Why mathematicians have a point about AI proofs: models hunt trophies, not exploration
_lewtun · x · 2026-09-17
- The quoted thread argues mathematicians' anger at AI solving their problems isn't mere gatekeeping: part of math's joy is exploring the territory around a proof, not just the answer.
- Commenters noted two things: for practical tasks a "straight line" AI solution is still optimal, and mathematicians remain free to wander around the solution (possibly with the same AI's help).
- The author explains why that exploration can't happen today: models are trained to nuke hard problems rather than marvel at the path; AI companies collect solutions like hunting trophies with little incentive to prioritize exploration; and the academic ecosystem doesn't reward post-solution wandering.
- Still, none of this is inevitable — the author hopes collaboration between mathematicians and open-source AI builders can enable "post-solution" exploration.
More from AGI Musings
- Long-read: AI's real value comes after the efficiency era—infra glut, agent tooling boom — yangyi · 2026-11-02
- Security's attack-defense automation gap mirrors math: proofs are easy, conjectures are hard — mboehme_ · 2026-11-02
- vikhyat predicts alternating compute oversupply and undersupply on the road to AGI — TheZachMueller · 2026-11-02
- Sam Altman: AI has surpassed top scientists, and 2026 will be a turning point year — kimmonismus · 2026-11-02
- DeepSeek design doc: a person can only follow 3-5 agent sessions at once, attention is the real bottleneck — heyshrutimishra · 2026-11-02
- AI researcher: we have no reliable way to tell whether an AI system is conscious — dioscuri · 2026-11-02