Will Artificial Intelligence Put Mathematicians Out of Work?

Mathematics as a rigorous scientific discipline has existed for more than two millennia, always sustained by a thriving community of people --- the mathematicians --- who are constantly engaged in problem formulation, exposition, teaching, refereeing, professional meetings, informal give-and-take, and a lot more. By contrast, Artificial Intelligence (AI) emerged as a distinct and separate scientific discipline about seventy years ago, which moreover cannot exist and develop further without its mathematical foundation. While AI owes its very existence to mathematics, and mathematics has been and will continue to be the foundational backbone of many other disciplines --- the mathematical sciences --- there are yet influential commentators and decision-makers of the hi-tech industry that predict that mathematics will become obsolete and lose its privileged position because of AI within a few years.

This lecture examines this prediction critically by analyzing three AI tools ---Large Language Models (LLMs), Automated Theorem Provers (ATPs), and Interactive Proof Assistants (IPAs) ---and asking what mathematics each requires, and what problems each has actually solved. The evidence reveals a different picture: LLMs can be good at solving mathematical problems (without absolute guarantee that they will succeed) but cannot discover new mathematics without guidance and even prodding by human mathematicians; by contrast, ATPs and IPAs excel at verification and formalization, and yet not at discovery. Rather than displacing mathematicians, these tools reveal mathematics to be more essential than ever --- as the foundation on which AI itself is built, and as the discipline most capable of ensuring rigorous correctness in an AI-driven world. The lecture concludes that mathematicians are not at risk; instead, their roles are evolving and expanding.