Toward a theory of evolution as multilevel learning

Abstract

We apply the theory of learning to physically renormalizable systems in an attempt to outline a theory of biological evolution, including the origin of life, as multilevel learning. We formulate seven fundamental principles of evolution that appear to be necessary and sufficient to render a universe observable and show that they entail the major features of biological evolution, including replication and natural selection. It is shown that these cornerstone phenomena of biology emerge from the fundamental features of learning dynamics such as the existence of a loss function, which is minimized during learning. We then sketch the theory of evolution using the mathematical framework of neural networks, which provides for detailed analysis of evolutionary phenomena. To demonstrate the potential of the proposed theoretical framework, we derive a generalized version of the Central Dogma of molecular biology by analyzing the flow of information during learning (back propagation) and predicting (forward propagation) the environment by evolving organisms. The more complex evolutionary phenomena, such as major transitions in evolution (in particular, the origin of life), have to be analyzed in the thermodynamic limit, which is described in detail in the paper by Vanchurin et al. [V. Vanchurin, Y. I. Wolf, E. V. Koonin, M. I. Katsnelson, Proc. Natl. Acad. Sci. U.S.A. 119, 10.1073/pnas.2120042119 (2022)].

Thermodynamics of evolution and the origin of life

“…Indeed, learning, by definition, decreases the uncertainty in knowledge and thus should result in entropy decrease. In the accompanying paper, we describe deep, multifaceted connections between learning and evolution and outline a theory of evolution as learning (18). In particular, this theory incorporates a theoretical description of major transitions in evolution (MTE) (19,20) and multilevel selection (21)(22)(23)(24), two fundamental evolutionary phenomena that so far have not been fully incorporated into the theory of evolution.…”

“…The crucial step in treating evolution as learning is the separation of variables into trainable and nontrainable ones (18). The trainable variables are subject to evolution by natural selection and, therefore, should be related, directly or indirectly, to the replication processes, whereas nontrainable variables initially characterize the environment, which determines the criteria of selection.…”

“…According to Sir Francis Bacon's famous motto scientia potentia est, better knowledge and hence improved ability to predict the environment increases chances of an organism's survival and reproduction. However, derivation of the Malthusian fitness from the properties of the learning system requires a detailed microscopic theory such as that outlined in the accompanying paper (18). Here, we look instead at the theory from a macroscopic perspective by developing a phenomenological description of evolution.…”

“…Evolutionary temperature is yet another key term in our vocabulary (Table 1), after entropy, which emerges as the inverse of the Lagrange multiplier β that imposes a constraint on the average loss function [2.2], that is, defines the extent of stochasticity of the process of evolution. Roughly, free energy F is the macroscopic counterpart of the loss function H or additive fitness (18), whereas, as shown below, partition function Z is the macroscopic counterpart of Malthusian fitness,…”

“…The relation between the loss function and fitness is discussed in the accompanying paper (18) and in Ideal Mutation Model.…”