Meeting Sumnmary: The thermodynamics of computation is a topic of long-standing interest in both the physics and computer science communities, with major practical applications for issues ranging from the design of artificial digital systems to the foundations of physics. The revolution in non-equilibrium statistical physics of the past two decades, sometimes summarized as “stochastic thermodynamics”, has provided a major advance in our ability to investigate this topic.
Despite the great success of this recent work, we currently know essentially nothing about the fundamental thermodynamics of computational systems that: (i) are distributed, with many spatially separated non-identical subsystems (e.g., arranged in a hierarchical, modular fashion); (ii) are not at thermodynamic equilibrium (and in general, not even in a stationary state); and (iii)
Have both substantial thermodynamic costs of communication among the subsystems and thermodynamic costs of the information processing within the subsystems. In addition, we currently know almost nothing about how the thermodynamic costs of computations are related to properties like accuracy, robustness, and fault tolerance.
In this workshop, we will exploit new developments in non-equilibrium statistical physics and information processing theory to develop foundations for theories of the physical processes underlying computation. In doing so, we will start to address the fundamental gaps in our understanding of the thermodynamics of computation. We aim to develop greater understanding of the tradeoffs among factors like computational power, computational fault-tolerance and robustness, and thermodynamic efficiency, with a particular focus on how the communication network relating different subsystems affects these factors. We will consider both naturally occurring distributed computational systems, like brains and genetic networks, as well as artificial, electronic distributed computational systems.