Mathematicians get a say in how OpenAI reports its math

The company that pulled out of a Caltech math hackathon is now asking mathematicians to advise it on how it talks about math. An independent advisory group will weigh in on how OpenAI assesses and communicates new mathematical results, the company said on X.

At a glance

• OpenAI's post lists three jobs for the group: judging and communicating new mathematical results, upholding academic and professional standards, and building tools that support mathematical research and learning.
• The stated goal is to put mathematicians at the center of shaping how AI supports mathematical understanding, and how the benefits of that work reach the wider community.
• The post names no members, does not say whether the group sees results before publication, and gives no schedule for when its advice starts.

If you missed the argument: Caltech mathematicians put their names to an open letter against a mathathon at the school, saying the event is likely to have destructive impacts, according to Reddit. A day later OpenAI stopped sponsoring it.

The group is described as independent of OpenAI, and its remit covers both how the company behaves and what it ships. On the behaviour side sit the assessment of new mathematical results, the way those results are communicated, and academic and professional standards. On the shipping side sits a single line about tools for mathematical research and learning.

Two words in that brief carry most of the weight. Assessing is the part where someone decides whether a model really produced a new result. Communicating is the part where that decision turns into a post, a paper or a line on a chart. In mathematics the two steps are normally kept apart: a proof goes to referees who did not write it before a journal prints it.

The third strand is building rather than judging, and OpenAI gives it one sentence. The company frames the whole arrangement as a question of who shapes the technology, saying it wants mathematicians at the center of how AI supports mathematical understanding and how the benefits of that work spread beyond the field.

What the announcement does not carry is a roster. No member is named, nothing is said about whether the group looks at a claimed result before OpenAI publicizes it, and the independence of the group is asserted rather than spelled out. In our view that is a thin remit for a body whose whole value would be vouching for the kind of claim that set off the Caltech row in the first place.

Who joins, and when

The checkpoint worth waiting for is the membership list. An advisory group means very little until you can see whose names are on it, what they are shown and at what point. OpenAI has not said when the group begins work, how often it meets, whether its advice will be published, or whether it has any say over announcements the company has already scheduled. None of that is dated in the post.