 |
Center for Nonlinear Studies |
This talk introduces PHAMES (Physics-informed, Hybrid AI, and Mathematics for Engineering and Sciences), a transformative initiative at the University of Arizona aimed at advancing AI methods rooted in physical laws and mathematical rigor. PHAMES seeks to overcome the limitations of both traditional modeling and “off-the-shelf” AI by developing hybrid approaches applicable to critical domains: energy systems, aerospace, communications, new materials, and networks of influence. With strong interdisciplinary collaboration and an open invitation to national labs like Los Alamos, PHAMES is designed as a service-oriented, cross-campus AI infrastructure supporting research, education, and societal resilience.
In the second part of the talk, I will present recent technical work on sampling decisions (https://arxiv.org/abs/2503.14549) based on harmonic path integral diffusion, which provides a principled foundation for modeling, sampling, and learning in extreme or high-uncertainty regimes. This framework demonstrates how rigorous mathematical methods can enhance the transparency, robustness, and adaptability of AI models -- goals that are central to the PHAMES vision and our potential collaboration with Los Alamos.
Host: Sergei Tretiak (T-1)