SAMTAL@SU: … and then AI became Good at Physics. What might that mean for physics education?
Seminar
Date:Wednesday 28 May 2025
Time:12.05 – 12.55
Location:AlbaNova: FA 42
AI can now solve physics problems — and not just the easy ones. In this seminar, we’ll explore how chatbots have improved and what that means for university teaching. Join the conversation over lunch.
Welcome!
To the Teaching & Learning event: SAMTAL@SU lunch seminar with the researchers Bor Gregorcic and Giulia Polverini from the physics education at the Department of Physics and Astronomy, Uppsala University.
The event is aimed at researchers who teach university physics or other science subjects at SU.
Abstract
Today, large language model-based chatbots can often be observed demonstrating impressive reasoning and explanatory capabilities, but this has not always been the case. Initially, these models struggled with basic physics tasks, but have substantially improved since. The arrival of vision-capable chatbots marked another important step for their usability in physics and physics education. While their image interpretation abilities have also been far from impressive to begin with, these are advancing quickly as well. In our talk, we will illustrate the progression of LLM-based chatbots' abilities from late 2022 to Spring 2025 and unpack some of the steps undertaken in their development that made them better at physics. Most importantly, we want to stimulate a discussion about how we as physics educators should adapt our practices now that AI can increasingly perform the physics tasks that we want our students to learn.
Bor Gregorcic
Bor Gregorcic is an associate professor in physics education at the Department of Physics and Astronomy, Uppsala University. His research interests revolve around the use of modern technology in physics teaching and learning. Teaching-wise, he is most active in the preparation of pre-service physics teachers.
Giulia Polverini is a PhD student at the Department of Physics and Astronomy, Uppsala University. Her research is focused on the performance of AI tools on physics tasks and the application of AI in physics education.
