Changes, page 37 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 37 of 37: changes 7201 to 7383 of 7383, newest first. Dates are when the change was recorded.

2026-10-29
HeldBy this siteGary Marcus
Held record marcus-a-cern-for-ai-and-the-global-governance-bfbde1
e1 is dated 2023-10, which is not inside the period described (2023-10-01 to 2023-10-26); it can only be context
2026-10-29
Assessment publishedBy this siteGary Marcus
Current AI cannot reliably anticipate the consequences or harms of actions it recommends.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-b58928-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
A CERN for AI safety and global AI safety governance should be created, provided they address a balanced range of risks.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-891fe2-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
There is no way to guarantee that large language models reliably stick to facts.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-87d6e2-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
The world can likely benefit enormously from AI, in ways extending far beyond the productivity gains seen in recent months, though there are clearly risks too.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-7a56ea-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
AI risk discussion should focus much more on misuse of LLMs by criminals and terrorists, not just on AI takeover scenarios.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-568684-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
AI need not be superintelligent to create serious problems; the real issue is control.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-3e0a1c-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
We need to consider both current AI risk and AGI risk.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-2fd9f9-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
People should not listen to those who deny there is any AI risk, such as Marc Andreessen, because enormous money is at stake.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-2e9236-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Some AI problems may be too complex for individual labs and of insufficient financial interest to profit-driven large AI companies, which still holds true.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-2113d8-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-a-cern-for-ai-and-the-global-governance-1afc82
The final-text check failed: proposition_faithful, citations_correct.
2026-10-29
Assessment publishedBy this siteGary Marcus
It seems to Marcus that Anthropic's Constitutional AI idea is clever and the best of the current AI safety ideas, but still inadequate, riddled with the usual LLM problems such as unreliability, excessive dependence on similarity to training data, hallucinations and unreliable reasoning.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-0e36a9-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
The combined AI safety efforts of large tech companies have not yielded much.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-a-cern-for-ai-and-the-global-governance-04e3d6-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
SoftBank has sold its entire Nvidia position, which is an ominous sign for generative AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-5-recent-ominous-signs-for-generative-d71145-a2.
2026-10-29
Assessment publishedBy this siteGary Marcus
Meta's financing games have been revealed, presented as one of five recent ominous signs for generative AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-5-recent-ominous-signs-for-generative-c9f2b4-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-5-recent-ominous-signs-for-generative-b91f1e
The anthropic screen reads the claim differently.
2026-10-29
Assessment publishedBy this siteGary Marcus
CoreWeave's stock fell 20% in one week, which is an ominous sign for generative AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-5-recent-ominous-signs-for-generative-97a70e-a2.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-5-recent-ominous-signs-for-generative-6a6791
The two reviews differ: anthropic historical_observation/not_applicable/mixed, openai historical_observation/not_applicable/contradicted.
2026-10-29
Assessment publishedBy this siteGary Marcus
Oracle's debt has been downgraded, which is an ominous sign for generative AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-5-recent-ominous-signs-for-generative-0bdad9-a2.
2026-10-29
Assessment publishedBy this siteGary Marcus
Knowing more about exactly what it is, beyond brain size, that makes the human mind unique would tell us an immense amount both about ourselves and about how we might want to build AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-5-myths-about-learning-and-innateness-0b2f3f-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
What Marcus said to Joscha Bach 26 months earlier still holds, 100%.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-26-months-of-ridicule-and-failure-c233ef-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-26-months-of-ridicule-and-failure-9b97b9
The final-text check failed: conditions_shown, rationale_supported_by_evidence; overclaims: The evidence of progress (e4) would complicate those claims..
