Mathematicians Issue Declaration: AI Solving Open Problems as Benchmarks Harms Mathematics

Thom_Wolf · x · 2026-09-12

The mathematical community published a "Math and AI" declaration (amplified by Thom Wolf) arguing that while LLMs can now solve major open problems, AI companies using them as benchmarks is detrimental to mathematics — the goals of AI companies and mathematicians are severely misaligned, part of a broader alignment issue for science and society.

Core argument: famous problems serve as lighthouses; their value lies in the new insights and methods they generate, refined through years of community discussion into textbooks. Models riffling to solutions strip out that process.

Wolf's pithy take: AI solving these problems yields 1 bit of signal while killing the 99 other bits of conceptual understanding mathematicians would have extracted from wrestling with them.

Related event: 25 Fields Medalists Led by Terence Tao Sign Open Letter Warning of AI's Severe Misalignment in Mathematics(79 posts)→

Original post →

More from AGI Musings

AGI Musings channel →