Building Brain-Inspired Computers with Nonlinear Dynamical Materials

Stan Williams, Texas A&M
Location: Online

Friday, 22 October 2021 02:30PM PDT

*Register at the following link to join seminar:
Access link will be sent to your email address after registration


With the near saturation of Moore’s scaling of transistors, there has been a recent explosion in activity and creativity to find new modes of computation that will continue to advance exponentially with time even though transistor circuits only improve modestly.  Much of the inspiration for new ways of computing comes from what little we understand about the brain.  We actually don’t know how the brain computes, but many different possibilities have been proposed, e.g. multinary logic, Neural Networks of all kinds, extensions of Hebbian learning via spike-timing dependent plasticity, Boltzmann/Ising machines, Hopfield networks, Bayesian inference and Markov chains, to name a few.   These possibilities are not necessarily mutually exclusive – the brain may use some combination of several of them or even use a higher order generalization that contains most of them, since many share mathematical similarities.  How to express these computational approaches efficiently in an electronic circuit is a significant challenge.  Since the brain itself is a highly nonlinear dynamical system, an appropriate area to investigate is nonlinear dynamical circuit theory.  This is the realm of the Principle of Local Activity, which provides a basis for inventing and building new generations of nanoscale memristor-based oscillators and amplifiers that emulate the integrate and fire dynamics of neurons to produce action potentials for signal processing, learning and computation. Combining such devices enables the design of ‘neuromorphic’ circuits that are biased at the Edge of Chaos, where the emergence of complex patterns and behavior in a homogeneous medium are found.  Thus, the goal of my research is not to improve transistors but rather to create new analog electronic circuit elements that can replace hundreds to thousands of transistors in a brain-inspired computer.   Can we emulate neural computation using new electronic-ionic-thermal devices that exhibit dynamical behavior?  These could be based on properties such as negative differential resistance and/or capacitance that arise from thermally activated transport, Mott transitions, molecular redox chemistry, or even electro-mechanical response in soft matter.  Can we intentionally design materials that will function in the Goldilocks zone of normal room temperature?  This is a rapidly growing multi-disciplinary research field that requires coordinated efforts across a broad range of fields from neurophysiology, chemistry, materials, solid-state physics, electrical engineering and computer science.  Early results are promising, with several experimental demonstrations of neuromorphic circuits outperforming digital CMOS and even a quantum computer by several orders of magnitude in terms of energy efficiency and time to solution for certain types of ‘hard’ problems.