Mathematicians Address AI's Impact on the Field

<p>In late July, mathematician Terence Tao from UCLA warned that the field of mathematics is entering a "turbulent period" due to advancements in artificial intelligence (AI). He made this statement during a lecture at the International Congress of Mathematicians, highlighting concerns that AI's capabilities are causing a "crisis in our mathematical values and practices."</p><p>Earlier that week, a researcher at Anthropic used the AI model Claude to disprove an 87-year-old mathematical conjecture. Many mathematicians at the conference expressed surprise at the focus on AI, indicating they had not considered its implications deeply.</p><p>Attention to AI's role in mathematics has increased following OpenAI's announcement that its AI systems solved the Navier-Stokes problem, one of the Millennium Prize problems that has challenged mathematicians since the 1930s. David Dritschel, a mathematician at the University of St. Andrews, described the Navier-Stokes problem as a "career-killing problem" due to its complexity. OpenAI reported that its AI agents solved the problem in less than four days.</p><p>This development has prompted an existential reflection within the mathematics community. Although OpenAI's solution relied on nearly a century of human research, mathematicians are recognizing the significant influence AI now holds over the discipline. Max Weinreich, a math professor at CUNY Baruch, remarked on the rapid changes AI is bringing to the field.</p><p>Tristan Buckmaster, a mathematician at NYU, claimed that he and his collaborator, Levent Alpöge from Anthropic, had made significant progress on the Navier-Stokes problem before OpenAI's announcement. Buckmaster suggested that OpenAI may have accessed their work through interactions with its AI agents. In response, OpenAI stated that it began working on the Millennium Problems after hearing rumors of solutions and that it did not reference Buckmaster's inputs.</p><p>The situation has raised concerns among mathematicians regarding the direction of AI in mathematics. A statement from 25 Fields Medal winners criticized the trend of AI companies solving mathematical problems primarily for publicity, arguing that this approach is harmful to the scientific integrity of mathematics.</p><p>While solving mathematical problems is a key aspect of the discipline, the process often leads to the development of tools that benefit other areas. For example, the Apollo missions not only achieved the moon landing but also advanced technology that has had lasting impacts, such as in the computer-chip industry.</p><p>Mathematicians hope that insights from OpenAI's proof of the Navier-Stokes problem will facilitate the exploration of other mathematical challenges. Historical precedents, such as the work on Fermat's Last Theorem, have shown that significant breakthroughs can yield valuable tools for various applications.</p><p>However, the complexity of AI-generated proofs poses challenges for human understanding. Many AI-generated solutions are difficult for mathematicians to comprehend, not due to their complexity, but because of the time required to analyze them. There is concern that while AI can accelerate problem-solving, it may not enhance human understanding of the proofs.</p><p>To adapt to these changes, mathematicians may need to shift their focus from merely solving problems to effectively communicating the underlying ideas of their solutions to benefit the broader community. However, establishing metrics for evaluating communication effectiveness remains challenging.</p><p>As AI continues to evolve rapidly, mathematicians face the task of determining how best to respond to these developments. The "turbulent period" Tao described is likely to persist as AI capabilities advance.</p>