Title: Thermodynamic Computing

Abstract: Concepts from thermodynamics are ubiquitous in computing systems today –

e.g. in power supplies and cooling systems, in signal transport losses, in device

fabrication and state changes, and in the abstractions and methods in machine learning.

In this talk I propose that thermodynamics should be the central, unifying concept in

future computing systems. In particular, I suppose that computing systems of the future

will thermodynamically evolve in response to electrical and information potential in their

environment and that thermodynamic evolution is the unifying idea that addresses the

central challenges of energy efficiency and self-organization in technological systems. I

present a few results from a novel thermodynamic neural network model that addresses

the core assumptions of this approach concretely. Although the talk focuses on the

domain of computation, the ideas are generic and derive from simple observations of the

everyday world. A key conclusion of the work is that causation is the product of

evolution, an idea that inverts the current philosophy of computation and challenges

many common assumptions about existence.