Sleep as a Trojan Horse (to find a unifying computational principle central to biological computation)

The brain’s fundamental challenge is the equivalent of competing in a game show without knowing the topic, the format, the level of complexity, the strength of the competition, or even the language. With your life on the line, how might you best prepare? What organizing principles will maximize your likelihood of success when faced with an unpredictable and arbitrarily difficult task? Almost nothing that we do is hard wired — there hasn’t been time to evolve circuits for scrolling, skateboarding, or even writing. As a result, the brain must adapt to the world in which it finds itself.

The criticality hypothesis suggests evolution has solved this problem. Drawing from concepts in statistical physics, the criticality hypothesis proposes that healthy brains operate near a tipping point, giving rise to a dynamical regime that maximizes raw computational capacity. At criticality, neural networks exhibit scale-invariant dynamics with optimal information processing properties: maximal dynamic range, enhanced signal transmission, and the ability to coordinate activity across multiple spatial and temporal scales simultaneously.

In this talk, I will present experimental evidence that criticality serves as a fundamental organizing principle across multiple domains of neuroscience. I will examine how developing neural circuits converge toward critical dynamics as they mature, proximity to criticality determines the capacity for learning, and how sleep may function to restore criticality, potentially solving a longstanding mystery in neuroscience. Finally, I’ll show that various neurological disorders—from epilepsy to Alzheimer's disease—involve systematic deviations from the critical regime. Using data from my lab, I will demonstrate how the same mathematical framework that describes phase transitions in physics can illuminate diverse aspects of brain function. These findings suggest that despite the brain's biological complexity, universal physical principles may govern neural computation, offering new perspectives on everything from cognitive development to therapeutic interventions in neurodegeneration.

The implications extend beyond neuroscience: if criticality represents an optimal computational regime shaped by evolutionary pressures, it may inform our understanding of intelligence, learning, and the fundamental constraints that govern complex adaptive systems.