Thirty of the world's top mathematicians gathered in Berkeley, California, for a secret meeting to challenge OpenAI's latest AI model, o4-mini. The AI, a reasoning large language model, was tasked with solving complex mathematical problems devised by the mathematicians. Despite their best efforts, the mathematicians were stunned by the AI's capabilities, with o4-mini solving problems that would typically stump even the brightest graduate students.
The meeting aimed to test o4-mini's mathematical prowess and push the boundaries of AI's problem-solving abilities. The mathematicians were divided into groups and tasked with creating challenging problems, with a $7,500 reward for each problem the AI couldn't solve. O4-mini's performance was impressive, with the AI solving complex problems in mere minutes, whereas human experts would take weeks or months to accomplish the same tasks.
The AI's ability to reason and provide solutions was likened to working with a "strong collaborator". Ken Ono, a mathematician at the University of Virginia, noted that o4-mini's capabilities are "approaching mathematical genius". The meeting sparked discussions about the future role of mathematicians in an AI-driven world. As AI models like o4-mini continue to advance, mathematicians may shift their focus from solving problems to posing questions and interacting with AI models to discover new mathematical truths.
While o4-mini's performance was impressive, some mathematicians expressed concerns about the potential risks of relying too heavily on AI-generated solutions. Yang Hui He, a mathematician at the London Institute for Mathematical Sciences, noted that o4-mini has mastered "proof by intimidation," presenting solutions with confidence, which can be intimidating. Overall, the meeting highlighted the rapid progress AI has made in mathematical problem-solving and the potential for AI to augment human capabilities in mathematics.
