AI use leads UC Berkeley math professor to see 'no point' in grading homework

Borcherds gave the interview on a Tuesday, shortly after OpenAI said its agents had apparently solved the Navier-Stokes existence and smoothness problem. He currently grades on completion if a student hands in their homework. He concedes that the approach is flawed but says he has no workable replacement. Berkeley has no formal campuswide AI policy. The Berkeley Division of the Academic Senate has asked all faculty members to include AI-use expectations in their course outlines. Each class's AI policy is up to faculty discretion. The university has tested detection software, but these tools aren't adequate enough to be standardized on campus. Borcherds, a 1998 Fields Medal winner, also said he asked ChatGPT what students should do about their math careers and got no clear answer. Separately, the Association for Human Mathematicians urged mathematicians to "discontinue their work with OpenAI" over the Navier-Stokes claim. For a founder, the signal is that correctness-based assessment is breaking down in a top math department. That creates demand for grading and verification methods that show how an answer was reached, not just whether it is right. It also shows that institutions are slow to change their policies, so products that help faculty adapt may find buyers before formal rules arrive.