2026-10-29
Assessment publishedBy this siteGary Marcus
Despite tens of billions of dollars invested since 2022, deep learning systems are still unreliable and still produce toxicity and hallucinations.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-26-months-of-ridicule-and-failure-8dd634-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
There has been no progress toward solving the AI alignment problem between 2022 and mid-2024.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-26-months-of-ridicule-and-failure-39298b-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Current AI technology, which Musk is now wildly overhyping, remains exactly as incorrigible as it ever was.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-26-months-of-ridicule-and-failure-2a1e62-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-fe41ce
The two reviews differ: anthropic forecast_outcome/elapsed/supported, openai forecast_outcome/elapsed/unresolved.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-f9f222
The final-text check failed; overclaims: This statement says the contractual definition stayed unchanged, so it was in force throughout 2025..
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-f904d8
The two reviews differ: anthropic forecast_outcome/elapsed/supported, openai forecast_outcome/elapsed/unresolved.
2026-10-29
Assessment publishedBy this siteGary Marcus
AI will presumably eventually solve compositionality, factuality and reasoning, but current techniques have not been adequate.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-f890fa-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
His predictions about generative AI's technical and economic limits were largely correct, but he was wrong about investor behavior.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-efc3b5-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
There could be no GPT-5 level model, meaning a huge across-the-board leap by community consensus, throughout 2025.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-e6b4e8-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Marcus suspects, without having tested it, that changes of state and the persistence of objects will still cause problems for Google's Veo 2, though it seems better than Sora.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-e2e649-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Possibly none of the Marcus-Brundage tasks will be reliably solved by end of 2025.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-e23736-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Commercial artists and voiceover actors have perhaps been hardest hit by AI so far.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-ddafbc-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-d39923
The anthropic review does not accept e4 as supports.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-ccb6d1
e1 measures something else (company OpenAI is not Leading AI model developers (e.g., OpenAI)); it can only be context unless it is derived with stated inputs; e2 measures something else (company OpenAI is not Leading AI model developers (e.g., OpenAI)); it can only be context unless it is derived with stated inputs; e3 measures something else (company Nvidia is not Leading AI model developers (e.g., OpenAI)); it can only be context unless it is derived with stated inputs
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-cc59d3
The two reviews differ: anthropic forecast_outcome/elapsed/supported, openai forecast_outcome/elapsed/unresolved.
2026-10-29
Assessment publishedBy this siteGary Marcus
In 2025 there may be a large-scale cyberattack in which generative AI plays an important causal role.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-cad01f-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-c5591b
The anthropic screen reads the claim differently.
2026-10-29
Assessment publishedBy this siteGary Marcus
By the end of 2025 truly driverless cars, in which no human is required to attend to traffic, will continue to see usage limited to a modest number of cities, mainly in the West and mainly in good weather, and human drivers will still make up a large part of the economy.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-c35871-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Corporate adoption of AI is far more limited than most people expected, and total profits across all companies, except hardware companies like Nvidia that profit from chips rather than models, have been modest at best.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-b9a0eb-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
OpenAI's o3 will likely be quite expensive when released.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-b25cff-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
With medium confidence, by the end of 2025 neurosymbolic AI will become much more prominent.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-acc12a-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
By the end of 2025 OpenAI will have continued to preview products months, and perhaps years, before they are solid and widely available at an accessible price.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-a757a5-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-a281a1
e1 measures something else (company Inflection AI is not AI model companies (OpenAI, Inflection, Adept, Character.AI, others)); it can only be context unless it is derived with stated inputs; e2 measures something else (company Adept is not AI model companies (OpenAI, Inflection, Adept, Character.AI, others)); it can only be context unless it is derived with stated inputs; e3 measures something else (company Covariant is not AI model companies (OpenAI, Inflection, Adept, Character.AI, others)); it can only be context unless it is derived with stated inputs; e4 measures something else (company
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-9d5c34
The two reviews differ: anthropic historical_observation/not_applicable/supported, openai historical_observation/not_applicable/unresolved.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-9a1ed8
The two reviews differ: anthropic forecast_outcome/elapsed/contradicted, openai forecast_outcome/elapsed/unresolved.
2026-10-29
Assessment publishedBy this siteGary Marcus
With medium confidence, by the end of 2025 a technical moat will continue to be elusive, with more convergence on broadly similar models across the US and China, and some European systems catching up to roughly the same place.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-942130-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-9034fd
The final-text check failed; overclaims: In 2025, generative-AI fabrications in a paid government report led to its revision and a partial refund..
