Abstract by Paul Bodily
Computational Creativity: Machine Learning in Music Composition
Artificial intelligence is an integral part of our day-to-day lives. We speak commands to our phones, filter out spam, get directions, and so on. Though computers are helpful for many things, when it comes to creativity they lag woefully behind. Of particular interest is the way in which people understand and create music to enlighten, empower, and inspire themselves and others.
Using machine learning and constraint programming techniques, we are seeking to model creativity in a way that enables computers to understand and compose music in terms of human-level concepts. This includes problems such as inferring global structure, generating meaningful motifs, effectively combining harmony and melody, and creating lyrics that are syntactically and semantically coherent. These technologies are relevant for recommendation and generative music systems. Our goal is to develop systems that would empower and inspire individuals—particularly those lacking musical proficiency—to more effectively use the music composition process for coping with life challenges, such as anxiety, depression, and learning and memory.