UC Press Blog

The first issue of the newly launched journal, Music and Data, from the University of California Press, is dedicated to the founding of a new field called, “AI Music Studies.” It consists of seven articles and two introductions from scholars who attended and presented at The First International Conference in AI Music Studies (December 10-12, 2024, Stockholm, Sweden), which was organized by Bob L. T. Sturm, Elin Kanhov, and Andre Holzapfel, guest editors of Music and Data's new issue. We sat down with two of the issue's editors, Bob L. T. Sturm and Elin Kanhov, to ask them to tell us more.

What inspired you to organize this conference?

Back in early 2023, the MUSAiC project team was having lunch and someone made a joke about seriously studying the collected works of AI music generation systems (like Boomy and an early version of Suno) as seriously as studying the works of Beethoven or Mozart. We riffed on the idea and laughed about writing academic papers arguing for stylistic epochs in the prodigious production of AI composers—as if there are any grounds to treat the Boomy song "Stonks Go Up" by Stonky Stonks with the same seriousness as the lieder “Erlkönig” by Franz Schubert. Around the same time, The Atlantic published an essay by US English professor Matthew Kirschenbaum reflecting on his encounter with the then newly released ChatGPT: “Prepare for the Textpocalypse: Our relationship to writing is about to change forever; it may not end well.” The idea that, on the internet and maybe beyond, machine-generated text was going to become the rule and human-crafted text the exception was becoming realistic. Could it be the same for music? Even a year earlier the CEO of Boomy mused about AI music systems creating and releasing “100 billion songs a year,” thus swamping all of human musical creation. The MUSAiC team pondered this funny situation more and began to realize that the musical and sound quality of AI-generated output was only going to improve: “A flood is coming—are we prepared?” in other words. 

In the research domain of music information retrieval, which aims to make music as accessible as Google search makes text, we could see some interesting problems ahead. How will recommendation work when music catalogues consist of billions of recordings, most of which no one has listened to (even in its creation)? Are the methods and methodologies of the discipline prepared for this coming flood of data? Sturm assembled 12 scholars from a variety of disciplines and wrote a conference paper submitted to the 2023 International Symposium on Music Information Retrieval. We argued that there is a lack of formal studies of a new music ecosystem that is quickly materializing: music generated using artificial intelligence, “AI music”. We proposed analyzing this ecosystem using perspectives of many disciplines, including engineering, musicology, science and technology studies, and economics. We proposed “a new kind of music studies”. Our submission was rejected, as well as revised submissions to two other conferences. We were so convinced by the need for a sustained multidisciplinary effort to address these emerging questions that we decided to just organize our own conference. (We posted our rejected paper on SocArXiv, but a revision was accepted and published at AI Music Creativity 2024.)

It’s been almost two years since the conference. What has happened since then?

It’s no surprise we are now seeing streaming services such as Deezer claiming they are receiving 75,000+ music tracks a day they suspect of being AI-generated. Last year, Spotify claimed they deleted over 75,000,000 “spammy tracks” from their catalogue. Online music generation platforms are accessible, music can easily be distributed to many digital service providers, and pay-per-play streaming media incentivizes flooding catalogues with music tracks in the hopes that some will find their ways to popular playlists. We now see “fake artists” popping up and garnering millions of streams. AI-generated music is charting in many places. The identities of deceased artists are being hijacked to release new music to game the pay-per-play remuneration system. AI music generation platforms are being used to traffic hate speech through popular musical forms. We are even seeing the emergence of AI music award shows, e.g., The AI Song Contest (founded in 2020 by academics, but moving mainstream). 

It’s no surprise then that the AI Music Studies conference has continued! The Second International Conference in AI Music Studies was organized earlier this year at the Department of Music at the University of Nottingham March 30-April 1. And the third edition of the conference will occur December 1-3 2027 at the Department of Musicology at the University of Oslo. The successful continuation of this conference shows the need for a forum where voices from a broad array of disciplines can gather to critically discuss AI music generation technology and its application. 

What would you like to see happen as AI Music Studies develops?

For the future of AI Music Studies, we’d love to continue seeing scholars from the humanities and social sciences bring their expertise to the complexities of AI music to counter the otherwise engineering and computer science-heavy scholarship on AI music. Journals such as Music and Data are crucial for making that possible. We expect much more research will be needed in media and platforms studies, music aesthetics and philosophy, gender studies, economics, law, and environmental studies, to name a few central disciplines, to the study of AI music. Finally, we hope to see AI Music Studies contribute to the development of public policies in the arts, and to support sustainable, thriving and human-centred art practices at large.

We invite you to read the first issue of Music and Data, "AI Music Studies," for free for a limited time.

Print copies of Music and Data's inaugural issue 1.1 can be purchased on the journal’s site. 

For ongoing access to Music and Data, please ask your librarian to subscribe and/or purchase an individual subscription. 

Music and Data is also available as part of our institutional UC Press Music Subject Collection, which packages all of our music journals together for one low subscription price. You may want to consider asking your librarian to subscribe to the full collection.