Csaba Szepesvari says AI labs' marketing-value math benchmarks will soon run dry
CsabaSzepesvari · x · 2026-09-13
RL researcher Csaba Szepesvari argues AI companies will soon exhaust significant open math problems of marketing value and will stop using them as benchmarks — the sooner the better. Quanquan Gu disagrees narrowly, saying he isn't convinced labs should be asked to stop treating important math problems as benchmarks. The exchange continues Szepesvari's critique that a recent letter's sole request (stop benchmarking unsolved math problems) isn't working.
More from AGI Musings
- Amid AI math discourse, Thurston's 1994 'On Proof and Progress in Mathematics' resurfaces — keenanisalive · 2026-09-13
- David Deutsch: Stop measuring knowledge or creativity, remove constraints instead — NinaDSchick · 2026-09-13
- Critics Claim Anthropic and OpenAI Are Building a 'Legal Cartel' via Safety Push — AlexTensor · 2026-09-13
- Radiologist AI Debate Shows People Think in Memes, Not About Labor Reallocation — AndyMasley · 2026-09-13
- The End of Sisyphus: AI won't free us, it forces us to choose which mountains exist — YogeshMalik · 2026-09-13
- Yishan Defends Dario/Sama's AI Pacing Plan Against 'China Will Win' Objections — yi_ding · 2026-09-13