Caltech: AI Mathathon Held Despite OpenAI's Withdrawal

At Caltech, students are preparing a "mathathon" to explore the use of AI on theoretical problems. Despite an open letter with over 700 signatures and the withdrawal of a $1 million support from OpenAI, the event remains on the agenda, and its promoters assure that they want to regulate the use of AI.

OpenAI withdraws its support, event still scheduled for October 30

OpenAI has withdrawn its support for the event and reclaimed the equivalent of $1 million in tokens. Dan Roberts, head of research at OpenAI, stated on X that the rapid pace of AI advancements in mathematics is disruptive. Despite this withdrawal, the organizers are maintaining the event, which has been announced with adjustments, and have set it for October 30. They are in discussions with other AI companies to replace the canceled resources. OpenAI's withdrawal came after criticism regarding the supposed presence of "slop" in AI and OpenAI's initial $1 million backing.

Unprecedented technical resources sought from Anthropic and OpenAI

To launch AI agents on computation-heavy problems, the organizers have reached out to several companies, having little capital to start with. During a Zoom meeting, Caiman Moreno-Earle requested $1 million from Anthropic, explaining that he would have been satisfied with $50,000. Anthropic agreed to provide $1 million in computing power, quickly followed by a promise of another million dollars from OpenAI. The promised resources thus reached a total of $2 million.

What the organizers want: to regulate AI and preserve the role of humans

The student organizers claim they want to explore how humans and AI can collaborate in mathematics and say they did not seek to provoke controversy. They have solicited professors from Caltech, who are being considered as judges, as well as other experts, and state that they have developed a framework for the responsible use of AI with them. Their cooperation with Anthropic and OpenAI aimed, according to them, to engage in critical dialogue with these companies. They present the mathathon as an opportunity to build connections between communities and assure that they take the received criticisms seriously.

A structured opposition: open letter, feared marketing, and methodological doubts

An open letter has gathered over 700 signatures calling for the event's cancellation. The signatories believe that the format, as supported by Anthropic and OpenAI, could have destructive effects on the mathematical community. They contest a presentation phrase questioning the role of the mathematician if AI solves conjectures faster, a judgment they describe as motivated and empirically dubious. Other critics fear a marketing takeover by AI companies and worry that such a setup could lead to hasty and misunderstood results. Caiman Moreno-Earle claims that some people consulted privately later signed the public letter.

An acknowledged generational gap and concerns about careers

The organizers are sticking to their plan despite calls for cancellation and defend the idea that young mathematicians cannot wait for formal rules on AI to emerge. Caiman Moreno-Earle, 19, says he has been demoralized by the speed at which AI companies tackle difficult theorems and initiated a multi-day mathathon. About a month after the idea was launched, a letter denouncing "slop" emerged, while professors expressed methodological skepticism and reaffirmed that only the progress of human knowledge matters. The controversy resurfaced the previous week when OpenAI claimed to have solved a problem with a $1 million prize from an institute. The flyer, published on September 4, announced an event that was later confirmed with modifications. Alvan Arulandu states he does not want to cancel and situates the debate in the tension between preserving the human element and the rise of system capabilities. Moreno-Earle presents the initiative as a necessity and asserts that students cannot wait for rules to be set by seniors or companies.