Discovering sensorimotor agency in cellular automata using diversity search

The field of artificial life studies how life-like phenomena such as agency and self-regulation can self-organize in computer simulations. In cellular automata (CA), a key open question is whether it is possible to find environment rules that self-organize robust "individuals" from an initial state with no prior existence of things like "bodies," "brain," "perception," or "action." Here, we leverage recent advances in machine learning, combining algorithms for diversity search, curriculum learning, and gradient descent, to automate the search of such "individuals." We show that this approach enables us to systematically find environmental conditions in CA leading to self-organization of basic forms of agency, i.e., localized structures that move around and react in a coherent and highly robust manner to external obstacles, maintain their integrity, and have strong capabilities to generalize to new environments. We discuss how this approach opens new perspectives in artificial intelligence and synthetic bioengineering.