Examining the influence of AI on music Artificial intelligence technology is playing an increasingly prominent role in the music industry, from curating personalised playlists to producing tracks. The MusAI research programme brings together researchers from across a wide range of disciplines to look critically, through the lens of music, at the cultural implications of AI, which tend to have been ignored in recent years, as Professor Georgina Born explains. The way that we experience music is changing, with more and more people using AI-based systems that recommend new tracks based on their previous listening habits, rather than listening to the radio or going to concerts or gigs to encounter new sounds. At the same time, generative AI technology is also being widely used to produce music without the direct input of a composer or artist. “There’s a specialist AI music start-up sector focused on socalled production music, which generates music for TV, films, games and advertising,” explains Georgina Born, Professor of Anthropology and Music at University College London (UCL). Technologies like AIdriven recommendations—either through traditional recommender systems or new conversational agents—increasingly mediate our relationship with music and threaten to undermine the principle of universality, which public service media organisations like the BBC and others are committed to upholding. “Universality is the idea that it is really important for social cohesion as well as for the toleration of difference that people have shared experiences of information and culture, including music, because it generates a sense of shared community and shared cultural knowledge, of pluralistic cultural belonging,” explains Professor Born.
MusAI project Highly personalised playlists, by contrast, encourage fragmentation rather than shared cultural experiences, one of the issues that Professor Born is addressing as the director of the ERC-backed MusAI research programme, which researches the relationship between AI and music. The MusAI programme encompasses ten research projects, several of which are interdisciplinary in nature, bringing together researchers from very different backgrounds. “In some of our projects, researchers from the humanities and social sciences have been working closely together with computer scientists and data scientists over long periods,” Professor Born adds.
www.euresearcher.com
Overlaid Diagram: A directed acyclic graph of temporal interrelations at work in a musical genre, from the MusAI project “Sonic-Social Genre: Towards Multimodal Computational Music Genre Modelling”/ Credit and copyright: Owen Green. Dr. Artemi-Maria Gioti and Dr. Xenia Pestova Bennett in rehearsal for Gioti’s work Bias II, composed for MusAI’s artistic research strand “Permeable Interdisciplinarity: Algorithmic Composition, Subverted”/ Credit and copyright: Georgina Born.
This interdisciplinarity is a feature, for example, of a project centred on investigating the nature of musical genre and how it can be computationally modelled. “Genre is a very well-developed term in the humanities, also in relation to music. We talk often about musical genres and sub-genres; it’s one of the ways in which people describe the history and the experience of music,” says Professor Born. The way that computer scientists have modelled genre has historically been weak. However, through close interdisciplinary collaborations, the project is developing methods that better bridge computer science and the humanities, cultivating mutual understanding and improved forms of modelling of cultural processes. In doing so, the project innovates, identifying new ways for quantitative and qualitative methods to interrelate. It thereby creates new computational approaches to both music and culture, with broader implications for the digital humanities in general. MusAI researchers are also working across disciplinary boundaries in a project focused on developing alternatives to
the commercial big tech platforms’ recommender systems, a project in which Professor Born has collaborated for several years, sharing knowledge, ideas and insights, with an American computer scientist, Professor Fernando Diaz, an expert on recommender systems based at Carnegie Mellon University. “We’ve been training Fernando in social science ideas, like universality, while he’s been training us in how recommender systems work technically,” she continues. “This radical interdisciplinarity has resulted in an innovative new recommender paradigm based on a new metric, which is designed to encourage more common experiences.” The MusAI alternative recommender system represents a significant shift from the way conventional, personalised recommender systems work, which is typically based on maximizing a measure of an individual’s satisfaction, based on their previous listening history. This personalised recommender paradigm tends to narrow people’s tastes and limit their musical horizons. “There’s quite a large body of research now showing that
71