OpenAI tries again to fix broken trust with mathematicians

OpenAI has a pattern: its artificial intelligence systems make real breakthroughs in solving complex mathematics problems. But then the company announces these achievements in ways that frustrate and upset the mathematicians who work in this field. Over the past few months, this has happened repeatedly, damaging OpenAI's reputation among people who actually study and work on math problems for a living.

Now OpenAI is attempting damage control. The company says it will consult with an independent advisory group made up of top mathematicians before making announcements about math-related achievements. The idea sounds sensible: talk to the experts first, get their feedback, then tell the world.

But here's the problem: even this attempt to fix things is turning into a mess. Mathematicians who have spoken to reporters—including someone actually on the new advisory group—describe the whole process as confused and disorganized. It has many of the same red flags as OpenAI's previous rushed announcements that caused the problems in the first place.

The larger issue is about respect. These mathematicians feel that OpenAI treats their field like a box to check, rather than genuinely engaging with the community it affects. Until OpenAI changes how it listens and communicates, creating more advisory groups probably won't help.