Artificial Intelligence1 min read

Caltech professor no longer grades homework's because everyone uses AI

By · Published by Everything Blog

In short

Caltech currently lacks a standardized institutional policy guiding how mathematics faculty should manage generative artificial intelligence, prompting individual instructors to adapt their course syllabuses independently. Professor Yi Ni, who has spent 17 years at the university, now scores problem sets purely on completion rather than correctness. Advanced chatbots currently operate at the level of graduate students, rendering traditional take-home accuracy grading obsolete. Students caution that leaning on these digital shortcuts prevents authentic comprehension ahead of high-stakes, in-person examinations.

Key points

  • Detecting artificial intelligence in math problem sets is nearly impossible due to the la…: Detecting artificial intelligence in math problem sets is nearly impossible due to the lack of identifiable stylistic traits.
  • Caltech currently lacks a unified institutional policy governing how mathematics instruct…: Caltech currently lacks a unified institutional policy governing how mathematics instructors should address artificial intelligence tools.
  • Overreliance on artificial intelligence tools risks leaving students unprepared for their…: Overreliance on artificial intelligence tools risks leaving students unprepared for their in-person university examinations.

Sources checked by our analysis pipeline: 1

Professor Yi Ni had no laptop or projector. He instead rapidly filled eight blackboards with chalk scrawl during an introductory math class at Caltech on Friday.

Though the instruction was entirely analog, the specter of AI loomed large.

Ni, a professor at Caltech for 17 years, told me that he no longer grades homework on accuracy because students are using AI to complete it. Now, he grades it only on completion.

"There is a problem in the education process," Ni said. "Virtually all the homework in college math, you can use AI."

Chatbots, once derided for their inability to perform such basic tasks as counting the r's in "strawberry," can now do math as well as a graduate student.

And if you're an undergraduate student, that's kind of a big deal. Some, Business Insider has reported, are reconsidering whether to study math at all.

From my perch near the back of the lecture hall on Friday, however, I didn't see any students prompting ChatGPT.

In many ways, the class resembled a typical early morning university lecture (for undergrads, 10 a.m.…

Artificial Intelligence