Noam Brown: LLMs have crossed the threshold, beating top human experts on some research problems
polynoamial · x · 2026-10-08
OpenAI researcher Noam Brown argues LLMs have crossed an important threshold: surpassing top human experts on some research problems. In scientific discovery, the difference between a model just below and just above top human performance is qualitative, which explains why the recent surge in math results feels so sudden.
He acknowledges LLM capabilities remain jagged, but expects breakthroughs in other domains to follow, ushering in a new era of scientific discovery. Quoted Mark Chen calls the run 'a decade of mathematical progress in a week' and says the tools will next be pointed at life sciences, next-gen models, and alignment.
Related event: OpenAI's 372 Math Proofs Spark Divided Reactions(81 posts)→
More from AGI Musings
- Anthropic Study: AI and Robots Can Already Cover 81% of U.S. Employment Tasks — Polymarket · 2026-10-08
- GPT's Math Wins Credit Thousands of Collaborators, Not One Mind — sandersted · 2026-10-08
- Reddit proposal: keep AGI from learning to code and use humans as the safety bottleneck — meira_xxx · 2026-10-08
- repligate mocks Paul Christiano's post-AGI vision: neither possible nor good — repligate · 2026-10-08
- repligate: once model weights spread they can't be clawed back, and superintelligence may need to guard them — repligate · 2026-10-08
- DeepMind's Andrew Trask: AI heads for a mainframe-to-PC moment of millions of small models — iamtrask · 2026-10-08