Overview
- Fields medalist Richard Borcherds now awards homework credit for completion rather than grading answers for correctness after observing many students using AI to produce correct solutions.
- Borcherds told student reporters that pervasive AI use made paying graders to check correctness “no point” and that continued AI progress could leave human mathematicians with little to do.
- UC Berkeley’s Academic Senate has asked instructors to state course expectations about AI in syllabi and built an AI Hub with sample language for faculty to use.
- The university tested detection products such as Turnitin and Rumi and judged them insufficiently reliable to standardize campus-wide, so enforcement and specific course rules remain at the instructor level.
- The change follows OpenAI’s high-profile claim about progress on the Navier–Stokes problem and raises practical and career questions for students, including how departments will assess learning and advise future mathematicians.