Sponsor
Physics
Co-Sponsor(s)
Chemistry
London Memorial Public Lecture - Tunable Matter as a Unifying Paradigm for Complex Emergent Function
Monday, March 2,
-
Speaker(s): Andrea Liu
How do collections of things-proteins in cells, cells in tissues or neurons in brains -act together to do complicated things that approach intelligence? To explore this foundational question, I will argue that it is useful to think
of complex function as emerging from systems made of tunable matter: collections of physical components that interact with each other in ways that can be adjusted to produce collective behavior. We already encounter this idea in familiar settings. The genome, for example, is tuned over generations through mutation and selection to enhance survival, while the brain adapts far more rapidly by tuning connections between neurons, and AI networks are trained by adjusting network parameters. I will introduce the broader concept of tunable matter, which extends far beyond genes and brains to many other systems. Through a few accessible examples, I will show how this perspective might allow us to build a unifying framework for the emergence of collective function
in biological systems, and how similar principles can be used to design electronic circuits that learn AI tasks on their own, without a computer, at remarkably low energy cost.
of complex function as emerging from systems made of tunable matter: collections of physical components that interact with each other in ways that can be adjusted to produce collective behavior. We already encounter this idea in familiar settings. The genome, for example, is tuned over generations through mutation and selection to enhance survival, while the brain adapts far more rapidly by tuning connections between neurons, and AI networks are trained by adjusting network parameters. I will introduce the broader concept of tunable matter, which extends far beyond genes and brains to many other systems. Through a few accessible examples, I will show how this perspective might allow us to build a unifying framework for the emergence of collective function
in biological systems, and how similar principles can be used to design electronic circuits that learn AI tasks on their own, without a computer, at remarkably low energy cost.