Disclaimer: her talk might challenge how you think about learning and memory!

When we speak of computation, learning, and prediction, we typically think of organisms endowed with nervous systems. Neuronal networks use memory to extract regularities from a dynamic world, anticipate future states, and flexibly adapt behavior. However, both freely living single-cell organisms ans cells embedded in tissues, also operate in fluctuating, metastable environments where responses depend on the integration of past experience and present context. We therefore ask: how do the computational capabilities of single cells differ from those of neurons? Are neurons uniquely specialized for learning and prediction, or do they represent one realization of a more general dynamical principle?

Combining quantitative experiments with theoretical modeling, we demonstrate that even epithelial cells utilize a form of molecular working memory that enables them to integrate temporally varying signals, navigate changing environments, and exhibit basic forms of learning. Based on these findings, we develop a unified framework of dynamical learning that applies to both biochemical network in single cells as well as neuronal networks. In this view, computation, memory, and learning are not exclusive hallmarks of neuronal circuits, but rather learning is a fundamnetal property of living systems operating in non-stationary environments.