Mathematician Daniel Litt: AI grinds computations well but can't build new math theory
stuffyokodraws · x · 2026-09-03
University of Toronto mathematician Daniel Litt and a16z's Lisha Li discuss AI's impact on mathematics. Key points:
- Models are good at a much narrower slice of math than headlines suggest: grinding long computations, pulling technical ideas from more papers than any human could read, and applying known techniques better than almost anyone.
- What they can't do is build theory, or hold a vague philosophy long enough to make it precise — most of what Litt does for a living.
- The conversation covers how mathematicians raided an AI proof for parts to break other problems, and why a thousand AI mathematicians might all turn out to be the same mathematician.
Takeaway: AI is a powerful computation and retrieval tool, but theory-building remains a distinctly human mathematical skill.
More from Models
- Mocked benchmark shows DeepSeek's document chart grounding badly misaligned — burkov · 2026-10-18
- Dev complains Claude's habit of saying "not done" is his biggest pain point — jasonkneen · 2026-10-18
- Open-weight TYGER speech decision model makes 80ms turn-taking calls for full-duplex voice agents — solyarisoftware · 2026-10-18
- Paying Google AI subscriber blocked: custom-domain Workspace accounts can't access AI Studio — ChrisRackauckas · 2026-10-18
- Running Ling 3.1 Flash (560B MoE, 25B Active) Across 4 DGX Sparks via llama.cpp RPC — BerryBlush_Rogers691 · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18