AI labs report major math results and spark backlash over process

In short: AI labs are announcing solutions to difficult math problems, but their fast, competitive approach is causing conflict with mathematicians.

What's going on

Over the past year, OpenAI, Anthropic, and other AI labs have reported progress on several long-standing math problems. In at least one case, OpenAI said it solved a Millennium Prize problem, a set of seven famous math challenges that each come with a $1 million reward.

These announcements would normally be a cause for celebration in academic math. Instead, they have often turned into public disputes. According to reporting from The Verge, critics say some AI labs rush announcements, do not follow common academic habits, and create confusion about what has been proven and what still needs checking (like claiming a bridge is safe before the final inspection).

Several specific concerns keep coming up. Mathematicians have questioned whether AI systems relied on other researchers’ work without clear credit, including unpublished ideas. There have also been allegations around “scooping,” which is when someone tries to publish first mainly to beat others to it.

OpenAI has responded by setting up an independent advisory group of prominent mathematicians. The group is meant to help guide how results are presented and released. But some mathematicians told The Verge that the effort itself has been messy, and they still have basic questions about how much power the group will have.

What to watch

A big open question is whether AI labs will slow down and work more closely with the math community, especially on peer review, credit, and sharing details. Another is what happens if OpenAI releases many more claimed results at once, which researchers fear could overwhelm the normal process for carefully checking proofs.

Source: The Verge AI