How can non-equilibrium statistical physics be used to uncover fundamental physical constraints of computation in dynamic, highly non-equilibrium systems like computers? (photo: iStock)

Ten percent of carbon burned on Earth comes from devices like your cellphone, quietly computing even when you aren’t looking at it. 

Developing a greater understanding of the thermodynamics of computation like that done by your cellphone is critical to reducing energy use. It is also critical to understanding a host of deep, long-standing scientific problems. 

The thermodynamics of computation governs the amount of energy used by complex systems that are not at equilibrium because they are constantly processing information and evolving. This includes everything from computers to the human brain, yet research in the field is almost non-existent. 

To address the knowledge gap, the Santa Fe Institute is convening a panel of experts Aug. 15–17 for a workshop, "The Thermodynamics of Natural and Artificial Distributed Computational Systems," to identify challenges, opportunities, and priorities to push forward scientific investigations of this topic. 

“Advances in non-equilibrium statistical physics over the last 20 years provide us with the tools for the first time to investigate the energetic attributes of non-equilibrium systems, which is central for everything from physics to biology to not cooking the planet,” said SFI professor David Wolpert. “In other words, we suddenly have this massive opportunity in science as a whole and we have no idea what we will find through that door.”

Wolpert, along with former SFI Complexity Postdoctoral Fellow Joshua Grochow, a computer scientist at the University of Colorado, Boulder, and other SFI collaborators, plans to discuss how the recent breakthroughs in non-equilibrium statistical physics can be applied to uncover fundamental physical constraints of computation in dynamic, highly non-equilibrium systems like computers. 

They will focus on both naturally occurring and artificially distributed computational systems with three chief characteristics: they are distributed, comprising a set of spatially separated non-identical subsystems; the subsystems interact with one another in a hierarchical, modular network; there are substantial thermodynamic costs of communication among, and within, those subsystems. 

The National Science Foundation–sponsored* event will ultimately inform research to advance fundamental understanding and practical applications such as reversible computing.

Findings from the workshop will be collected in a report to the NSF, posted to the preprint server arXiv, and submitted to the Proceedings of the National Academy of Sciences.

* NSF Award 2145170