcus-what-scaling-does-and-doesnt-buy-27c6ae

    The openai screen reads the claim differently.

  101. Assessment publishedBy this siteGary Marcus

    The visual polish of generative image systems is superficial and hides a lack of understanding.

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

    Assessment marcus-what-scaling-does-and-doesnt-buy-03425c-a1.

  102. HeldBy this siteGary Marcus

    Held record marcus-what-matters-or-should-matter-at-fa47fe

    The anthropic review does not accept e1 as supports.

  103. HeldBy this siteGary Marcus

    Held record marcus-what-matters-or-should-matter-at-bdcf36

    the conclusive event e6 is not outcome evidence in the packet

  104. HeldBy this siteGary Marcus

    Held record marcus-what-matters-or-should-matter-at-69f008

    The anthropic screen reads the claim differently.

  105. Assessment publishedBy this siteGary Marcus

    OpenAI may not be able to maintain its rapid growth trajectory.

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

    Assessment marcus-what-if-gpt-5-didnt-meet-expectations-fea05a-a1.

  107. Assessment publishedBy this siteGary Marcus

    It is plausible that OpenAI was unsure it could profit from GPT-5 given training and inference costs.

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

    Assessment marcus-what-if-gpt-5-didnt-meet-expectations-cb2837-a1.

  110. HeldBy this siteGary Marcus

    Held record marcus-what-if-gpt-5-didnt-meet-expectations-5e36b2

    The anthropic screen reads the claim differently.

  111. Assessment publishedBy this siteGary Marcus

    If GPT-5 disappoints or is indefinitely deferred, AI valuations could deflate rapidly.

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

    Assessment marcus-what-if-gpt-5-didnt-meet-expectations-22191a-a1.

  112. Assessment publishedBy this siteGary Marcus

    Generative AI revenue is not yet there and might never materialize.

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

    Assessment marcus-what-if-generative-ai-turned-out-ff3251-a1.

  113. Assessment publishedBy this siteGary Marcus

    The US might fail to get needed consumer protections because it is fostering generative AI, which may not grow as expected.

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

    Assessment marcus-what-if-generative-ai-turned-out-fd0ed2-a1.

  114. Assessment publishedBy this siteGary Marcus

    Current generative AI is built on autocompletion rather than factuality.

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

    Assessment marcus-what-if-generative-ai-turned-out-f20765-a1.

  115. Assessment publishedBy this siteGary Marcus

    It is foolish to imagine that generative AI's challenging core problems will all suddenly be solved.

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

    Assessment marcus-what-if-generative-ai-turned-out-dfbc5a-a1.

  118. HeldBy this siteGary Marcus

    Held record marcus-what-if-generative-ai-turned-out-cc9d57

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

  120. Assessment publishedBy this siteGary Marcus

    Other potential paying customers of generative AI, beyond coders and undergraduates, may lose heart quickly.

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

    Assessment marcus-what-if-generative-ai-turned-out-a2de4b-a1.

  121. HeldBy this siteGary Marcus

    Held record marcus-what-if-generative-ai-turned-out-8a711e

    The final-text check failed: citations_correct.

  122. Assessment publishedBy this siteGary Marcus

    Scaling has helped only partially; no system can yet be guaranteed honest, harmless, or helpful.

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

    Assessment marcus-what-if-generative-ai-turned-out-81abbd-a2.

  123. Assessment publishedBy this siteGary Marcus

    We should not assume generative AI revenues will grow 1000x, which is highly speculative.

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

    Assessment marcus-what-if-generative-ai-turned-out-7e707a-a1.

  125. Assessment publishedBy this siteGary Marcus

    If user disillusionment like Evans' is representative, the generative AI field at current valuations could end fairly swiftly.

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

    Assessment marcus-what-if-generative-ai-turned-out-76889c-a1.

  126. Assessment publishedBy this siteGary Marcus

    If hallucination is not solved soon, the generative AI bubble could easily burst.

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

    Assessment marcus-what-if-generative-ai-turned-out-68af17-a1.