2026-10-29
Assessment publishedBy this siteGary Marcus
AI agents will be heavily hyped in 2025 but far from reliable except possibly in very narrow use cases.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-6d352a-a2.
2026-10-29
Assessment publishedBy this siteGary Marcus
There are many GPT-4 level models, a price war, and very little moat among LLM providers.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-67a324-a3.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-66331f
The anthropic review does not accept e3 as contradicts.
2026-10-29
Assessment publishedBy this siteGary Marcus
Through 2025 hallucinations will continue to plague generative AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-64d30d-a3.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-5a33a0
The two reviews differ: anthropic forecast_outcome/elapsed/supported, openai forecast_outcome/elapsed/unresolved.
2026-10-29
Assessment publishedBy this siteGary Marcus
There is still no robust solution to LLM hallucinations.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-597446-a3.
2026-10-29
Assessment publishedBy this siteGary Marcus
OpenAI has no profits, does not expect any until at least 2029, burns billions, and lacks a clear technical moat.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-58c0ec-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
OpenAI's long-term security, especially at high valuations, is far from assured.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-57b086-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
AI-generated misinformation became increasingly widespread, but its effects have thus far been less than Marcus feared; he remains quite concerned and thinks vigilance continues to be required.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-570d43-a2.
2026-10-29
Assessment publishedBy this siteGary Marcus
With medium confidence, by the end of 2025 few companies, and even fewer consumers, will adopt o3 at wide scale because of concerns about price and robustness relative to that price.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-5672a6-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
By end of 2025 less than 10%, probably less than 5%, of the workforce will be replaced by AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-559339-a2.
2026-10-29
Assessment publishedBy this siteGary Marcus
2024 was a year of AI disillusionment for some but not all; his prediction was overstated.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-550827-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
With medium confidence, by the end of 2025 Sora will continue to have trouble with physics.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-4cddd6-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-4acef5
The openai review found the criteria unfair to the speaker's words.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-3fb203
The two reviews differ: anthropic historical_observation/not_applicable/supported, openai historical_observation/not_applicable/unresolved.
2026-10-29
Assessment publishedBy this siteGary Marcus
His 2023 prediction that OpenAI might someday be seen as the WeWork of AI has not been confirmed, but OpenAI is far from out of the woods.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-3f1880-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
By end of 2025 few if any radiologists will be replaced by AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-2a5fb1-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Sora's troubles with physics have not gone away since its February 2024 announcement.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-249ada-a2.
2026-10-29
Assessment publishedBy this siteGary Marcus
In 2025 we may see o1-like models good at synthetic-data-friendly tasks but only incrementally better than GPT-4 elsewhere.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-236bfc-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Through 2025 reasoning errors will continue to plague generative AI.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-22c0aa-a3.
2026-10-29
Assessment publishedBy this siteGary Marcus
With medium confidence, by the end of 2025 companies will continue to experiment with AI, but adoption of production-grade systems scaled out in the real world will continue to be tentative.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-1e5178-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-1c5d29
The two reviews differ: anthropic historical_observation/not_applicable/supported, openai historical_observation/not_applicable/unresolved.
2026-10-29
Assessment publishedBy this siteGary Marcus
Valuations of major AI companies could start to fall in 2025.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-1a5f39-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-0ab350
The two reviews differ: anthropic forecast_outcome/elapsed/supported, openai forecast_outcome/elapsed/unresolved.
2026-10-29
Assessment publishedBy this siteGary Marcus
Copyright lawsuits over generative AI will continue throughout 2025.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-092660-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
His seven December 2022 predictions about GPT-4 largely held true for every model released in 2024, with prediction number seven still pending.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-25-ai-predictions-for-2025-from-marcus-054763-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-25-ai-predictions-for-2025-from-marcus-012f05
The openai review found the criteria unfair to the speaker's words.
