A Berkeley math professor stops the grade of homework because of AI

Richard Borcherds, a professor at the University of California Berkeley, has stopped the grade of homework for correct answers. Borcherds tells that many students used AI to get top scores. Borcherds sees no point in payment for readers to grade the work. The interview was on Tuesday. Before that, OpenAI told that its AI agents possibly did the proof of the Navier-Stokes problem.

What changed

Borcherds gives credit to each student who gives the work. Borcherds names this “not a very good system”, but has no alternative at this time. Borcherds does not know if students must study math.

The proof

The source gives no data from the class. It gives the opinion of Borcherds on the scores. Borcherds thinks that in some years, AI will do all work better than the mathematicians. This is an opinion, not a test result.

What is not known

The campus tried the detection tools Turnitin and Rumi. It found that they are not sufficient for all classes. Each teacher sets the rules for a class. Borcherds thinks the department can stay the same for some years.

Insight: Homework scores measure the tool, not the student

A study by UC Berkeley and 4 other universities found that after 10 minutes of ChatGPT, 1,222 test persons made more errors, or stopped. This was when the team removed the tool. The campus has no good detection tool. Homework done with AI shows the work that the tool can do, and the correct check must be work done without it.