Date:
Wednesday, November 29, 2017, 4:00pm to 5:15pm
Location:
WJH Basement Auditorium
Jean-Remi King (NYU - Poeppel/Melloni Lab)
Title: “Identifying the Neural Architecture of Human Cognition with Machine Learning and Neuroimaging”
Abstract: What computations are specific to human cognition? Are they supported by a particular neural architecture? Here, I will argue that the recent developments in 1) artificial intelligence and 2) temporally-resolved neuroimaging offer a new opportunity to address these foundational issues. In particular, I will show how we can track the unfolding of complex hierarchical processes, such as those supporting perceptual decision making and speech comprehension, and reveal the representations that the human brain encodes, selects, and maintains at each instant in time. The present studies show that brain activity can be automatically decomposed into a multitude of parallel, serial and recursive processing stages that can be directly compared and contrasted to Bayesian and deep neural networks. Finally, I will outline a few promising research directions to parse the recursive computations responsible for syntactic processing and conceptual composition. I argue that the present research program will pave the way to integrate the disciplines of biological and artificial cognition.