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

Commenters debated whether AI makes graded homework ineffective, and what homework is meant to accomplish. Some argued that working problems is essential practice: even collaboration or copying once pushed students to engage with the material, while AI can eliminate that effort and leave less self-motivated students unprepared for exams. Others said homework is primarily for learning, not grading, and that points turn education into a transaction while encouraging cheating. Several described courses where homework carried no grade, or counted only as completion, with mastery judged through exams.

Many favored shifting grades to in-person, proctored exams, oral questioning, interviews, or student presentations, sometimes retaining homework as optional practice or a prerequisite. Supporters said this would make grades better reflect what students know and remove the incentive to submit AI-generated work. Others emphasized that assessment should still encourage curiosity and learning for its own sake, and that AI can help independent learners by offering accessible feedback on proofs. Commenters also disagreed about the wider meaning: some saw a need to redesign university education, while others argued that AI changes assessment rather than mathematics or the value of teaching. One commenter questioned the professor’s framing, citing his AI-industry work; another suggested universities should focus less on industry training.