Math PhD Trains AI to Crack Open Research Problems, Reports Progress in Arithmetic Physics
shuchaobi · x · 2026-10-03
A mathematics-trained researcher describes shifting from Olympiad-style problems to open research questions. Realizing he lacked the raw genius to solve the problems he cared about, he moved into machine learning — and now focuses on training AI models to tackle those problems instead. Working with former grad-school friends who are now math professors, he reports meaningful progress on an open problem in arithmetic physics. He argues math is like a vastly expanded game of Go: structured environment plus intrinsic self-verification makes it ideal terrain for AI-driven discovery.
Related event: Math PhD Turns to Training AI to Tackle Open Problems, Makes Progress(2 posts)→
More from AGI Musings
- RSIArena Day 2: MiniMax joins late, GPT-6 Astra retrains from scratch after finding test leaks — my_cat_can_code · 2026-10-03
- NYU professor argues academics on social media should disclose financial conflicts of interest — ipeirotis · 2026-10-03
- Hinton: leading researchers now think an intelligence explosion may happen soon — geoffreyhinton · 2026-10-03
- Sutton shares Bitter Lesson country song as Mollick admits his agent-orchestration thesis was wrong — RichardSSutton · 2026-10-03
- Prediction: economical household robots will be hired by the job, not owned — jessi_cata · 2026-10-03
- Developer U-turn: frontier open-source models may be too dangerous to release — DrDatta_AIIMS · 2026-10-03