Abstract
Current Mixed Initiative Co-Creative systems (MICC) have opened new avenues for musicking with machines. However, these systems lack the ability for long-term adaptation between user and machine. Current ways of adapting are either unilateral, where the user adaptively learns to operate the AI system, or self-engineered, where musician-engineers modify their own systems over time. The recently started research project MICCDroP aims to address this limitation by developing AI systems for music performance based on lifelong learning, with a mix of reinforcement learning, curiosity-driven learning, and autotelic agents. These systems will be designed by putting rehearsing and practicing with AI as the key aspect of musicking. This allow us to explore the evolution of human-AI partnerships using ethnographic studies and creativity theory, as well as conduct artistic experimentations and live performances to inform future research directions. In this presentation, I will introduce the theoretical and practical underpinnings of the project, current design of musical spaces we want to explore, as well as future prospects.
Bio
Ken Déguernel is a CNRS Researcher in computer music. His main research topics are computational creativity for musical co-creativity, for which he develops new computer tools for musician-machine interactions and studies the emergent dynamics of co-creativity, and AI Music studies, for which he analyses the impact of AI technology on the music ecosystem, in particular on how AI music is listened to, created/performed and criticised. Website: deguernel.discordia.fr