Title: Computing with Physical Systems

Abstract: With conventional digital computing technology reaching its limits, there has been a renaissance in analog computing across a wide range of physical substrates. In this talk I will introduce the concept of Physical Neural Networks [1] and describe a method my group has developed to train any complex physical system to perform as a neural network for machine-learning tasks. We have tested our method experimentally with three different systems – one mechanical, one electronic, and one photonic – and have been able to show MNIST handwritten-digit classification using each of these systems. We are excited to try our method with other physical systems, especially ones that have sufficiently complex and energy-efficient dynamics that they can plausibly deliver an advantage over conventional digital computers. We are interested in many different physical platforms, but have a particular focus on photonic approaches to computing. As a highlight, I will discuss our recent demonstration [2] of an optical neural network that uses less than one photon per weight multiplication, which at first hearing might seem impossible but actually works!

[1] L.G. Wright*, T. Onodera* et al. Nature 601, 549-555 (2022)

[2] T. Wang et al. Nature Communications 13, 123 (2022)

Bio: Peter McMahon is an assistant professor in Applied and Engineering Physics at Cornell University, where he has been since 2019. Prior to joining Cornell he completed his Ph.D. and postdoctoral training at Stanford University. He is the recipient of Packard and Sloan Fellowships, an Office of Naval Research Young Investigator Program Award and a Google Quantum Research Award, and is a CIFAR Azrieli Global Scholar in Quantum Information Science.