A group of 25 Fields Medal winners, including Terence Tao and Pierre Deligne, has published a statement warning that artificial intelligence’s recent success in solving mathematical problems represents a fundamental misalignment between AI companies’ objectives and those of the mathematical community.
According to the statement posted on mathandai.org, while large language models have dramatically improved in mathematical capability and can now solve major outstanding problems, the rush to use such achievements as benchmarks for AI progress threatens mathematics itself. The mathematicians argue that solving problems is merely a tool—the actual goal is conceptual understanding and insight.
The statement explains that in traditional mathematics, solving a famous problem initiates a lengthy process of community discussion, simplification, and integration into mathematical knowledge. Solutions are carefully documented, new methods are isolated, and previous work is properly cited. This human-driven process eventually produces textbook presentations accessible to students and sometimes yields ideas that become foundational tools decades or centuries later.
In contrast, according to the statement, AI solutions are often announced hastily without proper writeups, without isolating new methods, and without adequate citation of prior work. This raises what the mathematicians describe as “severe attribution and plagiarism questions.” More fundamentally, without human mathematicians taking responsibility for developing and integrating AI-generated ideas into the mathematical canon, those ideas cannot become fully realized, and the crucial transmission of knowledge between mathematicians is broken.
The mathematicians frame this as part of a broader alignment problem affecting multiple fields. They note that years of traditional training in any creative profession serve dual purposes: producing results and developing understanding. As AI systems, built on vast bodies of human work, become capable of producing results directly, these dual goals diverge. The statement suggests this represents a threat to intellectual work across society.
The authors acknowledge that AI could enhance and accelerate mathematical study. However, they argue that whether these changes ultimately benefit or harm mathematics depends on decisions made by humans controlling the technology. They call for urgent action by the mathematical community, by AI companies, and by society broadly to address these alignment issues before similar problems manifest across other forms of intellectual work.
Key facts
- 25 Fields Medal winners, including Terence Tao and Pierre Deligne, issued the warning
- The statement emphasizes that problem-solving is a tool, not mathematics’ primary goal of conceptual understanding
- AI solutions are often announced without proper documentation, method isolation, or citation of prior work
- Traditional mathematics integrates solutions through community discussion and careful writeups over time
- The mathematicians warn this misalignment mirrors broader threats to intellectual work across multiple professions