2026-10-29
Assessment publishedBy this siteGary Marcus
AI sweatshops result from substituting shallow data for deep understanding in AI design.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-seriously-embarrassing-hours-for-e8c38d-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Genuine AI would not require vast amounts of training data.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-seriously-embarrassing-hours-for-e1f1ce-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
CNET was the first casualty of putting too much faith in ChatGPT, publishing AI-written stories with many mistakes.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-seriously-embarrassing-hours-for-dcb322-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-24-seriously-embarrassing-hours-for-cd2743
The final-text check failed: rationale_supported_by_evidence; overclaims: OpenAI's statement to TIME confirmed only that Sama workers in Kenya contributed to a toxic-content detector later built into ChatGPT..
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-24-seriously-embarrassing-hours-for-8ada97
The openai review found the criteria unfair to the speaker's words.
2026-10-29
Assessment publishedBy this siteGary Marcus
In choosing what kind of AI to build, sometimes the slower road is the better road.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-seriously-embarrassing-hours-for-6b01b4-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Anthropic's Claude, despite its alignment focus, still quickly hallucinates like ChatGPT.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-seriously-embarrassing-hours-for-3a1915-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-24-seriously-embarrassing-hours-for-361e6a
The openai review found the criteria unfair to the speaker's words.
2026-10-29
Assessment publishedBy this siteGary Marcus
Genuine AI would not require sweatshop labor.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-seriously-embarrassing-hours-for-27172f-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-24-seriously-embarrassing-hours-for-222fb0
The two reviews differ: anthropic historical_observation/not_applicable/supported, openai historical_observation/not_applicable/unresolved.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-24-seriously-embarrassing-hours-for-1323fd
The openai review found the criteria unfair to the speaker's words.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-24-seriously-embarrassing-hours-for-118c22
The anthropic review does not accept e4 as contradicts.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-24-seriously-embarrassing-hours-for-0f41ab
The anthropic review reads the claim differently.
2026-10-29
Assessment publishedBy this siteGary Marcus
Tesla FSD is getting better, and he raises the question of whether it is getting enough better.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-hours-in-ai-news-b04c94-a2.
2026-10-29
Assessment publishedBy this siteGary Marcus
The generative AI investment bubble may not last forever, as signalled by some late-stage VC firms sitting out.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-hours-in-ai-news-69af4a-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
Generative AI is starting to get a tiny bit of a 'Scarlet Letter' feel, which he presents as possible foreshadowing.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-24-hours-in-ai-news-05a396-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-1984-but-with-llms-f72a52
hedged language is not scored, even on a plain reading
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-1984-but-with-llms-ca2320
The final-text check failed; overclaims: But 64% is short of 'pretty much every', the review stops at 2023, and it does not test any effect on views..
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-1984-but-with-llms-bbafdd
The openai review found the criteria unfair to the speaker's words.
2026-10-29
Assessment publishedBy this siteGary Marcus
If a system's input is deliberately altered, it may wind up faring poorly on other benchmarks, a cost Musk may not wish to bear.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-1984-but-with-llms-b17699-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
No LLM has fully coherent beliefs, because LLMs repeat conflicting material from inconsistent training data instead of reasoning their way to a coherent view.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-1984-but-with-llms-8e842e-a3.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-1984-but-with-llms-8dfaae
The openai review found the criteria unfair to the speaker's words.
2026-10-29
Assessment publishedBy this siteGary Marcus
Musk's plan to rewrite Grok's training data will probably be harder than he expects, because it cannot be done by hand and automating it to his satisfaction will be difficult.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-1984-but-with-llms-74f9eb-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-1984-but-with-llms-54c24a
The openai review found the criteria unfair to the speaker's words.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-1984-but-with-llms-513b59
The anthropic review does not accept e3 as supports.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-1984-but-with-llms-3c3539
The final-text check failed; overclaims: But every company ranged from neutral to left, with none clearly right..
2026-10-29
Assessment publishedBy this siteGary Marcus
Even if LLMs are not AGI, they could easily become the most powerful form of mind control ever invented.
