Please note the change in date and format for this research meeting. This meeting will be held on Tues., Nov. 30, 2021; (Previously scheduled for Nov. 23rd.) This meeting will also be held in a fully remote format via Zoom.
Abstract: Language, and more generally the principle of compositionality, provides a window through which humans generalize their knowledge between radically different settings. We will discuss how compositionality can be incorporated into robotic models to enable them to act rationally in new scenarios while following novel commands. We then apply this principle to creating machines that reason about sequences of actions and about social interactions. Finally, we will present a new dataset and analysis that helps shed light on how language is processed by the brain.
MIT Building 46 | Brain and Cognitive Sciences Complex, 43 Vassar Street, Cambridge MA 02139