  127. HeldBy this siteGary Marcus

    Held record marcus-what-if-generative-ai-turned-out-6882fa

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

  128. HeldBy this siteGary Marcus

    Held record marcus-what-if-generative-ai-turned-out-5e1337

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

  129. Assessment publishedBy this siteGary Marcus

    OpenAI could have difficulty living up to its $29 billion valuation.

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

    Assessment marcus-what-if-generative-ai-turned-out-559b22-a1.

  130. Assessment publishedBy this siteGary Marcus

    Generative AI startups valued in the low billions might well eventually collapse if they earn only tens or hundreds of millions annually.

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

    Assessment marcus-what-if-generative-ai-turned-out-559b22-2-a1.

  131. Assessment publishedBy this siteGary Marcus

    If generative AI isn't going to have the impact people seem to expect, maybe we should not be building our world around the premise that it will.

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

    Assessment marcus-what-if-generative-ai-turned-out-466d76-a1.

  132. Assessment publishedBy this siteGary Marcus

    Generative AI has many serious unsolved core problems, including hallucination, unreliable tool use, and month-to-month instability.

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

    Assessment marcus-what-if-generative-ai-turned-out-30855d-a2.

  134. HeldBy this siteGary Marcus

    Held record marcus-what-if-generative-ai-turned-out-26c457

    hedged language is not scored, even on a plain reading

  135. HeldBy this siteGary Marcus

    Held record marcus-what-if-generative-ai-turned-out-25f062

    The openai screen reads the claim differently.

  136. Assessment publishedBy this siteGary Marcus

    Generative AI is among the most hyped things the author has ever seen.

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

    Assessment marcus-what-if-generative-ai-turned-out-2235c2-a1.

  137. Assessment publishedBy this siteGary Marcus

    If hallucinations aren't fixable, generative AI probably won't earn a trillion dollars a year or have the expected impact.

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

    Assessment marcus-what-if-generative-ai-turned-out-1c49e9-a1.

  138. Assessment publishedBy this siteGary Marcus

    Coding and mediocre copywriting use cases are nowhere near enough to justify current generative AI valuations.

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

    Assessment marcus-what-if-generative-ai-turned-out-090824-a2.

  139. Assessment publishedBy this siteGary Marcus

    There is no reason to think that the hallucination problem, part of the immense technical problems facing the search use case, will be solved soon.

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

    Assessment marcus-what-if-generative-ai-turned-out-07e6c8-a2.

  140. Assessment publishedBy this siteGary Marcus

    Expectations for AI will at the very least be significantly reframed, even if a full AI winter does not arrive.

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

    Assessment marcus-what-has-and-has-not-changed-in-the-d990b8-a1.

  141. Assessment publishedBy this siteGary Marcus

    GPT-5 was not released in the week before 2024-08-18, despite many people hoping it would be.

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

    Assessment marcus-what-has-and-has-not-changed-in-the-bf6ece-a2.

  142. Assessment publishedBy this siteGary Marcus

    There is further converging evidence, as of late May 2026, that tokenmaxxing (the push to maximize AI token usage) is declining.

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

    Assessment marcus-what-happens-next-after-the-decline-778ba6-a2.

  143. Assessment publishedBy this siteGary Marcus

    If search results become flooded with useless misinformation, the value of search would go to zero, which could kill Google.

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

    Assessment marcus-what-google-should-really-be-worried-e7efcc-a1.

  144. Assessment publishedBy this siteGary Marcus

    Large language models, as a tool for tricking search engines, may fundamentally change search engine manipulation.

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

    Assessment marcus-what-google-should-really-be-worried-d0cf64-a1.

  146. Assessment publishedBy this siteGary Marcus

    The Bialek fake-website hoax is probably only a small visible part of a much larger problem.

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

    Assessment marcus-what-google-should-really-be-worried-bc0bd0-a1.

  147. Assessment publishedBy this siteGary Marcus

    Web search is likely to face a major, growing threat from rings of fake, mutually reinforcing ad-selling websites.

