AI in music performance: the human-AI performance loop

Music is a distinctly human artform, one that has been written, taught, shared, and reinterpreted for thousands of years. The rise of powerful artificial intelligence (AI) is reshaping musical creation in ways subtle and profound. Here I propose the human-AI performance loop as a framework for understanding these developments through three principal roles: the AI listener derives musical information from sound and movement; the AI coach guides practice using performance evidence and musicians' learning goals; and the AI partner performs with the musician in a collaborative dynamic process. With the human musician at the centre of the framework, human → AI → human exchanges become coupled through shared information and musicians' actions. Evaluating these roles will require multilayered evidence to support their adoption: accurate and musically relevant analysis, demonstrated improvements in musician performance and knowledge through coaching, and effective and supportive coordination between musicians and AI partners. Audience beliefs about AI will also shape reception of the creative output, making musical purpose and honest representation important considerations in an AI music performance system. The human-AI performance loop provides a basis for investigating how these roles can extend support between lessons and rehearsals, broaden opportunities for ensemble performance, and contribute to musicians' learning and enjoyment of the musical experience.