AI-rewritten: This is a summary of an article from New Scientist, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

On October 7, OpenAI released 722 mathematical papers generated by an AI model covering topics from major mysteries to minor curiosities. Many of these documents are hundreds of pages long and contain complex computations that have not yet been formalized into steps verifiable by a computer. Consequently, there is little assurance regarding their accuracy, though three papers have already been retracted. Despite this uncertainty, preliminary reviews suggest some results could be significant.

Several papers address Millennium problems, seven challenges considered among the most important in mathematics, each offering a $1 million prize from the Clay Mathematics Institute. Since 2000, only one has been solved. OpenAI claims its model has solved a problem related to the Navier-Stokes equations of fluid dynamics. The company also targets the Riemann hypothesis regarding prime numbers by proving a variation called the quasi-Riemann hypothesis, which deals with how far primes deviate from their expected positions. Additionally, papers claim partial proofs for the Birch and Swinnerton-Dyer conjecture and the Hodge conjecture, while another purports to expand a solution to the Kakeya conjecture into four dimensions.

Other results include minuscule improvements in algorithmic power, such as claiming integer multiplication could be performed 2⁻¹⁸² faster than previously thought. While this represents an improvement smaller than one divided by 14 million billion billion billion times the age of the universe, other speed gains for Fourier transforms are slightly more significant. One paper also suggests that coloring points on a plane so no two adjacent points share a color requires more than five colors, narrowing the known range to between five and seven.

Experts warn it could take years for mathematicians to verify this large volume of work. If AI mathematics continues at this pace, it may present either a major benefit or a serious problem for the field, but no prediction is possible until the results are validated.