Mathematicians warn AI threatens the future of their field

An international group of mathematicians has sounded the alarm over the accelerating progress of artificial intelligence, warning that its rapid development poses a significant threat to the future of mathematical research. The concern is that AI is not just a tool but is beginning to encroach on the fundamental processes of mathematical discovery and validation.

This warning comes in the wake of AI models demonstrating advanced logical reasoning capabilities, including solving long-standing mathematical problems. The "Leiden Declaration on Artificial Intelligence and Mathematics," published by 16 researchers and supported by numerous colleagues and organizations, calls for immediate action to address the challenges presented by AI's integration into the field. The declaration has garnered over 1,200 signatures from researchers worldwide within its first few days.

“

As I myself conduct basic research in AI, I am both enthusiastic about the possibilities AI can offer and worried about how misuse of AI affects society and the research community.

Martin Trapp, an associate professor in machine learning at the KTH Royal Institute of Technology in Stockholm, expressed a dual perspective: "As I myself conduct basic research in AI, I am both enthusiastic about the possibilities AI can offer and worried about how misuse of AI affects society and the research community." He views the declaration as a crucial first step toward ensuring AI tools are used thoughtfully.

The declaration outlines potential threats, such as AI models producing seemingly credible but ultimately flawed arguments, making it difficult to discern true mathematical proofs from fabricated ones. While acknowledging AI's potential to accelerate research processes, Trapp cautioned, "Language models and AI agents can speed up parts of the research process. But they also easily lead to losing control, and to producing trivial, uninteresting, or incorrect results." Recommendations within the declaration target individual mathematicians, research institutions, funders, policymakers, and the AI industry, urging transparency, rigorous verification, and thoughtful regulation.

“

Language models and AI agents can speed up parts of the research process. But they also easily lead to losing control, and to producing trivial, uninteresting, or incorrect results.