A group of 25 Fields Medalists has issued a statement criticizing the current trend of AI companies racing to solve complex mathematical problems. Signatories, including Terence Tao, Peter Scholze, Maryna Viazovska, and Maxim Kontsevich, argue that prioritizing rapid results over scientific methodology threatens both research and education.
The mathematicians emphasize that scientific progress relies on understanding methods rather than just obtaining binary true or false answers. They warn that the rush to announce AI-generated proofs prevents proper peer review, attribution, and the integration of new findings into existing mathematical frameworks, potentially stifling the development of new ideas.
The statement follows OpenAI's announcement regarding the Navier-Stokes equation, one of the Millennium Prize problems. OpenAI reported that a system of approximately 10,000 agents found a solution in 88 hours, with an additional 17 hours for formalization in Lean. The company stated it would not claim the Clay Institute prize.
The OpenAI announcement sparked controversy regarding the potential use of unpublished work by mathematicians Tristan Buckmaster and Levent Alpge. While OpenAI conducted an internal review and denied any influence from Buckmaster's queries, the incident highlighted growing tensions regarding data sharing and intellectual property in the mathematical community.
In a separate development, Anthropic recently announced that its Claude model formalized a proof for Fermat's Last Theorem. The process, which involved translating existing mathematical reasoning into code for computer verification, took 11 days. The Fields Medalists noted that such AI applications depend heavily on human management.
