UC Berkeley Professor Stops Grading Due to AI Use
A professor at UC Berkeley has reportedly decided against grading assignments after noticing a significant increase in students using AI to boost their grades. Richard Borcherds, who is a well-known mathematician and received the Fields Medal in 1998, commented, “Student use of AI for homework is now widespread.” He believes that grading based on correctness prompts students to rely on AI instead of doing the work themselves, which ultimately punishes honest students.
Borcherds expressed that evaluating assignments for AI-generated work seems pointless and a waste of both time and money. Instead, he has shifted to giving students credit simply for completing their assignments, regardless of whether their answers are accurate. He admits he doesn’t have a better solution at this moment.
He remarked, “There is no easy way to tell the difference between human and AI generated work.” Sometimes it’s possible to spot AI due to certain typical errors, but the technology has advanced so much that it rarely makes blunders anymore.
In a discussion with UC Berkeley’s student newspaper, Borcherds voiced his worries about the future of mathematics. He warned that if AI keeps advancing at this pace, it might reach a point where it outperforms human mathematicians in every aspect.
“I have no idea what to advise,” he said, even mentioning that he asked ChatGPT for insight on this issue, only to find that it was equally puzzled about the future.
He reflected on his uncertainty about advising students to pursue mathematics, highlighting that just five years ago, it seemed like a stable career choice amidst the AI revolution.
Borcherds likened the current scenario to the ongoing existence of physical journals, which he believes have been irrelevant for quite some time. “There have been plenty of technical breakthroughs in the past,” he explained, “but this is the first time we’re seeing something like this in mathematics.”
The university has also experimented with AI detection tools like Turnitin and Rumi, but found them insufficient for a standardized solution, leaving professors to decide how to address AI in their classes.
“If the worst effect of AI is that it complicates homework assessments, then I’ll be quite relieved,” Borcherds noted, hinting at his concerns for potentially bigger challenges ahead.
Additionally, there have been discussions about the “Cobra Effect” created by AI detection tools, which may not only be identifying AI-generated text but also changing how students write. Some are noticing that sophisticated writing is being penalized while generic styles pass through unnoticed.
For instance, a student who wrote an essay on Kurt Vonnegut’s “Harrison Bergeron” saw a high AI-generated score due to the word “devoid.” Once they replaced it with “without,” the score dropped dramatically with no real change in the essence of the essay. This exemplifies how current detection systems focus largely on superficial statistical patterns rather than meaningful authorship evaluation.
The foundational technology for these detection tools could be part of the issue. They analyze various textual features to determine the likelihood of machine-generation, leading students to either simplify their writing or utilize AI to produce content that fits within acceptable parameters.
Wynton Hall from Breitbart News has expressed concern about the unintended consequences of evolving technology. He describes the current moment as standing on the edge of great potential while also navigating the challenges it brings.
