signal量子位2026-10-05
Hinton Just Published His First RSI Paper
Hinton and 21 other authors, including Yoshua Bengio, published a paper titled "What if automating AI R&D triggers an intelligence explosion?" examining recursive self-improvement. Data cited from Anthropic shows AI-generated approved code rose from low single digits in January 2025 to over 80% by May 2026, while Claude's autonomous R&D work share increased from 1% in March 2026 to 26% in August. The paper estimates that full automation of AI R&D could accelerate progress tenfold in about 1.5 years, compressing a year of progress to roughly five weeks.
- for who
- Researchers and practitioners focused on AI recursive self-improvement and the risk of an intelligence explosion
- what changes
- AI R&D automation could expand lab research teams from thousands to millions of AI researchers and drastically shorten development cycles, though the paper stresses current evidence is insufficient to confirm an explosion has begun
- to do
- Read the paper and track AI R&D automation trends to assess their impact on your research field
key points
- Hinton and 22 authors publish first RSI paper on AI-driven intelligence explosion
- AI-generated code share jumps from low single digits to over 80% in a year
- Claude's autonomous R&D work share rises from 1% to 26% in six months
#hinton#rsi#intelligence explosion#ai research automation
score
score 6 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source