25 Fields Medalists Criticize AI Labs’ Rush to Announce Math Proofs

Twenty-five Fields Medalists have signed an open letter criticizing AI labs’ competition to solve famous mathematical problems. They argue that the race to announce results can move faster than mathematicians can verify, explain and credit the work behind them.

One OpenAI mathematical proof cited in the current debate remains unverified, according to TechCrunch. The letter says rushed announcements can come before a proper writeup, the identification of new methods and ideas, and citations to relevant earlier work. Those steps, the signatories argue, are how a proposed result becomes something other mathematicians can assess and build on.

Often these solutions are announced in a rush, leaving no time for a proper writeup, the isolation of new methods and ideas, and citing relevant previous work of others.

The Fields Medalist signatories, in their open letter

The dispute is not limited to whether a particular proof is correct. The signatories say AI-conceived ideas still need mathematicians to develop and integrate them into the mathematical canon, the accumulated work that researchers use and teach. Without that work, they warn, the human transmission chain between mathematicians could be lost.

What a rushed announcement can leave unfinished

• A proper written account of the proposed solution.
• Identification of the methods and ideas that matter.
• Citations and credit for relevant previous work.
• Development and integration of the idea into mathematical literature.

The letter also points to the imbalance created by well-funded frontier labs. The mathematicians warned that a lab that finds a promising path to a discovery can spend tens of millions of dollars on large-language-model inference to beat the original researchers to a proof. They say that dynamic could encourage secrecy around promising research paths.

The intervention follows the Leiden Declaration, released in June by a mathematicians’ working group. That document offered recommendations for mathematicians, institutions and policymakers on how language-model proofs could change the field. The new letter adds the authority of 25 Fields Medal recipients to a related argument: AI may produce valuable results, but their value depends on whether people can understand, communicate and incorporate them into the discipline.