25 Fields Medalists including Terence Tao declare AI benchmark-chasing severely misaligned with mathematics
sytelus · x · 2026-09-12
- Terence Tao and 24 other Fields Medalists issued a joint declaration on Math and AI at mathandai.org, open for additional signatories.
- Core claim: LLMs can now solve major open problems, but AI companies' push to use math problems as benchmarks is detrimental to the science of mathematics — the goals of AI companies and the mathematical community are severely misaligned.
- Famous problems serve as landmarks; their value lies in community digestion of new ideas, textbook simplification, and eventual broad diffusion — a process at odds with benchmark-chasing. The authors frame this as part of broader alignment issues affecting science, creative professions, and society.
- The retweeter complains that no one who ever turned down a Fields Medal has signed.
More from AGI Musings
- Ethan Mollick: METR long-horizon benchmark effectively saturated as agents hit 18-week tasks — emollick · 2026-09-12
- ResNet by hand: a 10-step walkthrough showing how one identity matrix enabled thousand-layer networks — ProfTomYeh · 2026-09-12
- On AI's punch-card phase: most programmers never touched a real computer until the 1970s — _AustinCalvert_ · 2026-09-12
- awilkinson: We're still in the punch cards phase of AI — and it's only year 3 — _AustinCalvert_ · 2026-09-12
- Investor: last time 'we're all going to die' alarmists won, it cost hundreds of thousands of lives via anti-nuclear fear — firstadopter · 2026-09-12
- John Carmack slams big-lab doomers: "If you believe in risk, work in the open" — beffjezos · 2026-09-12