Mataton controversy at Caltech

A new event at Caltech called “Mataton,” organized by a group of undergraduate students to “promote the responsible use of AI in mathematics,” quickly became controversial last week.

In the first round, scheduled for October 30 to November 1 this year, 100 teams of mathematicians will work on open-ended math problems for 40 hours, with each team receiving approximately $20,000 worth of tokens to use in cutting-edge artificial intelligence models. The teams that make the most progress will advance to the second round, which will last six months. Organizers say the event aims in part to expose mathematicians outside artificial intelligence companies to powerful models.

However, on September 10, a coalition of current and former Caltech mathematicians published an open letter signed by more than 700 of their colleagues, including four Fields Medal recipients, asking for Mataton to be cancelled. The letter stated that Mataton’s accelerated format would encourage the use of “crappy” mathematics that other mathematicians would have to spend time and effort correcting and checking. Because OpenAI and Anthropic sponsored the event, the letter also accused artificial intelligence companies of turning open-source math problems into promotional competitions that drain math of its human and social resources.

Although the organizers subsequently changed some of Mataton’s rules, including requiring second-round participants to publish their work as preprints on the arXiv repository, they refused to cancel the event. However, OpenAI withdrew from the sponsorship; the website currently says Anthropic will provide $600,000 in loans.