Cédric Villani warns of AI's impact on science

Cédric Villani, a former French parliament member and Fields Medal winner, has raised concerns about how artificial intelligence is affecting scientific research, especially in mathematics. He discussed these issues in an interview with Rebecca Fitoussi on the radio program Un monde, un regard.

Villani's intellectual journey

Villani’s career has been shaped by deep focus and a drive to understand complex subjects. He described how his early struggles and later work on mathematical theories like optimal transport and lambda damping were all part of his ongoing pursuit of knowledge.

Transition from politics to academia

After serving in the French parliament, Villani has returned to his academic roots. He still remembers the excitement of political life, describing the parliamentary environment as full of energy and intense debate, even though he now focuses on research.

Bridging science and politics

Villani’s political involvement, first with the movement La République en marche and later with an ecological group, was motivated by a desire to connect science and politics. He believes that in times of environmental and technological challenges, experts must be heard in political decisions.

Challenges of political courage

Villani argued that the real challenge in politics was not just providing scientific information but also having the political courage to act on it. He noted that many problems have clear solutions, but implementing them can be difficult due to political resistance.

Mathematicians caution against AI overreach

In 2025, many top mathematicians saw AI as a useful tool. But by 2026, they realized it could potentially replace them, especially when controlled by large corporations. Villani, along with 24 other Fields Medalists, warned about this in an op-ed in Le Monde.

Setting boundaries for AI use

Villani acknowledged that experts might wrongly believe they are irreplaceable. He urged mathematicians to define the limits of AI use and to decide what is acceptable in their field. He stressed that authorship should not be given to someone who simply let a machine generate a result.