Hinton, Bengio and 20 top researchers warn an AI intelligence explosion may be months away
新智元 · wechat · 2026-10-03
A report by 22 leading AI figures — including Geoffrey Hinton, Yoshua Bengio, OpenAI's chief scientist Jakub, Anthropic co-founder Jack Clark and Meta's Dawn Song — was released on Cambridge's CASP platform. It documents soaring AI R&D automation: Anthropic's AI-written code went from single digits in early 2025 to over 80% by May 2026, fully AI-autonomous R&D rose from 1% to 26% in five months, with 30,000 agents making over a billion decisions in August alone; OpenAI and Google report similar figures. The core thesis: the bottleneck shifts to "AI R&D automation rate," creating a feedback flywheel — dedicating frontier inference compute to R&D agents would equal 2–20 million top scientists. With estimated returns-to-research r of 1.2–1.9, fully automated AI R&D could speed technical progress 10x within 17 months, compressing a year of progress into five weeks. The report warns of risks arriving years early, humans losing control (up to marginalization or extinction), and broken checks-and-balances, calling for circuit breakers, sandboxing and societal preparation.
More from AGI Musings
- Blaise's new book hailed as year's best; his space-GPU Project Suncatcher just launched — paulnovosad · 2026-10-23
- If LLMs become the universal interface, who indexes the MCP services? — ThomasBuildLab · 2026-10-23
- Bindu Reddy: bearish on superintelligence until robotics breaks the digital loop — bindureddy · 2026-10-23
- AI layoffs arrive: Barclays cuts 5,000, eBay 9%, Amex 3,500, Netflix ~1,000 — MFordFuture · 2026-10-23
- 632 tests find no reliable link between AI search visibility and branded or direct traffic — lilyraynyc · 2026-10-23
- Random programs reliably evolve into self-replicators: why 'life is computation' may reshape how you think — paulnovosad · 2026-10-23