Complex systems, particularly those that incorporate evolutionary mechanisms, offer the possibility of (apparent) creativity not present in most computer systems. The fundamental distinction between evolutionary systems (and artificially intelligent systems in general) and more traditional computer systems is that evolutionary (and AI) systems typically search for results rather than computing them in a more direct manner. Search has the potential to surprise and to appear creative because a true search may find a result that one didn't expect and perhaps didn't even know existed. Yet the creative potential of search is limited to the search space. The better one knows the structure and content of the search space, the less creative the search results will appear. We argue that the most creative searches require search spaces that include processes. In nature, biological evolution, and especially evolutionary arms races offer good examples of such search spaces: what evolves in natural evolution are biological and chemical processes. In computer systems, genetic programming is the best example of a search mechanism whose search space includes processes. Yet the use of genetic programming-like searches in agent-bases systems presents very serious challenges. The two primary approaches to dealing with those challenges are internal simulations and dispensable agents.
