A warning on using AI in mathematics

gerardsans · x · 2026-07-20

A reminder that mathematicians experimenting with AI should understand its current limits.

The quoted thread argues that AI can help with some math tasks, but users should verify every suggestion because the model’s outputs are partial, biased by training data, and not combinatorially complete or grounded. In open-ended problems where solutions may not exist and verification is expensive, AI can quietly waste time and budget without warning.

Related event: AI in Research: High Verification Costs and the Expert Sweet Spot(8 posts)→

Original post →

More from AGI Musings

AGI Musings channel →