Fifteen mathematicians who have won the Fields Medal, widely regarded as the Nobel Prize of mathematics, have signed a manifesto accusing artificial intelligence companies such as OpenAI of pursuing goals that are «profoundly misaligned» with those of the mathematical community and of posing a «generalized threat to intellectual work.» The statement was released on Friday and invites other researchers to add their names.
The signatories argue that the race to solve famous problems — including one of the seven Millennium Prize Problems, whose resolution OpenAI recently announced — is distorting the purpose of mathematical research. In their view, solving a problem is only a tool and a means toward the main objective: conceptual understanding and insight. Forgetting this, they warn, could turn the tool against the very goal it is meant to serve.
According to the manifesto, AI-generated solutions are frequently announced hastily, without allowing time for proper writing, for the isolation of new methods and ideas, or for citing relevant prior work by other authors. «As in all creative professions, this raises serious problems of attribution and plagiarism,» the mathematicians write. They also caution that without mathematicians willing to take charge of developing and integrating AI-conceived ideas into the mathematical canon, those ideas would never be fully realized and the crucial human chain of transmission between mathematicians would be lost.
The signatories describe mathematics as a community that functions in many ways as a miniature version of humanity, composed of individuals who use a wide variety of approaches but are united by fundamental values. Its most valuable resources, they say, are students and ideas, which are cultivated with great care. Students are often given problems primarily to develop skills that will allow them to advance in research and other fields, and ideas are disseminated through talks, private conversations, and detailed writing that connects them to previous work. These processes invariably require time and are based on human interaction.
The manifesto acknowledges that the mathematical capabilities of AI systems have improved drastically in recent months, to the point of being able to solve important open problems in many fields. But it argues that the eagerness of AI companies to use mathematical problems as benchmarks is harmful to mathematical science and to the mathematical community. The authors place this within broader alignment problems affecting other scientific and creative professions, as well as society as a whole.
They note that the most famous problems have long served as benchmarks and beacons for measuring a better understanding of the field. Solving one of them has been a clear sign of new perspectives and interesting methods, which are then studied by a community of mathematicians through a long and arduous process of conversations, debates, and simplifications. Ideally, that process ends with results presented in a textbook accessible to any undergraduate or graduate student. Some mathematical ideas endure even longer, becoming tools understood and used by the general population decades or centuries later.
The signatories warn that the increasingly accelerated mass production of «true/false» claims could destroy fertile ground rather than give life to new ideas. They call for measures to ensure that, as AI transforms the way mathematicians work, the original purpose of mathematical work is not lost from sight.