Deep Optimisation: Multi-scale Evolution by Inducing and Searching in Deep Representations

The evolution of life on earth has been driven by a small number of major evolutionary transitions. These transitions have been characterized by individuals that could previously replicate independently, cooperating to form a new, more complex life form. For example, archaea and eubacteria formed eukaryotic cells, and cells formed multicellular organisms. However, not all cooperative groups are en route to major transitions. How can we explain why major evolutionary transitions have or haven't taken place on different branches of the tree of life? We break down major transitions into two steps: the formation of a cooperative group and the transformation of that group into an integrated entity. We show how these steps require cooperation, division of labor, communication, mutual dependence, and negligible within-group conflict. We find that certain ecological conditions and the ways in which groups form have played recurrent roles in driving multiple transitions. In contrast, we find that other factors have played relatively minor roles at many key points, such as within-group kin discrimination and mechanisms to actively repress competition. More generally, by identifying the small number of factors that have driven major transitions, we provide a simpler and more unified description of how life on earth has evolved.

The theory of evolution links random variation and selection to incremental adaptation. In a different intellectual domain, learning theory links incremental adaptation (e.g., from positive and/or negative reinforcement) to intelligent behaviour. Specifically, learning theory explains how incremental adaptation can acquire knowledge from past experience and use it to direct future behaviours toward favourable outcomes. Until recently such cognitive learning seemed irrelevant to the 'uninformed' process of evolution. In our opinion, however, new results formally linking evolutionary processes to the principles of learning might provide solutions to several evolutionary puzzles - the evolution of evolvability, the evolution of ecological organisation, and evolutionary transitions in individuality. If so, the ability for evolution to learn might explain how it produces such apparently intelligent designs.