We design and build a sub-symbolic artificial chemistry based on random boolean networks (RBN). We show the expressive richness of the RBN in terms of system design and the behavioural range of the overall system. This is done by first generating reference sets of RBNs and then comparing their behaviour as we add mass conservation and energetics to the system. The comparison is facilitated by an activity measure based on information theory and reaction graphs but tailored for our system.
The system is used to reason about methods of designing complex systems and directing them towards specific tasks.