For Responsible Artificial Intelligence Serving Inventive and Diverse Musical Creation
Opinion column published in Culture et Recherche n°147 (autumn-winter 2024, 138 pages), pp. 32-33
In the face of the myth of automating creative tasks, proposing alternative visions has become essential to ensure the survival and diversity of the musical ecosystem. The Institute for Research and Coordination in Acoustics/Music (IRCAM), where musical creation intersects with scientific research and technological development, advocates for a musical craftsmanship that integrates artificial intelligence (AI) responsibly and inventively from algorithm design to applications in artistic creation.
Challenges to Address
The spectacular rise of new technologies driven by artificial intelligence (AI) offers promising opportunities for artistic creation, particularly in sound and music. While AI provides advancements that align well with the needs of the creative and cultural industries, it also exacerbates the ecological and ethical challenges of digital technology and raises economic and artistic challenges that disrupt the ecosystem of musical creation.
An artistic and cultural challenge : The concentration of technological resources among a few dominant players, the standardization of algorithms, and the inevitable persistence of biases in datasets cause a high risk of increasing uniformity of AI-generated music. Traditional music, experimental music, and, more broadly, music that falls outside the mainstream industry’s norms face the risk of being increasingly marginalized. Adding to this, the fact that AI-generated content will inevitably spill into future datasets highlights a real threat of regression and impoverishment. Beyond musical creation, the issue of cultural diversity represents a profound anthropological challenge for our societies.
An economic challenge: A recent study commissioned by CISAC estimates that by 2028, 24% of music creators’ incomes will be at risk due to AI; this is equivalent to a €10 billion loss over the next 5 years. For instance, music generation applications based on text prompts, capable of quickly producing a piece of music from a brief verbal description without paying session fees or royalties, directly compete with human-driven musical creation. Major imbalances between industrial production and artisanal creation in the music ecosystem are expected in the coming years.
A Different AI is Possible
In light of these issues, we aim to promote a vision of AI that serves unique artistic projects, capable of blending musical and technological expertise in a spirit of experimentation and hybridization, so that AI contributes responsibly to the enrichment of a fully diverse creation.
Responsibility begins with the design of algorithms. Research into frugal AI, which directly addresses the energy efficiency and ecological challenges of AI, must ultimately offer sustainable alternatives to current systems, significantly reducing needs in terms of energy and data. This approach aligns with the use of curated datasets that are tailored to specific applications and remain respectful of the consent of the individuals whose data is used, contrasting with the current race to collect massive datasets exploiting legal gray zones. More broadly, new economic models that ensure fair distribution and remuneration for artists must be envisioned and explored.
Following the EU AI Act’s recommendations on data transparency, we also advocate for transparency of the usages, covering means and purposes. Concerning the purposes, this includes systematically informing the public about the use of AI in producing publicly accessible content and ensuring its traceability from creation to distribution. At the means level, it involves specifying the processes employed to produce AI-generated content, as was done for General De Gaulle's Appeal of 18 June, whose voice recreation realized in January 2023 relied primarily on human interpretation choices.
Beyond these practical proposals, we want to promote diversity, both in technological innovations and in the uses of AI for musical creation. Alongside music generation from textual prompts, other experiences—participatory and immersive—and other interaction modes—combining sound and gesture, improvisers and machines, artists and audiences —should be explored. The diversity and plurality of technological and musical practices are essential for the sustainability of the creative ecosystem, where majority and minority music practices mutually enrich each other. Furthermore, market mechanisms alone are likely insufficient to sustain diversity in the AI era. We call for reaffirming and generalizing the principle of cultural exception, particularly through redistribution mechanisms supervised by public authorities at the European level.
AI, like other technological innovations that have shaped the history of music, is a powerful tool for stimulating new desirable ways of experiencing listening, performing, and creating sound and music. More than ever, it is up to musicians to embrace these tools to shape the present and future of music.
Signatories (by alphabetical order) :
Frédéric Amadu, CTO, Ircam Amplify
Pavlos Antoniadis, Collaborator, IRCAM, STMS*
Greg Beller, Product Owner Ircam Forum, IRCAM
Maher Benchekroun, Lead PMO, Ircam amplify
Cyril Béros, Musical Production Director, IRCAM
Frédéric Bevilacqua, Research Director, IRCAM, STMS*
Nathalie Birocheau, CEO, Ircam Amplify
Clément Canonne, Research Director, CNRS, STMS*
Nicolas Dauban, Engineer, Ircam Amplify
François-Xavier Féron, Research Fellow, CNRS, STMS*
Jean-Louis Giavitto, Research Director, CNRS, STMS*
Nadia Guerouaou, Post-Doctorate, IRCAM, STMS* and FEMTO-ST
Pierre Guillot, Developer, IRCAM
Thomas Hélie, Research dDirector, CNRS, STMS*
Elias Karam, Head of Business Development, Ircam Amplify
Carlo Laurenzi, Computer-Music Designer, IRCAM
Serge Lemouton, Computer-Music Designer, IRCAM
Jean-Yves Le Porcher, Sound Experience Director, Ircam Amplify
Frank Madlener, Director, IRCAM
Nicolas Misdariis, Research Director, IRCAM, STMS*
Markus Noisternig, Head of Musical Research and Research Fellow, IRCAM, STMS*
Nicolas Obin, Associate Professor, Sorbonne University, STMS*
Charles Picasso, Engineer, CNRS, STMS*
Johannes Regnier, Computer-Music Designer, IRCAM
Axel Roebel, Research Director, IRCAM, STMS*
Sébastien Roux, PhD student and Computer-Music Designer, Sorbonne University, STMS*
Pierre Saint-Germier, Research Fellow, CNRS, STMS*
Romain Simiand, CPO, Ircam Amplify
Coralie Vincent, Engineer, CNRS, STMS*
Hugues Vinet, Director of Innovation and Research Means, IRCAM
* The STMS research lab is a joint research unit (UMR 9912) housed at IRCAM under the supervision of CNRS, Ircam, the French Ministry of Culture and Sorbonne University.