An Artificially Intelligent Jazz Performer

This paper presents an intelligent agent model for simulating the behavior of a jazz bass player during live performance. In jazz performance, there is a strikingly large gap between the instructions given in a chord grid and the music actually being played. To bridge this gap, we integrate Artificial Intelligence (AI) methods within the intelligent agent paradigm, focusing on two critical aspects. First, the experience acquired by musicians in terms of previously heard or played melodic fragments, which are stored in the agent's ''musical memory''. Second, the use of these known fragments within the evolving context of live improvisation. In previous papers, we have presented a model for an improvising bass player, emphasizing the underlying problem solving method. In this paper, we describe the fully implemented model and give more details on how melodic fragments are represented, indexed and retrieved.