Research

Developing AI systems that learn and think with people
Building AI systems designed to collaborate with humans as thought partners — learning and thinking with people rather than instead of people.

Probing the limits of learning and teaching with minimal examples
Investigating the fundamental limits of learning from very few examples, including less-than-one-shot learning, dataset distillation, and soft-label prototypes.

Studying parallels between human and artificial intelligence representations
Understanding how representations in artificial neural networks relate to representations in the human mind. We use representational alignment as a tool to study these parallels and improve AI systems.