Fields Medal-winning mathematician issues warning against AI use

Summary

Twenty-five Fields Medalists, including Terence Tao and Peter Scholze, signed a joint statement warning that the growing role of AI in mathematics has created a 'serious mismatch' between companies' tendency to focus on benchmarks and the community's need for slow, shared understanding and proper attribution.

Why it matters

This rare joint statement from the highest echelons of the mathematical world indicates that the field's epistemic culture—peer discussion, citation, and transmission to students—is under structural threat from a different incentive logic. This shifts the AI-mathematics debate from 'can models solve problems' to 'what constitutes knowledge once a solution is produced.' This concern extends to any field built on cumulative, human-mediated understanding.

Highlights

• The statement does not ban AI. However, it conditions the acceptance of results as mathematical knowledge on proper writing, attribution, and integration.
• Three specific risks are highlighted: hasty publication due to the failure to isolate new methods, ambiguity in attribution and plagiarism, and the inability of ideas to enter the mathematical canon.
• Formal proof assistants like Lean 4 are positioned as a verification layer that distinguishes between 'solving a problem' and 'producing reliable and transferable mathematical knowledge.'

Critical take

The signatories' concern that benchmark-focused publication weakens mathematical culture is well-founded. However, the article reduces this nuanced stance to a broad anti-AI narrative. It then transitions to a Lean 4 setup guide, as if the deep epistemological questions regarding understanding, pedagogy, and transmission raised by the statement could be solved merely by formalization.