M1: Conversations with our Digital Selves: the development of an autonomous music improviser

Mark d’Inverno is a critically acclaimed jazz pianist (BBC TV & National Radio, BBC Music Magazine, Guardian, Observer, Jazz Review) who has toured nationally and internationally since the 1980s. He is also a full professor at Goldsmiths, University of London, taking on the role in 2006. He is recognised as a leading academic researcher exploring the relationship between Artificial Intelligence and Human creativity, and his invited talks included improvised piano performances with AI systems. He has over 250 peer-reviewed publications, was formerly Director of Research at Goldsmiths, and currently holds honorary research positions at Instituto de Investigación en Inteligencia Artificial (IIIA) Universitat Autònoma de Barcelona and Politecnico di Milano, Italy.

This paper describes a series of real-time music improvisation systems we have created over the last 15 years. The systems are all based on variable-order Markov models. We explain our iterated design process focused on creating collaborative AI systems to enhance human musical experiences. We have used the systems to play piano, drums and even to control a set of live effects and loopers. The systems have performed at live concerts with professional musicians, with some performances being broadcast on national radio. The system is sufficiently well-developed that professional and amateur musicians can now use it in various improvising contexts without our direct support.