Emergent dynamics in neural populations recording
Modeling in neuroscience, and biology more generally, has firmly established itself in cases where it's been possible to reduce a system to a relatively small number of constitutive parts. In contrast, modern experiments often measure activity of thousands of heterogeneous neurons at a time. Building computational models at this level of detail has proven difficult (and maybe not useful), and we lack intuition about how to interpret results of such experiments. Put another way: When should we be surprised by what we see in high throughput recordings? I will show that simple models can explain seemingly surprising results of high throughput experiments in fields as diverse as neuroscience and immunology (my primary focus here will be on neuroscience), and I will argue that success of these models signals emergence of simpler, collective descriptions of complex biological systems. The goal now is to identify those collective degrees of freedom, and how they interact with each other. Overall, this progress raises hopes for building predictive models of the nervous system without neurons as the degrees of freedom.