DrWave Ngampruetikorn

My research seeks to understand the fundamental principles that govern how microscopic components conspire to generate new, collective behavior at the macroscopic scale—from magnetism and superconductivity to bird flocking, synchronous neural activity and learning in deep networks.

I am particularly interested in treating AI systems not merely as an engineering problem but as a new frontier for fundamental physics. Current research themes in my group include the principles of in-context learning, the universality of neural computation, and the application of physics-motivated concepts, such as coarse-graining, to make AI more interpretable. Recent work has explored when in-context learning can generalize beyond its training distribution, the tradeoffs between generalization and specialization under concept shift, how data coarse graining can improve model performance and extreme synergy in spin glass models.