Physics of Learning and Neural Computation

This collaboration, directed by Surya Ganguli at Stanford University, aims to elucidate fundamental scientific principles of learning and neural computation underlying modern artificial intelligence. To achieve these aims, we treat AI as a complex physical system, and we employ and develop powerful techniques from physics and mathematics, in concert with computer science and theoretical neuroscience, to understand and improve how large neural networks learn, compute, scale, reason, and imagine. More information can be found on the collaboration's website.