Human cognition demonstrates a remarkable ability to transcend limited experience to form highly general, abstract ideas. Understanding how the mind and brain accomplish this has been a central challenge in multiple scientific fields. The recent success of large language models in AI has led to the development of models capable of performing certain abstract tasks, but their ability to do so can be unreliable and relies on more data than individual humans receive in an entire lifetime. Moreover, the nature of the representations that are learned needs to be better understood. Recent research has begun to synthesize ideas at the interface of neuroscience and AI, leading to new approaches to modeling abstraction in natural and artificial systems.
The Wu Tsai Institute's Center for Neurocomputation and Machine Intelligence invites you to a small, informal two-day workshop that brings together researchers from different disciplines to share ideas and spark further work in this exciting area. Please RSVP for the workshop and view the working agenda below.
Workshop Agenda
12:00 PM | Lunch (11th floor common area)
1:00 - 3:00 PM | Session 1 (Workshop Room 1116)
- 1:00 - 1:40 PM | Taylor Webb (UCLA): The Relational Bottleneck
- 1:40 - 2:20 PM | Awni Altabaa (Yale): Relational Convolution Networks
- 2:20 - 3:00 PM | Declan Campbell / Tyler Giallanza (Princeton): Modeling of Context and Abstractions for Control
3:00 PM | Break (11th floor common area)
3:30 - 5:30 PM | Session 2 (Workshop Room 1116)
- 3:30 - 4:10 PM | Stefano Fusi (Columbia): Geometry of Abstraction
- 4:10 - 4:50 PM | Kamesh Krishnamurthy / Simon Segert (Princeton): Learning Symmetries and Counting
- 4:50 - 5:30 PM | Michael Wu / Abhishek Bhattacharjee (Yale) : Relational Learning in Computer Systems
6:00 - 8:00 PM | Dinner at Silliman Hall (505 College Street)
8:00 - 9:00 AM | Breakfast (11th floor common area)
9:00 - 10:20 AM | Session 3 (Classroom 1167)
- 9:00 - 9:40 AM | Jeff Johnston (Columbia): Abstractions and Multi-Task Learning
- 9:40 - 10:20 AM | Shanka Mondal / Taylor Webb (Princeton): Applying the Relational Bottleneck to Complex Visual Inputs
10:20 - 10:30 AM | Coffee (11th floor common area)
10:30 AM - 12:00 PM | Session 4 (Classroom 1167)
- 10:30 - 11:00 AM | Matteo Alman (Columbia): TBA
- 11:00 AM - 12:00 PM | Brainstorm Session