25 Fields Medalists Against Turning Mathematics into a Benchmark for AI

25 laureates of the Fields Medal — the highest honor in mathematics — have published a joint declaration titled "Math and AI." Among the signatories are Terence Tao, Peter Scholze, Pierre Deligne, Alessio Figalli, and other renowned mathematicians.

The trigger was the rapid growth in the capabilities of language models. Over recent months, the authors note, AI has learned to solve major open problems across various areas of mathematics, and these successes make headlines even beyond scientific circles.

The complaint is about misplaced goals. For AI labs, solving famous problems has turned into a benchmark, whereas the goal of mathematics is not a ready answer, but understanding and new ideas.

The authors remind us: famous problems have historically served as landmarks. Solving them was a sign of new methods and ideas that the community would then spend years digesting — in talks, discussions, and textbooks.

"Mass production" of proofs at an accelerated pace, by contrast, threatens to destroy the living culture of the science. Solutions are often announced in haste, without full derivations or references to predecessors, which raises questions of attribution.

In the authors' view, without living mathematicians to develop AI's ideas and weave them into the scientific canon, the chain of knowledge transmission will be lost. The problem, they emphasize, is broader than a single discipline — it is part of the general threat to intellectual labor in science and the creative professions.

Image: This SVG image was created by Medium69 .
Cette image SVG a été créée par Medium69 .
Please credit this : William Crochot CC BY-SA 4.0 · License

Image: Ivonne Vetter CC BY-SA 2.0 de · License