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

    Assessment marcus-what-google-should-really-be-worried-b96161-a1.

  149. HeldBy this siteGary Marcus

    Held record marcus-what-google-should-really-be-worried-6eb7c7

    The openai review does not accept e1 as supports.

  150. HeldBy this siteGary Marcus

    Held record marcus-what-google-should-really-be-worried-2777b1

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

  151. Assessment publishedBy this siteGary Marcus

    It would be irrational for scammers not to use these new tools (LLMs) to vastly expand their reach, and hence vastly expand their profits.

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

    Assessment marcus-what-google-should-really-be-worried-1f054a-a1.

  152. Assessment publishedBy this siteGary Marcus

    Chat-style search currently has obvious truthfulness problems, and it is uncertain how far it can get.

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

    Assessment marcus-what-google-should-really-be-worried-1bc095-a1.

  153. Assessment publishedBy this siteGary Marcus

    Automatically generated fake websites, rather than ChatGPT-style search, may ultimately be the single biggest threat Google ever faces.

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

    Assessment marcus-what-google-should-really-be-worried-063aae-a1.

  154. Assessment publishedBy this siteGary Marcus

    Some important tasks may simply not improve with scale.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-f42cc7-a1.

  155. Assessment publishedBy this siteGary Marcus

    There may not be enough available compute in the world to make scaling maximalism feasible.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-e47a81-a1.

  156. Assessment publishedBy this siteGary Marcus

    Capturing the pragmatic implications of language without cognitive models and common sense is really hard.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-e1c19a-a1.

  157. Assessment publishedBy this siteGary Marcus

    If performance plateaus at about 80% on even one important task like pragmatics, that may mean game over for scaling.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-ddba63-a1.

  160. Assessment publishedBy this siteGary Marcus

    There may not be enough data in the world for scaling maximalism to work.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-96d64b-a1.

  161. Assessment publishedBy this siteGary Marcus

    Scaling maximalism will not get us to AGI.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-8c2a45-a1.

  162. Assessment publishedBy this siteGary Marcus

    AI scaling is an empirical observation, not a law of nature that is guaranteed to keep holding.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-83395d-a1.

  163. HeldBy this siteGary Marcus

    Held record marcus-what-game-over-for-the-latest-paradigm-693075

    The anthropic review does not accept e1 as supports.

  164. Assessment publishedBy this siteGary Marcus

    If the goal is AGI, the AI field should stop putting so much of its effort into scaling alone.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-43b4a0-a1.

  165. Assessment publishedBy this siteGary Marcus

    A system that gets syntax and semantics right but fails on pragmatics or commonsense reasoning is not AGI you can trust.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-3449c2-a1.

  166. Assessment publishedBy this siteGary Marcus

    On the Ruis et al. pragmatics task, the best model reached only 80.6%, and for most models scaling had at best a negligible effect.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-30ad12-a1.

  167. Assessment publishedBy this siteGary Marcus

    If the Merrill, Warstadt and Linzen proof is even close to correct, scaling maximalism may soon be in real trouble.

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

    Assessment marcus-what-game-over-for-the-latest-paradigm-18df58-a1.

  168. Assessment publishedBy this siteGary Marcus

    Marcus does not think OpenAI is in any kind of imminent danger, but he also does not think it is out of the woods.

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

    Assessment marcus-what-exactly-are-the-economics-of-f5a11c-a1.

  169. Assessment publishedBy this siteGary Marcus

    Investment in driverless cars, around $100 billion, vastly exceeds the revenue they have generated so far.

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

    Assessment marcus-what-exactly-are-the-economics-of-e23039-a1.

  170. Assessment publishedBy this siteGary Marcus

    LLMs will probably become cheaper to operate over time as hardware cheapens and efficiency improves.

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

    Assessment marcus-what-exactly-are-the-economics-of-ded8c5-a1.

  173. HeldBy this siteGary Marcus

    Held record marcus-what-exactly-are-the-economics-of-843ad8

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

  174. Assessment publishedBy this siteGary Marcus

    Coding will remain a solid use case keeping generative AI prominent for a long time, and programmers will not abandon it.

