Marc Serra Garcia, AMOLF, Amsterdam

Leuchs-Russell Auditorium, A.1.500, Staudtstr. 2

Location details

Abstract:

Physical systems can be thought of as computers: their dynamical variables store information and evolve according to update functions acting as pre-programmed algorithms. Although nature offers a vast range of physical phenomena, traditionally, CMOS electronics has been used for computation. et CMOS electronics typically consumes orders-of-magnitude more energy than thermodynamic limits allow, and mandates the transduction of all signals into the electronic domain. In this talk, I will provide an overview of our efforts to build computers that leverage more of what physics has to offer. I will do so through three examples, discussing speech recognition based on mechanical vibrations, self-learning algorithms that leverage symmetry and thermal noise, and digital circuits constructed out of nonlinear topological pumps. These phenomena are unconventional, in the sense that they deviate from standard practice in computer-building, yet present distinct advantages in their respective application domains.