First assessment published after both reviews and the final-text check passed.
Assessment marcus-1984-but-with-llms-2d4453-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-1984-but-with-llms-183cfc
The final-text check failed: rationale_supported_by_evidence; overclaims: The sentence is a rhetorical warning about intent: Altman and Zuckerberg will 'try to' do what Musk did, meaning shape LLMs to control minds. | Only limited searching on OpenAI's actions..
2026-10-29
HeldBy this siteGary Marcus
Held record marcus-10-million-says-we-wont-see-human-843a7d
The openai review does not accept e3 as supports.
2026-10-29
HeldBy this siteGary Marcus
Held record americansunlight-substack-com-cp-168074209-f8024d
The anthropic screen reads the claim differently.
2026-10-29
Assessment publishedBy this siteGary Marcus
Generative AI chatbots' lack of reasoning can directly contribute to the harmful effects of LLM grooming by disinformation actors.
First assessment published after both reviews and the final-text check passed.
Assessment americansunlight-substack-com-cp-168074209-db31c0-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record americansunlight-substack-com-cp-168074209-54e1ab
The anthropic screen reads the claim differently.
2026-10-29
HeldBy this siteGary Marcus
Held record americansunlight-substack-com-cp-168074209-533d41
The anthropic screen reads the claim differently.
2026-10-29
Assessment publishedBy this siteGary Marcus
Current AIs, being naive mimicry systems, degrade their own futures and training data whenever they repeat propaganda.
First assessment published after both reviews and the final-text check passed.
Assessment americansunlight-substack-com-cp-168074209-244111-a1.
2026-10-29
Assessment publishedBy this siteGary Marcus
If an LLM's training data or search-accessible data is tainted, the model may well regurgitate the tainted information.
First assessment published after both reviews and the final-text check passed.
Assessment americansunlight-substack-com-cp-168074209-15b585-a1.
2026-10-29
HeldBy this siteGary Marcus
Held record americansunlight-substack-com-cp-168074209-145d88
The anthropic screen reads the claim differently.
2026-10-29
Assessment publishedBy this siteGary Marcus
To the extent that such systems now supply news to a large number of people, o3's failure to note that the Pravda network is a known disinformation network is a serious failure with potentially serious consequences.
First assessment published after both reviews and the final-text check passed.
Assessment americansunlight-substack-com-cp-168074209-07451c-a1.
2026-10-28
Site correctionBy this siteGary Marcus
Marcus seriously doubts that, by the end of 2025, AI will be able to learn and master the basics of almost any new video game within a few minutes or hours and solve original puzzles in the alternate world of that video game.
A new assessment with the same outcome replaced the previous one.
Assessment gm-video-game-mastery-by-end-2025-a3.
2026-10-28
Site correctionBy this siteGary Marcus
Offered as an unlikely scenario that could nonetheless come true: by the end of 2026, the giant AI infrastructure plays (like Project Stargate) that President Trump championed after his inauguration will look like an unprofitable and underused mistake.
A new assessment with the same outcome replaced the previous one.
Assessment gm-stargate-unprofitable-mistake-2026-a2.
2026-10-28
Site correctionBy this siteGary Marcus
Filed under 'Medium confidence': at the end of 2025, OpenAI's Sora will still have trouble with physics.
A new assessment with the same outcome replaced the previous one.
Assessment gm-sora-physics-trouble-2025-a3.
2026-10-28
Site correctionBy this siteGary Marcus
The current generation of AI agents should be removed from the market until they can be fixed.
A new assessment with the same outcome replaced the previous one.
Assessment gm-recall-open-ended-agents-a2.
2026-10-28
Site correctionBy this siteGary Marcus
Filed under 'Medium confidence': by the end of 2025, few companies (and even fewer consumers) will adopt OpenAI's o3 at wide scale, because of concerns about price and robustness relative to that price.
A new assessment with the same outcome replaced the previous one.
Assessment gm-o3-limited-adoption-2025-a2.
