Greetings from Berkeley~

The Berkeley Artificial Intelligence Research (BAIR) seminar series is hosted by the BAIR Lab and showcases dissertations and job talks of BAIR Ph.D. students and postdoctoral researchers.
 
These talks are held in person on the 8th floor of Berkeley Way West (BWW) at 11:10 am on Wednesdays. For those who join in person, the seminar will be followed by pizza in the kitchen of BWW8. If you are unable to attend in person, please use this link to join seminar over Zoom. (Password: 8006)

Wednesday, 5/14/25's speaker is Kaylene Stocking, a final-year Ph.D. candidate advised by Prof. Claire Tomlin. Kaylene will present her dissertation.

Title: Between pixels and policies: Toward interpretable representations for (inter)action

Abstract: Many robots pick actions based on their internal representation of the surrounding environment. Today, this representation is often computed directly from camera images by using deep learning algorithms to extract relevant high-level features. However, unlike in other computer vision applications, robot representations must support closed-loop interaction, where current actions affect future observations. Furthermore, the need for safety and transparency in robotics motivates closer scrutiny of the contents and limitations of learned representations.
This talk will discuss research that offers two complementary perspectives on interpretable environment representations in robotics. First, we propose new representation learning algorithms that increase interpretability and improve the robot's ability to reason about other agents. Second, we examine the representations learned by existing deep learning models in robotics, including end-to-end autonomous driving and vision-language-action models. A common thread throughout both perspectives is making connections with findings about the representations that support human cognition and action. Together, these projects work toward a holistic understanding of how embodied agents should represent their environments.

Bio: Kaylene is a final-year PhD candidate in electrical engineering and computer sciences, advised by Professor Claire Tomlin, and a Graduate Fellow with the Kavli Center for Ethics, Science, and the Public. Her research lies at the intersection of robotics and machine learning. More broadly, she is fascinated by what goes into 'intelligent behavior' across many different realizations, from octopuses to humans to mobile robots. She is also passionate about science communication and broader impacts of research.

Wednesday seminar links: (Please do not distribute the below links beyond your organization.)
For those of you who are unable to join in person, please join via Zoom here (Password: 8006)
Unpublished Playlist of Talk Recordings 
Unpublished Playlist of 2024 Recordings
Speaker Slides
Click here to add the BAIR Seminar to your calendar
  
Hope to see you at a seminar~

The BAIR Admin Team
Berkeley Artificial Intelligence Research (BAIR) Lab
https://bair.berkeley.edu