Markus Maier (HfPh München) will join us online on Thursday December 18th, 4 PM (CET) and deliver a talk titled A Philosophical Perspective on the Physics of AI.
Abstract
Despite our full access to their internal structure, deep neural networks (DNNs) are epistemically opaque to us; we cannot extract meaningful explanations from existing micro-level descriptions. Broadly speaking, “Physics of AI” denotes the application of physics to the study of DNNs in order to gain insights into their inner workings and (better) understand these systems. By analogy with complex phenomena in physical systems, physicists have recently begun to explicitly apply conceptual frameworks from theoretical physics – such as effective field theory (EFT) and renormalization – to DNNs. This talk will analyze The Principles of Deep Learning Theory (Roberts, Yaida, and Hanin, 2022) as one particular example of this approach from a philosophical, but technically informed, perspective.
Please join us in WebEx:
https://unistuttgart.webex.com/unistuttgart/j.php?MTID=mafe337cf67ee380f3a6816b9109d5ac5