2026-10-28
HeldBy this siteGary Marcus
Held record gm-no-superhuman-ai-by-end-2025
The openai review found the criteria unfair to the speaker's words.
2026-10-28
HeldBy this siteGary Marcus
Held record gm-no-hallucination-free-chatbots-by-2025
The openai review found the criteria unfair to the speaker's words.
2026-10-28
HeldBy this siteGary Marcus
Held record gm-no-10k-line-bugfree-code-2029
The openai review found the criteria unfair to the speaker's words.
2026-10-28
Assessment publishedBy this siteGary Marcus
Filed under 'Medium confidence': through the end of 2025, a technical moat will continue to be elusive; instead there will be more convergence on broadly similar models across both the US and China, and some systems in Europe will catch up to roughly the same place.
First assessment published after both reviews and the final-text check passed.
Assessment gm-moat-convergence-2025-a3.
2026-10-28
Site correctionBy this siteGary Marcus
By the end of 2024, hallucinations by large language models will not have been reduced to the level of human experts.
A new assessment with the same outcome replaced the previous one.
Assessment gm-hallucinations-expert-level-2024-a4.
2026-10-28
Site correctionBy this siteGary Marcus
OpenAI will not release GPT-5 in 2024.
A new assessment with the same outcome replaced the previous one.
Assessment gm-gpt-5-not-in-2024-a4.
2026-10-28
Site correctionBy this siteGary Marcus
During 2024 and 2025, leading AI figures will further revise and lower their public expectations for AI, such as AGI timelines.
A new assessment with the same outcome replaced the previous one.
Assessment gm-expect-revisionism-2024-2025-a4.
2026-10-28
Site correctionBy this siteGary Marcus
Marcus seriously doubts the projections of near-term AGI, including Dario Amodei's projection that we will have AGI in 2-3 years from August 2023.
A new assessment with the same outcome replaced the previous one.
Assessment gm-doubt-agi-in-2-3-years-2023-a3.
2026-10-28
HeldBy this siteGary Marcus
Held record gm-copyright-suits-continue-2025
The final-text check failed: citations_correct.
2026-10-28
Site correctionBy this siteGary Marcus
Marcus suspects that OpenAI's GPT-6 Astra will show loads of problems with open-ended real-world tasks, despite its success on ARC-AGI.
A new assessment with the same outcome replaced the previous one.
Assessment gm-astra-open-ended-task-problems-a2.
2026-10-28
Site correctionBy this siteGary Marcus
If 'great' AI research agents do not arrive by the end of 2025, everything later in the AI 2027 scenario's timeline is pushed back.
A new assessment with the same outcome replaced the previous one.
Assessment gm-ai2027-research-agents-moved-back-a3.
2026-10-28
Site correctionBy this siteGary Marcus
Throughout 2025, AI agents will be endlessly hyped but far from reliable, except possibly in very narrow use cases.
A new assessment with the same outcome replaced the previous one.
Assessment gm-agents-hyped-unreliable-2025-a3.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-wall-street-punishes-big-tech-by-q1-2025
The openai review found the criteria unfair to the speaker's words.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-oracle-collapses-without-openai
The openai review found the criteria unfair to the speaker's words.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-openai-next-round-by-july-2025
The openai review found the criteria unfair to the speaker's words.
2026-10-28
Site correctionBy this siteEd Zitron
If OpenAI does not convert to a for-profit structure by the end of 2025, it will be effectively dead.
A new assessment with the same outcome replaced the previous one.
Assessment ez-openai-convert-by-end-2025-or-dead-a2.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-openai-broadcom-asic-not-in-2026
The two reviews differ: anthropic forecast_outcome/not_due/unresolved, openai forecast_outcome/not_due/not_adjudicable.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-openai-another-round-by-march-2027
The final-text check failed; overclaims: An OpenAI funding round announced and signed by 2027-03-31, or none by then, with any statement by OpenAI about its need for capital..
