Talk Title: Establishing the neural foundations of physical intelligence

Talk Abstract: My research aims to understand the neural representations and computations that subserve physical intelligence—creatively and efficiently assembling dexterous interactions with the physical world to solve complex challenges. I study these neural computations with high-throughput electrophysiology in rich and flexible tasks performed via robotic manipulanda in virtual environments. Although we can associate behaviors with specific neural activity patterns through passive observation, identifying the precise mechanisms that shape these patterns necessitates causal intervention. By formalizing hypotheses about neural computations as dynamical systems, I directly test them using targeted optogenetic and electrical perturbations. Through this approach, I resolved fundamental questions about the motor cortex, identifying a low-dimensional dynamical system that governs neural activity during reaching movements and is contained within a neural subspace exhibiting a privileged causal relationship with behavior. Additionally, I have assembled a brain-wide spiking activity map in a dynamic foraging task, wherein future decisions rely on computations integrating past interactions with the environment, including the history of choices and rewards that shape the subjective value of presently available actions. I have devised a data-driven framework to reconstruct the dynamical systems that shape brain-wide neural activity and reveal the key dynamical structures that underlie specific cognitive computations. Looking forward, I will integrate these methods to study a novel, cognitively demanding physical construction task, establishing the precise neural computations that enable physical intelligence, multi-step planning, and compositional reasoning.

This talk will not be live streamed