Speakers
- Dr. Ipek Ozkaya
- Dr. Doug Schmidt: Director, Operational Test and Evaluation
- Dr. Michael Hilton: Professor, Carnegie Mellon University
”How can SWE education take advantage of generative AI tools?”
Dr. Schmidt I try to find tasks that are boring or mundane that can be turned over to LLMs. For example, documentation, figuring out dependencies, refactoring code, splitting code into platform independent & dependent portions, etc.
Students who are better programmers are usually more adept at using gen AI effectively because they understand what to look for.
Dr. Hilton I think it would be irresponsible to teach software engineering and not touch on generative AI.
Students with lower self-efficacy often struggles with LLMs, especially when it produces wrong code. These students have a harder time making judgments of LLM output.
My concern is that it will grow the gap between students who are ahead and students who are behind.
00:24:00 Dr. Schmidt I encourage my students to use LLMs as low-cost, high fidelity tutors. I encourage my students to use Glasp to ask questions on lecture transcripts. Quiz yourself using the LLM. If you give LLMs content to work from, they’re actually quite accurate. You only run into problems when you ask LLMs open ended questions.
00:26:00 Dr. Schmidt I encourage students to use paid versions of LLMs, like ChatGPT 4.0.