Proofs vs. explanations: what would it take for AI to replace mathematicians?
Alchemistwiza · reddit · 2026-10-14
Following the 'AI will replace mathematicians' debate, the author draws a key distinction: finishing a proof isn't finishing the math. Anthropic's Sept 4 Fermat project formalized a proof with an internal Claude model, but its own write-up says formalization can't replace human-understandable exposition; OpenAI's Oct 6 math release likewise plans to improve presentation. Verified results are valuable, but explanations let others reuse ideas. Proof generation alone won't obsolete mathematicians — reliable explanation, connection of results, and problem selection would be the stronger challenge. The author asks what demonstration would change minds.
More from AGI Musings
- banteg: creating out of resentment vs creating out of love — the PS5 hacker essay — banteg · 2026-10-14
- Mathematician Sylvia Serfati: most OpenAI results would deserve a Fields Medal — coherence · 2026-10-14
- "Hitting the bullseye doesn't make you a good shooter": an AI proofs debate — gerardsans · 2026-10-14
- Humans on AI: possible power-seeking incidents, White House AI order, RL at 63% of Reflection's run — gleech · 2026-10-14
- Data center panel asks 'Are We Screwed?' as Gary Marcus says AI industry may have screwed itself — GaryMarcus · 2026-10-14
- "Everything is now a token": ML vocabulary is being rewritten by the LLM paradigm — jongranskog · 2026-10-14