- When
- Wednesday, Jan 7, 2026, 3:30 – 4:30 pm
- Event interval
- Single day event
- Campus location
- Kincaid Hall (KIN)
- Campus room
- 102/108
- Accessibility contact
- chairpsy@uw.edu
- Event types
- Lectures/Seminars
- Event sponsors
- Department of Psychology
- Description
- Making image with minds and machines
Thomas Naselaris, Associate Professor of Neuroscience, University of Minnesota
I will discuss recent advances in using human brain activity to externalize the visual content of mental images. I will first consider the more straightforward challenge of decoding brain activity into reconstructions of seen images, and show how large neuroimaging datasets and tools from AI are accelerating progress. I will then show that models optimized to reconstruct seen images can generalize reasonably well to mental images—well enough that naive observers can reliably identify imagined pictures from their reconstructions. We will also see that there is much room for improvement. I will introduce a computational theory of mental imagery that explains why directly extending vision-based models to mental imagery may not be the most effective approach, and share several recent results in support of this theory.
Faculty host: Ariel Rokem
These lectures are made possible by a generous endowment by Professor Allen L. Edwards - Map
- View Google Map