ChatGPT and the End of Learning (www.theargumentmag.com)

A UC Berkeley instructor reports that widespread use of ChatGPT in his Introduction to Database Systems course let students produce flawless homework but left them ill-prepared for exams: assignment scores stayed high while handwritten exam averages fell about 15% below historical norms and dropped further on a second exam. The professor argues AI short-circuits the iterative debugging and first-principles thinking central to learning programming—students outsource problem solving to chatbots, miss building mental models, and then can’t perform when tools aren’t available. He also sees the downstream hiring consequence: candidates who leaned on AI during practice struggle to explain or defend solutions in interviews. The episode underscores a generational and pedagogical shift. Survey data cited shows millennial and Gen Z users adopt chatbots far more often (nearly a third use daily; 37% of under‑30 AI users rely on them to “get work done”), and younger cohorts are both more optimistic about AI and more worried it could amplify productivity to the point of replacement. For AI/ML practitioners and educators this raises technical and policy implications: curricula and assessments must be redesigned to evaluate conceptual understanding (not just output), onboarding should anticipate AI-assisted workflows, and regulators and institutions need new guardrails that distinguish novice learning from professional augmentation.