Claim Check
Lee Sedol, world-champion Go player (9-dan professional) · Interview with the Associated Press, Seoul, before the AlphaGo match · Source
What happenedA sceptical forecast: Lee predicted he would win 5-0 or 4-1. The machine won 4-1, so the score he predicted happened, but in AlphaGo's favour.
John Zimmer, Lyft co-founder and president · Essay 'The Third Transportation Revolution' (Medium), reported by PCWorld · Source
What happenedBy the 2021 deadline Lyft had sold its self-driving unit. Four years later its autonomous rides were still small pilots with safety staff on board.
Geoffrey Hinton, computer scientist (University of Toronto and Google) · Creative Destruction Lab conference 'Machine Learning and the Market for Intelligence', Toronto (video: 'Geoff Hinton: On Radiology') · Source
What happenedHe said deep learning would beat radiologists within five years, maybe ten. Ten years on, there are more radiologists, not fewer, and Hinton himself has said he was wrong about the timing.
Ray Kurzweil, futurist and director of engineering at Google · SXSW, Austin, and a statement to Futurism · Source
What happenedDue in 2029. Chatbots already pass short Turing tests. Kurzweil's 'valid' test and 'human-level intelligence' set a higher bar, so this one stays open.
Katja Grace and co-authors (AI Impacts, Oxford, Yale), survey of machine-learning researchers · Paper 'When Will AI Exceed Human Performance?' (arXiv; survey run in 2016) · Source
What happenedThis was the researchers' median forecast that AI would beat humans at high-school essays by 2026. It happened by 2023, earlier than forecast. The evidence is one study of argumentative essays.
Elon Musk, Tesla CEO · Tesla Autonomy Day, Palo Alto · Source
What happenedMusk promised a million robotaxis by 2020. The first small robotaxi service started more than five years later, and Bloomberg counted dozens of cars, not a million.
Sam Altman, OpenAI CEO · Essay 'Moore's Law for Everything' · Source
What happenedWithin five years, AI systems were routinely reading legal documents at law firms and answering health questions for hundreds of millions of people. The quality of that 'medical advice' is still debated.
Gary Marcus, cognitive scientist and AI critic · Post on X listing predictions for the end of 2024 · Source
What happenedHe was right that no GPT-5 came in 2024 and that rival models bunched together. But o1-style reasoning models made a large jump on hard maths in late 2024, so 'no massive advance' is disputed.
Elon Musk, CEO of Tesla and xAI · X Spaces interview with Nicolai Tangen, Norway's sovereign wealth fund (reported by Reuters) · Source
What happenedHis main forecast was 'probably' 2025, with 'within two years' as a hedge. In 2026 Musk himself kept placing AGI in the future, so by his own account it has not arrived.
Sam Altman, OpenAI CEO · Blog post 'Reflections' · Source
What happenedAI agents did start doing real work in 2025, especially coding agents, and some firms reported big output gains. Across most companies, use was still narrow and the profit impact small.
Mark Zuckerberg, Meta CEO · The Joe Rogan Experience, episode #2255 (reported by Business Insider) · Source
What happenedThere is no agreed test for 'a sort of mid-level engineer'. Meta reported that AI tools made its engineers more productive, not that an AI replaced one, so the claim cannot be cleanly scored.
Demis Hassabis, CEO of Google DeepMind and Isomorphic Labs · World Economic Forum panel, Davos (reported by The Register) · Source
What happenedIsomorphic's AI-designed drugs were not in trials by the end of 2025, and Hassabis moved the target to the end of 2026. (Other companies' AI-assisted drugs were already in trials, but this forecast was about Isomorphic.)
Yann LeCun, Meta chief AI scientist · World Economic Forum session 'Debating Technology', Davos (reported by TechCrunch) · Source
What happenedA sceptic's forecast, due between 2028 and 2030. He also said that by then 'nobody in their right mind would use them anymore' as the core of an AI system.
Dario Amodei, Anthropic CEO · Council on Foreign Relations, CEO Speaker Series, New York · Source
What happenedAmodei later said the forecast was true inside Anthropic and some partner companies, then added 'not uniformly'. Across the software industry, the share stayed far below 90%.
Dario Amodei, Anthropic CEO · Interview with Axios, 'Behind the Curtain: A white-collar bloodbath' · Source
What happenedThe window runs to 2030. So far there is a clear early hit to young workers in AI-exposed jobs, but nothing near half of entry-level jobs or 10–20% unemployment.