CAISH Social: Maths, Machines, and Misalignment

Last month, Terence Tao and other Fields Medallists published a statement on “a severe misalignment” between the goals of AI companies and those of the mathematical community. It followed OpenAI’s claim that an AI model had resolved a version of the Navier–Stokes Millennium Prize Problem. The claim is still being checked, and only a small number of experts can check it. If AI systems can now make progress on open problems in mathematics, similar progress may follow in other fields of research, including AI research itself. Some of these fields carry serious risks if the capability is misused or poorly overseen. We will discuss what happens when AI produces results faster than people can verify them. Join us for a casual evening of discussion, and food. No maths background needed.