2026-10-28
Site correctionBy this siteEd Zitron
By the end of 2025, Microsoft will have spent over $200 billion in capital expenditures since 2023.
A new assessment with the same outcome replaced the previous one.
Assessment ez-microsoft-200bn-capex-by-end-2025-a4.
2026-10-28
Site correctionBy this siteEd Zitron
By the end of 2026, the hyperscalers (Meta, Google, Amazon, and Microsoft) will have sunk over $1.3 trillion into generative AI since March 2022.
A new assessment with the same outcome replaced the previous one.
Assessment ez-hyperscalers-1-3tn-genai-by-end-2026-a4.
2026-10-28
Site correctionBy this siteEd Zitron
If generative AI becomes an albatross around the necks of Google, Microsoft, or Amazon (which Zitron believes it will), these companies will face serious pressure to curtail their investments and capital expenditures.
A new assessment with the same outcome replaced the previous one.
Assessment ez-big-tech-capex-pressure-albatross-a5.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-big-tech-560bn-ai-capex-2024-2025-if-promises-kept
The anthropic review found the criteria unfair to the speaker's words.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-big-tech-2tn-ai-capex-by-2027
The openai review found the criteria unfair to the speaker's words.
2026-10-28
Site correctionBy this siteEd Zitron
Because the funds Anthropic raised in early 2026 and its cash flow will not cover its costs, Anthropic will have to raise at least another $200 billion within a year of 2026-06-08.
A new assessment with the same outcome replaced the previous one.
Assessment ez-anthropic-raise-200bn-within-year-2026-a2.
2026-10-28
Site correctionBy this siteEd Zitron
Anthropic's revenue for calendar 2025 was about $4.6 billion.
A new assessment with the same outcome replaced the previous one.
Assessment ez-anthropic-2025-revenue-4-6bn-a2.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-anthropic-2025-net-loss-42bn
The final-text check failed: citations_correct.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-anthropic-2025-lost-over-8bn
The two reviews differ: anthropic historical_observation/not_applicable/supported, openai not_applicable/not_applicable/not_adjudicable.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-ai-three-quarters-to-prove-itself
The anthropic review found the criteria unfair to the speaker's words.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-ai-services-majority-token-billing-by-end-2026
The openai review found the criteria unfair to the speaker's words.
2026-10-28
Assessment publishedBy this siteEd Zitron
Within a few months of 2025-10-20, AI-native companies (products built on top of AI models) will face multiple price increases.
First assessment published after both reviews and the final-text check passed.
Assessment ez-ai-native-price-increases-few-months-a3.
2026-10-28
Assessment publishedBy this siteGary Marcus
Marcus seriously doubts that, by the end of 2025, AI will be able to learn and master the basics of almost any new video game within a few minutes or hours and solve original puzzles in the alternate world of that video game.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
Offered as an unlikely scenario that could nonetheless come true: by the end of 2026, the giant AI infrastructure plays (like Project Stargate) that President Trump championed after his inauguration will look like an unprofitable and underused mistake.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
Filed under 'Medium confidence': at the end of 2025, OpenAI's Sora will still have trouble with physics.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
The current generation of AI agents should be removed from the market until they can be fixed.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
Filed under 'Medium confidence': by the end of 2025, few companies (and even fewer consumers) will adopt OpenAI's o3 at wide scale, because of concerns about price and robustness relative to that price.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
Held record gm-no-superhuman-ai-by-end-2025
First assessment published after both reviews and the final-text check passed.
2026-10-28
HeldBy this siteGary Marcus
Held record gm-no-hallucination-free-chatbots-by-2025
The openai review found the criteria unfair to the speaker's words.
2026-10-28
Assessment publishedBy this siteGary Marcus
Held record gm-no-10k-line-bugfree-code-2029
First assessment published after both reviews and the final-text check passed.
2026-10-28
HeldBy this siteGary Marcus
Filed under 'Medium confidence': through the end of 2025, a technical moat will continue to be elusive; instead there will be more convergence on broadly similar models across both the US and China, and some systems in Europe will catch up to roughly the same place.
