A UC Berkeley professor has ceased grading students' homework for accuracy, citing the increasing ability of artificial intelligence tools to produce correct answers. Richard Borcherds, a mathematician and Fields Medal recipient, now awards credit based on assignment completion rather than the correctness of the work.
Borcherds told The Daily Californian, Berkeley's student newspaper, that if AI continues to improve at its current rate, human mathematicians may soon be surpassed in their capabilities by machines. He has begun grading assignments based on whether students complete them, acknowledging that this is a deviation from a traditional grading system but stating he has few alternatives.
This adjustment reflects the mounting challenges universities face as students gain access to sophisticated AI systems capable of completing academic tasks. Borcherds also expressed uncertainty about the future of mathematics as a field of study for aspiring professionals, noting that just five years ago, it seemed relatively insulated from the widespread disruption now affecting other professions.
He reported even asking ChatGPT for advice on the matter, but the AI itself could not provide a clear answer regarding future outcomes.
The professor suggested that academic institutions might be slow to adapt to rapid advancements in artificial intelligence, pointing to a general resistance to change within academia. He anticipates that mathematics departments may continue operating similarly to how they have for years, even as AI capabilities expand. Borcherds drew a parallel to academic journals that continue publishing print editions despite, in his view, having lost much of their relevance decades prior.
He stated that while technological advancements have historically displaced human labor, such as hand weaving by looms or horses by automobiles, this is the first time a similar impact has been observed in mathematics.
UC Berkeley has implemented guidelines requiring faculty to address the use of AI in their courses. Anne Marie Richard, the university's assistant vice provost for undergraduate education and chief academic technology officer, indicated that instructors are asked to include statements in their syllabi detailing course expectations, including policies on AI tools. The specific restrictions can vary significantly based on the course content and subject matter.
The university has investigated AI detection software, such as Turnitin and Rumi, but concluded that these tools were not sufficiently reliable for campus-wide implementation. Consequently, individual instructors retain autonomy in setting AI policies for their respective classrooms.