OpenAI's math reputation crisis

OpenAI turned spectacular mathematical breakthroughs into a reputational disaster through poor communication and announcement practices. The company is now consulting elite mathematicians to chart a better course.

Meet the advisory group

OpenAI announced a nine-member independent panel called AGMAI, stacked with Fields Medal winners and MacArthur fellows from Stanford, Harvard, Oxford, and Cambridge. The group is hosted by Princeton's Institute for Advanced Study.

What the panel will actually do

The group will advise OpenAI on reviewing and communicating mathematical results, assessing their significance, coordinating their release, and maintaining academic standards. Members get early access to OpenAI's research but receive no payment.

“The group will advise on the review and communication of emerging results.”

Independence claims raise questions

OpenAI says the panel operates independently with freedom to publicly challenge the company, but researchers are skeptical about how members were selected and why the group was formed in the first place.

Panel member defends the effort

Martin Hairer, an Imperial College London professor on the panel, wrote that the group is genuinely independent and that engaging with AI labs makes more sense than ignoring them. He stressed members face no financial incentives or restrictive agreements.

“The other fundamental fact is that our overarching aim is to represent as well as we possibly can the interests of the mathematical community.”

Concerns about elite bubble

Critics worry the nine elite academics don't represent ordinary mathematicians whose careers and research are being disrupted by OpenAI's push for trophy results. One researcher called it an ivory tower approach.

Different realities for different mathematicians

Researchers at elite institutions face a vastly different reality from most mathematicians. Panel members may not fully understand or speak to the concerns of the broader community being affected by AI.

The PR spin problem

Even panel member Hairer acknowledged that AI companies will likely try to spin the group's advice for their own PR benefit, which dwarfs anything the mathematicians could generate.

“We are very naïve to believe that the AI labs won't try to spin whatever we say in a way that suits their PR machine.”

The real problem to solve

AI companies have recently produced high-profile mathematical results with publication and dissemination falling far short of acceptable mathematical practice. The panel must address this fundamental issue.

The logjam ahead

OpenAI casually announced its unreleased model has resolved over one hundred long-standing open problems across mathematics. The advisory group's first priority is coordinating the release of these results.

“We are currently facing the very specific challenge of advising OpenAI on how to coordinate the release of a large number of significant results.”

Past missteps that sparked this

OpenAI faced accusations of scooping researchers, failing to credit human work its models built upon, and even stealing from users. These controversies raised doubts about whether the company understood or respected mathematics.

A simpler message might suffice

Some mathematicians question whether OpenAI needed an elite committee at all. The community's main concern wasn't about proving hard theorems but about the company respecting the discipline and its people.

Trust will determine success

Whether the panel can earn mathematicians' trust will determine how effective it is at containing fallout from OpenAI's missteps and softening the blow for future announcements.

Gathering community input

The panel is collecting feedback through conversations with colleagues, a public feedback form, and discussion on Proofs and Prompts, a blog that has become a key forum for mathematicians reckoning with AI's impact.