Physics Colloquium with Fan Wu

Fan Wu from Cornell University will present at this week's Physics Colloquium. Dr. Fan Wu's presentation is titled "Optics for AI and AI for Optics: From Optical Thermodynamic AI Accelerators to Machine-Learning–Enabled Reconfigurable Integrated Photonics".

The seminar will be presented at 4:00 p.m. on Thursday (Sep. 24) in Fisher 139. The coffee hour will be held at 3:30 in the Fisher Hall Lobby.

Abstract In the first part of this talk, I will introduce an optical thermodynamic framework that recasts complex nonlinear dynamics using entropy and statistical mechanics. By embracing the second law of thermodynamics, this framework provides a unified way to understand and harness these systems without tracking every degree of freedom. I will further show that equilibrium fluctuations in such systems can be used as a computational resource: through steady-state correlations, multimode systems natively perform matrix inversion, offering massively parallel, energy-efficient computing for AI. In the second part, a machine-learning-enabled reconfigurable photonic platform will be introduced to address the fabrication-reality gap in modern inverse-designed integrated photonics. In such system, the refractive-index landscape can be dynamically tuned by a top optical illumination on the chip. By combining this in-situ programmability with physics-informed machine learning, the same chip can perform various functionalities including mode conversion, matrix-vector multiplication, and dispersion engineering.

Bio Fan Wu obtained his Ph.D. in Optics from the University of Central Florida in 2020 and has conducted postdoctoral research at Cornell University since 2022. He will join the School of Engineering at Westlake University as an Assistant Professor in 2027. He was selected as a 2024 Springer Nature LSA, Rising Star of Light and received the 2024 Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship.