How I changed teaching after AI managed to do all my homework assignments (thelastsoftwareengineer.substack.com)

In a recent reflection on AI's impact on education, Vincent Hellendoorn shared how advancements in generative AI, predating the launch of ChatGPT, led him to reevaluate his teaching methods and design of assessments in his Machine Learning in Production course. Previously reliant on traditional assessment methods, Hellendoorn has shifted towards more interactive, in-person evaluations with teaching assistants (TAs) and reduced written assignments in response to students using AI to generate homework. He emphasized that while students are encouraged to use AI responsibly across many settings, this technology is also challenging evidence-based pedagogical practices, such as the emphasis on low-stakes assessments and feedback. Significantly, Hellendoorn adopted a strategy where students must now engage in oral check-ins with TAs to discuss their work, replacing written reflections which could be easily offloaded to AI. This approach fosters direct interaction and deepens understanding, even as it recognizes that some students may exploit AI to circumvent traditional learning. Additionally, the integration of autograding via LLMs for coding assignments has streamlined grading processes, allowing TAs to focus more on valuable interactions with students. This evolution in teaching emphasizes adapting to technological advancements while still aiming to equip students with essential engineering skills.