Rising AI Usage Correlates with Higher Failing Rates and Weakening Math Skills in Berkeley CS

The Daily Californian reports on a concerning trend in UC Berkeley's computer science department: as students increasingly use AI assistants like ChatGPT and GitHub Copilot for assignments, professors observe a rise in failing grades and a deterioration of fundamental math skills. Instructors note that students struggle with basic linear algebra, calculus, and discrete math—prerequisites for advanced CS topics—when they offload problem-solving to AI. Some students become over-reliant on AI-generated code and explanations, failing to develop their own understanding. The department is exploring assessment reforms, such as in-person exams without AI tools, to preserve learning outcomes. The situation underscores the double-edged nature of AI in education.