UC Berkeley Math Professor Halts Homework Grading as AI‑Generated Assignments Surge

Richard Borcherds, a leading mathematician at the University of California, Berkeley, announced that he will no longer assign or grade traditional homework in his courses, citing the growing prevalence of artificial‑intelligence tools that can produce solutions indistinguishable from those written by students.

Borcherds, who has earned a reputation for rigorous coursework and meticulous evaluation, explained that the rapid improvement of large‑language models makes it virtually impossible to verify the authenticity of written submissions. "When a student hands in a solution that could have been generated by an AI, there is no reliable way for me to tell the difference," he said, emphasizing that the integrity of the assessment process has been compromised.

The professor’s decision reflects a broader concern among educators about the ease with which generative AI can complete problem sets, essays, and code assignments. While many institutions have experimented with plagiarism‑detection software, those tools are less effective against content produced by models that can paraphrase or create original mathematical reasoning. As a result, Borcherds has shifted his assessment strategy toward in‑person examinations, where students must demonstrate their understanding under supervised conditions.

In moving to on‑campus exams, the department hopes to preserve the core learning objectives of its mathematics programs while discouraging reliance on unauthorized assistance. The change also aligns with a growing trend at universities nationwide to redesign curricula, incorporate oral defenses, and emphasize project‑based work that is harder to outsource to an algorithm.

The transition raises questions about the future of homework as a pedagogical tool. If AI can reliably solve standard problems, educators may need to rethink assignments to focus on higher‑order thinking, collaborative inquiry, or real‑world applications that require contextual judgment beyond formulaic answers. For now, Borcherds’ classes serve as a test case, and the university is monitoring outcomes to determine whether similar policies might be adopted across other departments.