Abstract

The connection between entropy production and the fitness of living organisms has been a recurring question at least since Schrodinger’s famous investigations. One way to consider this connection is to note that living organisms can be seen as catalysts, helping to transform accessible energy in their environment into entropy (typically released as heat) in a way that improves their fitness. Importantly though, organisms are “intelligent” catalysts; they perform computations to determine exactly how to try to improve their fitness. These computations in turn require energy to perform. As a result, in general, the smartest computational behavior of an organism is the most energetically costly, while the dumbest computational behavior is the least energetically costly. So the question at stake is two-fold:

1. At the scale of an individual organism (be it a unicellular creature, a single cell in a multicellular organism, or an individual in a eusocial species), what are the inherent physical constraints relating to thermodynamic efficiency, fitness, and intelligence?
2. At the scale of natural selection, how should the theorems of evolutionary biology be modified to account for the trade-off between the thermodynamic costs of performing computation and the fitness benefits of computations?

To address these questions, we propose to focus on recent results of stochastic thermodynamics, a revolutionary new development in statistical physics that allows us to investigate systems evolving arbitrarily quickly, while arbitrarily far from thermodynamic equilibrium. In particular, recent results such as fluctuation theorems, speed limit theorems, thermodynamic uncertainty relations, kinetic uncertainty relations and optimal stopping time theorems, all give us clues about how changing an organism’s computational behavior to improve its fitness  (e.g., by computing more accurately, more quickly, or more powerfully) necessarily increases the energetic requirements of the organism.