Abstract
As opposed to the traditional view wherein intelligence was believed to be a result of centralised complex monolithic rules, it is now believed that the phenomenon is multi-scale, modular and emergent (self-organising) in nature. At each scale, the constituents of an intelligent system are cognitive units driven towards a specific goal, in a specific problem space—physical, molecular, metabolic, morphological, etc. Recently, Neural Cellular Automata (NCA) have proven to be effective in simulating many evolutionary tasks, in morphological space, as self-organising dynamical systems. They are however limited in their capacity to emulate complex phenomena seen in nature such as cell differentiation (change in cell’s phenotypical and functional characteristics), metamorphosis (transformation to a new morphology after evolving to another) and apoptosis (programmed cell death). Inspired by the idea of multi-scale emergence of intelligence, we present Hierarchical NCA, a self-organising model that allows for phased, feedback-based, complex emergent behaviour. We show that by modelling emergent behaviour at two different scales in a modular hierarchy with dedicated goals, we can effectively simulate many complex evolutionary morphological tasks. Finally, we discuss the broader impact and application of this concept in areas outside biological process modelling.