25 Fields Medal Laureates Declare AI Companies Have a "Severe Misalignment" With Mathematics
Terence Tao, Peter Scholze, Marina Vyazovskaya, Cédric Villani, Maxim Kontsevich, and 20 other Fields Medal laureates have signed a declaration titled A Severe Misalignment of AI in Mathematics, warning that turning open mathematical problems into AI benchmarks is a dangerous path. The signatories argue that AI solutions are being published too quickly for proper verification, proof exposition is neglected, and new methods are not adequately distinguished from prior work, raising serious attribution and plagiarism concerns. The mathematicians do not deny recent AI progress in solving major open problems but stress that the value of mathematics lies not only in final proofs but in new ideas, explanations, and training the next generation. The declaration directly addresses the conflict of goals: AI labs optimize for speed and capability demonstrations, while the mathematical community prioritizes understanding, verification, and integration into the broader body of science.
Related: Terence Tao Warns of a Crisis in Mathematics Driven by AI, OpenAI’s Astra Solves 10 Longstanding Math Problems, OpenAI Model Autonomously Solves 80-Year-Old Geometry Conjecture