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

    Assessment marcus-what-exactly-are-the-economics-of-8145e7-a1.

  175. Assessment publishedBy this siteGary Marcus

    The financial foundation of so-called foundation models is not a house of cards, but it is hardly as robust as it seems.

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

    Assessment marcus-what-exactly-are-the-economics-of-7a45e0-a1.

  176. Assessment publishedBy this siteGary Marcus

    Microsoft is unlikely to let OpenAI go under anytime soon.

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

    Assessment marcus-what-exactly-are-the-economics-of-798954-a1.

  177. Assessment publishedBy this siteGary Marcus

    Generative AI could turn out to be a fad, though it is too early to tell.

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

    Assessment marcus-what-exactly-are-the-economics-of-775e40-a1.

  179. HeldBy this siteGary Marcus

    Held record marcus-what-exactly-are-the-economics-of-474ddd

    The anthropic screen doubts the attribution.

  180. HeldBy this siteGary Marcus

    Held record marcus-what-exactly-are-the-economics-of-46a432

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

  181. HeldBy this siteGary Marcus

    Held record marcus-what-exactly-are-the-economics-of-1a6ff1

    e1 is dated 2023, which is not inside the period described (2023-01-01 to 2023-08-14); it can only be context; e2 is dated 2023, which is not inside the period described (2023-01-01 to 2023-08-14); it can only be context

  183. HeldBy this siteGary Marcus

    Held record marcus-what-exactly-are-the-economics-of-017c1b

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

  185. HeldBy this siteGary Marcus

    Held record marcus-what-does-meta-ais-diplomacy-winning-d9349d

    The openai review reads the claim differently.

  187. HeldBy this siteGary Marcus

    Held record marcus-what-does-meta-ais-diplomacy-winning-aa6ed7

    The openai review reads the claim differently.

  189. HeldBy this siteGary Marcus

    Held record marcus-what-does-meta-ais-diplomacy-winning-8cdffa

    The final-text check failed: citations_correct.

  190. Assessment publishedBy this siteGary Marcus

    Cicero may change the field's view of machine learning as a universal solvent.

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

    Assessment marcus-what-does-meta-ais-diplomacy-winning-5491f5-a1.

  191. HeldBy this siteGary Marcus

    Held record marcus-what-does-meta-ais-diplomacy-winning-4362a5

    The openai screen reads the claim differently.

  192. Assessment publishedBy this siteGary Marcus

    Cicero's architecture differs profoundly from most AI approaches discussed in recent years.

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

    Assessment marcus-what-does-meta-ais-diplomacy-winning-41910d-a1.

  193. Assessment publishedBy this siteGary Marcus

    If Diplomacy were played with alternative rules, Cicero would probably need to be retrained almost from scratch.

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

    Assessment marcus-what-does-meta-ais-diplomacy-winning-409181-a1.

  195. HeldBy this siteGary Marcus

    Held record marcus-what-does-meta-ais-diplomacy-winning-28afe3

    The anthropic screen reads the claim differently.

  196. Assessment publishedBy this siteGary Marcus

    Cicero's Diplomacy results are, without question, genuinely impressive.

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

    Assessment marcus-what-does-meta-ais-diplomacy-winning-1a26ea-a1.

  198. Assessment publishedBy this siteGary Marcus

    Cicero cannot be immediately applied to tasks such as customer service or domestic robotics, or almost anything besides blitz Diplomacy.

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

    Assessment marcus-what-does-meta-ais-diplomacy-winning-0ec58b-a1.

  199. HeldBy this siteGary Marcus

    Held record marcus-what-does-meta-ais-diplomacy-winning-03d397

    The openai screen reads the claim differently.

  200. Assessment publishedBy this siteGary Marcus

    Improvement from GPT-2 to GPT-3 has been quantitative only, with zero qualitative change.

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

    Assessment marcus-what-does-it-mean-when-an-ai-fails-f2ac96-a1.