OpenAI Releases Hundreds of AI-Generated Maths Papers
OpenAI has published 722 papers in which an AI model explores a wide range of mathematical questions. The results have not yet been fully reviewed or computer-verified, and three papers have already been withdrawn. Some work addresses major Millennium Prize problems, including a weaker version of the Riemann hypothesis and the Birch and Swinnerton-Dyer and Hodge conjectures.
OpenAI published 722 papers on 7 October, generated by an AI model exploring many mathematical areas. The outputs range from famous open problems to tiny algorithmic gains. Because many are long, unformalised and not yet checked by computers, accuracy remains uncertain; three papers were withdrawn.
Several papers target Millennium Prize questions, including a weaker Riemann hypothesis variant and partial cases of the Birch and Swinnerton-Dyer and Hodge conjectures. Another claims to extend a Kakeya conjecture result to four dimensions. Others report marginal speedups, such as a theoretical multiplication improvement and modest Fourier-transform gains.
If these results withstand scrutiny, mathematicians and computer scientists could gain new tools and directions, potentially speeding research in cryptography, physics and computation. Yet the volume of unverified, partly unformalised work may strain peer review and amplify uncertainty. Universities, journals and funding bodies may need clearer standards for evaluating AI-generated proofs, while the public could encounter premature claims about solved famous problems. The immediate impact likely depends less on the papers’ number than on independent verification.