OpenAI's Math Claims Spark Mathematician Backlash

OpenAI has spent the past several months proving it can produce genuine mathematical results and then mishandle the announcement of those results almost every single time. The pattern has repeated often enough that mathematicians who work adjacent to the company describe a kind of fatigue, watching breakthroughs get rolled out in ways that confuse or alarm the people meant to evaluate them. OpenAI's latest fix is a new independent advisory group made up of practitioners from the field, meant to smooth over relations with a community it has managed to rankle more than once.

The problem is what the group has been asked to do first: help manage the release of a large batch of new results OpenAI says its unreleased model has generated, a prospect that has researchers bracing for disruption rather than celebrating progress. The Verge spoke with mathematicians familiar with the situation, including someone inside the advisory group itself, and the picture that emerges looks a lot like OpenAI's earlier stumbles into this territory. Formed on September 21st under the name AGMAI, the group was introduced through a blog post that, according to the people involved, did not exactly inspire confidence.