The final-text check failed; overclaims: A prediction stated without a medium or low confidence grade, with a named measure of the gap between leading models, would be scored..
2026-10-28
Assessment publishedBy this siteGary Marcus
By the end of 2024, hallucinations by large language models will not have been reduced to the level of human experts.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
OpenAI will not release GPT-5 in 2024.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
During 2024 and 2025, leading AI figures will further revise and lower their public expectations for AI, such as AGI timelines.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
Marcus seriously doubts the projections of near-term AGI, including Dario Amodei's projection that we will have AGI in 2-3 years from August 2023.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
Held record gm-copyright-suits-continue-2025
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
Marcus suspects that OpenAI's GPT-6 Astra will show loads of problems with open-ended real-world tasks, despite its success on ARC-AGI.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
If 'great' AI research agents do not arrive by the end of 2025, everything later in the AI 2027 scenario's timeline is pushed back.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteGary Marcus
Throughout 2025, AI agents will be endlessly hyped but far from reliable, except possibly in very narrow use cases.
First assessment published after both reviews and the final-text check passed.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-wall-street-punishes-big-tech-by-q1-2025
The openai review abstained.
2026-10-28
Assessment publishedBy this siteEd Zitron
Held record ez-oracle-collapses-without-openai
First assessment published after both reviews and the final-text check passed.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-openai-next-round-by-july-2025
The openai review found the criteria unfair to the speaker's words.
2026-10-28
Assessment publishedBy this siteEd Zitron
If OpenAI does not convert to a for-profit structure by the end of 2025, it will be effectively dead.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteEd Zitron
Held record ez-openai-broadcom-asic-not-in-2026
First assessment published after both reviews and the final-text check passed.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-openai-another-round-by-march-2027
The final-text check failed; overclaims: An OpenAI funding round announced and signed by 2027-03-31, or none by then, with any statement by OpenAI about its need for capital..
2026-10-28
Assessment publishedBy this siteEd Zitron
By the end of 2025, Microsoft will have spent over $200 billion in capital expenditures since 2023.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteEd Zitron
By the end of 2026, the hyperscalers (Meta, Google, Amazon, and Microsoft) will have sunk over $1.3 trillion into generative AI since March 2022.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteEd Zitron
If generative AI becomes an albatross around the necks of Google, Microsoft, or Amazon (which Zitron believes it will), these companies will face serious pressure to curtail their investments and capital expenditures.
First assessment published after both reviews and the final-text check passed.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-big-tech-560bn-ai-capex-2024-2025-if-promises-kept
The anthropic review found the criteria unfair to the speaker's words.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-big-tech-2tn-ai-capex-by-2027
The openai review found the criteria unfair to the speaker's words.
2026-10-28
Assessment publishedBy this siteEd Zitron
Because the funds Anthropic raised in early 2026 and its cash flow will not cover its costs, Anthropic will have to raise at least another $200 billion within a year of 2026-06-08.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteEd Zitron
Anthropic's revenue for calendar 2025 was about $4.6 billion.
First assessment published after both reviews and the final-text check passed.
2026-10-28
Assessment publishedBy this siteEd Zitron
Held record ez-anthropic-2025-net-loss-42bn
First assessment published after both reviews and the final-text check passed.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-anthropic-2025-lost-over-8bn
The openai review abstained.
2026-10-28
Assessment publishedBy this siteEd Zitron
Held record ez-ai-three-quarters-to-prove-itself
First assessment published after both reviews and the final-text check passed.
2026-10-28
HeldBy this siteEd Zitron
Held record ez-ai-services-majority-token-billing-by-end-2026
The openai review found the criteria unfair to the speaker's words.
2026-10-28
HeldBy this siteEd Zitron
Within a few months of 2025-10-20, AI-native companies (products built on top of AI models) will face multiple price increases.
The final-text check failed: rationale_supported_by_evidence; overclaims: Settling it would need an end date and a definition that he did not give, so under these criteria no evidence available by the cutoff can settle it..