Care for a little more AI?

Duminil-Copin (Université de Genève and IHES) writes that frontier models already outperform humans at many mathematical tasks. He expects AI to resolve famous conjectures, using the dying-percolation conjecture θ(p_c)=0 as his example, before humans would. In his view this loses the real value of open problems: the failed attempts, collaborations, auxiliary methods and intuition built across generations. Open problems act as "lighthouses" for communities, and quick AI solutions risk "decapitating entire fields before they have time to develop", which leaves young mathematicians without problems to grow on. He offers no simple remedy. He says he personally has chosen not to let AI replace his creative process and argues for keeping a culture of search, doubt and failure.

(summary plus short quotes only)

• "only a matter of time before the most famous conjecture in our field ... also falls to the bulldozers" (as quoted by Scientific American)
• the current use of AI "does not empower us, it petrifies us"
• "Let the lighthouses built for us shine a little longer. We owe it to our predecessors"
• "I have chosen not to rely on artificial intelligence to replace my creative process"