Theories of human cognition have, at each point in history, been colored by the dominant technology of the times. In the 17th century, without any technology reminiscent of human cognitive capabilities, Cartesian dualism, in which the human mind stood outside the realm of physical laws governing the human body, ruled the day. In the 20th century, with the advent of Turing computation, the analogy of the mind as a computer dominated. More recent advances in statistical machine learning have yielded exceedingly interesting Bayesian formalisms for quantifying sensorimotor and cognitive behavior.
In the 21st century, what technologies will further drive theories of human cognition? I believe the transformative technologies will not be artificial computing algorithms that provide metaphors for the mind, but physical technologies that can probe and perturb the biophysics of the brain. Indeed recent experimental revolutions in neuroscience have opened new vistas into immensely complex dynamical processes spanning many spatiotemporal scales, ranging from molecular dynamics to global neural activity patterns during behavior. These new experimental windows, amplified and empowered by national brain initiatives, now allow us to address one of the most striking conundrums in modern science: how do the cognitive capacities of the mind emerge from the biological wet-ware of the brain? However, experiment alone cannot answer this question; to obtain a meaningful understanding of such complex data, we desperately need integrative theories and new mathematical frameworks to concisely describe how important behaviors like perception, action, learning and memory emerge from the cooperative activity of high-dimensional, multi-scale neural processes.
My laboratory is dedicated to generating and testing such theories in close collaboration with experimentalists. In particular, two interrelated goals of my laboratory are (1) Elucidate how cognitive behaviors emerge from the microscopic dynamics of neurons and synapses embedded in networks; (2) Develop methods in high dimensional statistics and machine learning to extract meaningful predictive models of cognitive and sensorimotor phenomena from large neuroscientific datasets. Indeed my career has been defined by an intense desire to understand how non-intuitive macroscopic phenomena emerge from microscopic dynamics. For example, during my PhD in string theory, I explored how the geometry of space-time itself could emerge from statistical interactions in large non-geometric objects. Building on this rigorous training, my lab and I employ and invent a broad arsenal of theoretical tools from physics, mathematics, engineering, statistics and computer science to develop new conceptual frameworks for describing the relation between biophysics and behavior across many scales of biological organization.
We have achieved several successes along these lines, including understanding how attention, decision making, sequence memory, sensorimotor learning, and semantic learning can all emerge naturally from distributed, plastic neuronal circuits. In the future, we will build on these successes to explore several research directions, including how synaptic complexity contributes to learning and memory, how distributed plasticity in sensorimotor loops gives rise to imitation learning, how collective motor cortical dynamics generates complex movements and plans, and how the nonlinear dynamics of learning in deep neuronal networks can yield mathematical laws of human semantic cognition.