Top mathematicians dismiss AI math wins as "just evals," sparking debate on AI-driven discovery
omarsar0 · x · 2026-11-23
Elvis Omarsar reflects on top mathematicians describing recent AI math breakthroughs as "just evals for AI researchers." Key takeaways: ambitious benchmarks matter for pushing verifiable domains—and eventually open-ended ones—but there's real concern that AI solutions aren't producing new human understanding. He endorses Terence Tao's "Copernican" view of coexisting intelligences, arguing these math milestones are its first signs, and urges practitioners to adapt by building with AI agents and developing evals across domains.
Related event: Terence Tao: AI Is Winning Math Competitions but Teaching Us Almost Nothing(5 posts)→
More from AGI Musings
- AI beat humans at chess—then Chess.com exploded to 277M members — adamamcbride · 2026-11-23
- "Where are the philosophers?" Crypto veteran casts open-weight local models as a sovereignty movement — seanmcdonaldxyz · 2026-11-23
- Elon Musk declares the technological Singularity has arrived — Polymarket · 2026-11-23
- AI can't do the impossible: exposed IT weaknesses were always there, argues researcher — JHochderffer · 2026-11-23
- What rights do we owe a machine? Inside AI labs' debate over machine consciousness — jeremyakahn · 2026-11-23
- Mathematician claims Claude may have plagiarized his Riemann proof work, writes 16-page letter to Amodei — ValerioCapraro · 2026-11-23