Advertisement
Welcome to Physics at the library with the talk "A Physicist’s Toy Model for Learning with AI" by Tor Ole Bigton Odden, Associate Professor from Center for Computing in Science Education, Department of Physics, University of Oslo. How can a large language model write poetry, explain quantum physics, invent a plausible historical anecdote, and then confidently get a simple fact wrong? Chat-based large language models (LLMs) have become remarkably capable at so many things, yet it remains surprisingly difficult to build an intuition for what they are actually doing. In this talk, I will take a physicist’s approach to this problem and propose a conceptual toy model for understanding how chat-based LLMs behave. The model has three components which, together, help make LLMs' strange mixture of capabilities and limitations more understandable: the LLM as a role-playing machine, as a compression algorithm, and as a conceptual blender. These perspectives offer complementary ways of understanding how LLMs generate plausible responses, encode patterns from enormous bodies of human-produced text, and recombine ideas across contexts. Using these frameworks, I will explore where LLMs fit best in our classrooms, where they do not, and how we can use them in ways that genuinely support learning.
Tor Ole Bigton Odden is an associate professor of physics at the Center for Computing in Science Education, UiO. He uses ideas from physics to study how people learn, how machines learn, and how machines can help people learn. His research focuses on how computational tools can support student creativity and agency in science learning, how students build computational literacy, and applications of generative AI to do qualitative research at scale. Tor was born and raised in the USA, and holds a PhD in science education from the University of Wisconsin-Madison, a masters in physics from UW-Madison, and a bachelors in physics from St. Olaf College in Northfield, Minnesota.
___________________________
Coffee/tea are served. Everyone is welcome!
The talk will be streamed. You can watch the stream at Realfagsbiblioteket YouTube-channel.
Advertisement
Event venue
UiO : Realfagsbiblioteket, Moltke Moes vei 35,Oslo, Norway
Tickets
Concerts, fests, parties, meetups — all the happenings, one place.
Discover more events by tag
Technology in OsloArtificial-intelligence in OsloData-science in OsloMachine-learning in OsloEducation in Oslo