Terence Tao: AI Labs Racing Math Benchmarks Are Starting to Hurt Mathematics
Old-School8916 · reddit · 2026-09-07
Terence Tao posted an eight-post thread warning that AI labs racing to beat math benchmarks is damaging the field. After Stadlmann's preprint lowered the prime gap bound from 246 to 240 on Aug. 31, several labs rushed to post their own improvements. Tao argues the number itself never mattered — what mattered was Zhang reviving equidistribution work, Maynard's sieve becoming a standard tool, and Polymath8's open collaboration. He sketches a counterfactual where 2026-era labs crush GPY's near-miss, burying insights like Maynard's sieve in unread AI output and costing mathematics far more than the improved bound. He calls it a Goodhart's law problem, cites Erdős problem 126 as a positive AI-collaboration case, and proposes labs instead compete on who announces a genuinely new mathematical insight first.
More from AGI Musings
- New paper finds AI agents negotiate badly under uncertainty and break deals on purpose — xuanalogue · 2026-10-26
- NBER: Only 2.9% of US consumers pay for AI; retention as low as 20%, pricing power at risk — avicgoldfarb · 2026-10-26
- AI researchers guessed robots wouldn't outrun humans until 2034 — it already happened — ben_j_todd · 2026-10-26
- AI folks, stop saying "biology" when you mean drug development — KevinKaichuang · 2026-10-26
- ESR on the Bugpocalypse: locking down frontier models only hands the game to Chinese open weights — colinmcnamara · 2026-10-26
- Anthropic: junior tech hires fell from 27% to 9% as agent hours hit 6x employee hours — thione · 2026-10-26