Can a Computer be Lucky? And Other Ridiculous Questions Posed by Computational Creativity

Given the fragility of today’s intelligent systems, we consider the necessity of creativity in systems designed for artificial general intelligence. We examine an archetypical creativity “algorithm” suggested by Czikzentmihalyi in the context of computational systems, and, in particular consider the computability of such an algorithm. We argue that it is likely not computable, in the Turing sense, but that this need not necessarily preclude the building of computationally creative systems, and, by extension, (potentially) systems with a level of artificial general intelligence.