Meet Dr. Michael Levy

Dr. Michael Levy

Dr. Michael Levy

Advisor: Dr. Massamiliano Di Ventra

Institution: University of California San Diego

Bio:  My scientific journey began with an application for a science writing internship at the Economist. As a highschooler I took the train from my Long Island hometown to meet the London based editor in NYC. After exclaiming "But you're just a school boy!", he offered me a position filling a maternity vacancy over my freshman winter break. Armed with faith in my ability to understand science, I went back to school aiming to learn the process thereof. At Brown University I majored in physics: working in computational quantum chemistry and writing a polymer physics thesis modeling constrained DNA. My PhD in Biophysics from UC Berkeley provided the opportunity to develop neural models of camouflage and seashell pattern formation, study the dynamics of subcellular neuronal processes, and perform a mathematical analysis of computational dynamics at the edge of chaos and synchrony. My dissertation focused on how biological systems make strange feedback loops with distributed systems aligning their sub-components to make patterns which seemingly require more information than locally provided. I have taught in the Neuroscience, Bioengineering, Physics, and Cognitive Science departments at UC Berkeley, was a Lead Projects TA for the international Neuromatch Academy, and have taught algebra in both Middle School and San Quentin Prison. I come to the Intelligence Community after a short UC Berkeley post-doc stint developing methods to computationally steer the directed evolution of viral capsid proteins.

Abstract:  Memcomputing is a novel compute framework leveraging bi-directional logic and the physical time-dependence of circuit components to solve otherwise intractable problems. In a physical chip the problem solution would be near instantaneous and the Di Ventra group has found that surprisingly simulation of these circuits allow the solution of problems supposedly requiring quantum computation. In simulated dynamics of memcomputation scaling laws for the distribution of element activity and periods of system wide quiescence are qualitatively similar to experimentally measured distributions of similar observables in the nervous system. We are currently interpolating between models replicating neuronal activity distributions and our emerging understanding of the topological computations underlying the success of memcomputation. An understanding of how scaling laws signify the computational capability of a system would allow the development of new measurements of the computational capacity of a behaving network. This would be widely applicable to understanding the formation of patterns in nature, assessing the computational capacity of an emergent system, and quantifying the decision making capabilities of distributed organizations.