How AI Is Challenging the Culture and Future of Mathematics

Summary

AI systems are rapidly producing proofs of longstanding mathematical conjectures, prompting a crisis over what mathematicians value and how the field should operate. The article focuses on a September 8 announcement by OpenAI that its models had solved a famous mathematical problem, an event that intensified disputes about authorship, standards and the role of human researchers. Mathematicians can verify many AI proofs with formal systems, but often cannot understand why the arguments work: the proofs may omit important connections, bury key details and provide little sense of how the result fits into existing literature. That gap threatens a culture built around understanding, creative problem-solving and open exchange, while current career incentives still reward theorem-proving above interpretation or explanation. Researchers report that companies including OpenAI and Anthropic are competing for mathematicians, withholding more than 100 results in OpenAI’s case, and contributing to fears that open science is being damaged. AI-assisted papers are overwhelming journals and arXiv, some mathematicians have stopped sharing conjectures, and MathOverflow activity has declined as students ask models questions privately. The article presents two broad futures. Without institutional change, mathematicians could become button-pushers who cannot read the outputs or train successors with classical mathematical skills. With new incentives, AI could handle more mechanical discovery while mathematicians interpret proofs, create theories, formulate conjectures, rewrite opaque arguments and build communities around ideas. The field is already reconsidering student training, publication standards and hiring, including experiments with requiring talks alongside papers. The author notes that this future may be less attractive to people drawn to original creative authorship, but describes emerging efforts such as the Association for Human Mathematics and argues that mathematics could become stronger if it protects understanding, communication and human participation in its culture.