Washington- Twenty-five winners of the Fields Medal, the highest international recognition in mathematics, warned this Friday of a clash between the objectives of mathematics companies. artificial intelligence (AI) and those of the mathematical community, given the advance of models capable of solving research problems at an unprecedented speed.
In a statement posted on the website Math and AI under the title “A serious misalignment of AI in mathematics”, the signatories recognize that language models have improved drastically in recent months to the point of solving important problems in different fields of mathematics.
However, they consider it detrimental to the discipline if AI companies turn problem solving into a competition to measure their models.
Among the signatories are some of the most recognized mathematicians of recent decades, such as the American Terence Tao, winner of the Fields Medal in 2006; the Frenchman Cédric Villani, awarded in 2010; the German Peter Scholze, who received the award in 2018; and the Ukrainian Maryna Viazovska, distinguished in 2022.
The document is also signed by the Chinese Yu Deng, one of the winners of the 2026 Fields Medal.
1/17 | Impressive: China exhibits its advances in artificial intelligence and robotics in Shanghai. A realistic-looking robot head watches a visitor during the World AI Conference. – Ng Han Guan
The signatories warn that the increasingly rapid production of true or false mathematical statements can make it difficult to verify them, identify new methods, connect with previous work and, finally, incorporate them into the knowledge of the discipline.
They also raise issues of attribution and possible plagiarism when solutions are announced before they can be adequately studied.
The signatories do not reject the use of AI and recognize that it can improve and accelerate mathematical research, but they consider that the impact of this technology will depend on the decisions of those who develop and control it.