2026-10-27 · Source：Tom's Hardware
UC Berkeley's Richard Brocherds says homework now grades the chatbot, not the student — leaving proctored, in-person exams as the only way to measure what students actually know.
Richard Brocherds has stopped grading homework. The Fields Medal-winning mathematician's verdict on graded problem sets in the age of chatbots: evaluating answers written by AI is a waste of time and money.
The UC Berkeley professor's conclusion goes to the core of the academic integrity problem. When students can hand in coursework generated by a model, what gets graded is no longer the student's understanding but the machine's output — and, in Brocherds' view, paying for that evaluation accomplishes nothing. His response is to consolidate assessment into a single format: the in-person exam, which he frames as the only solution left to widespread AI use.
Who is making the call matters as much as the call itself. Brocherds holds a Fields Medal, the most coveted honor in mathematics, and teaches at one of the world's leading public universities. When a mathematician of that stature concludes that homework — a fixture of mathematical training for generations — can no longer function as a graded exercise, the decision reads as more than one classroom's housekeeping. It amounts to a judgment from inside elite mathematics on what generative AI has done to take-home work.
There is an economics to the argument. Grading consumes time and money — instructor hours, grader budgets — and if the submissions under review are machine-written, that spending measures a chatbot's fluency rather than a student's knowledge. Brocherds' remedy is blunt: stop financing the broken signal and reserve evaluation for the one setting where a student cannot be swapped out for a model.
The open question is how many institutions follow his lead. For now, a Fields Medalist has answered in the starkest form available: whatever happens outside a proctored room can no longer be trusted as a measure of learning, and only the supervised exam survives.
Tags: UC Berkeley · Richard Brocherds · AI Cheating · Academic Integrity · Generative AI · Higher Education
萬安算交所 AIXX · AIXX · Verified compute exchange · Enter spot hall