25 Fields Medalists Sign Declaration Warning Against Measuring AI Progress by Solved Problems

krishnan · x · 2026-09-17

A declaration signed by 25 Fields medalists, published by IMPA, argues the wrong unit is being used to measure AI progress. While accepting LLMs have made dramatic strides, the signatories object to treating famous problems as a scoreboard: research also produces judgment—which questions matter, why methods work, whether the next generation can build on results. They warn that when output becomes cheap, organizations mistake artifact volume for institutional learning. For AI-for-science labs, they recommend publishing provenance, checkable intermediate reasoning, verification status, and teachable expositions, and measuring time to community verification and reuse rather than time to first result.

Original post →

More from AGI Musings

AGI Musings channel →