Conference Agenda

We are in a moment of transition between two standards for AI-based music generation: one of custom models built for specific tasks and one of generic music models, such as pre-trained music language models (LM) that work at a broad level but lack fine-grained customizability. Substantial research in music technology has so far pursued the former approach by building corpora and models for specific music traditions. More recent projects, however, offer immense capabilities for music generation by harnessing LMs, such as Google’s MusicLM and Meta’s MusicGen. To what extent is it possible to bridge the gap between the old and new eras? What are the possibilities and limitations of such models for non-Western music traditions? How do multimodal AI models, which incorporate the listener’s embodied feedback, push the boundaries of pre-existing cultural frameworks for audience–performer interaction, listener engagement, and ecstatic experience? How might scholars, developers, performers, music educators, and audiences effectively utilize these new AI-driven tools?

These questions intersect with longstanding concerns in (ethno)musicology regarding virtuosity, human–machine interaction, musical ecstasy, and listener experience. Our proposed workshop reframes such issues for the new world of “generative AI,” drawing on inter/transdisciplinary approaches to better understand AI’s potential for music composition and accompaniment. The workshop centers on AI applications for Arab music performance, specifically accompanied vocal improvisation (mawwal) and solo instrumental improvisation (taqasim).

The first half of the workshop will feature 10-minute presentations (in-person and online) across the fields of music technology, music education, and ethnomusicology. The remainder of the workshop will focus on demonstrations of the AI computer application Mawaweel and a fine-tuned model of Meta’s MusicGen for Arab music improvisation. The audience will be invited to sing with the machine and engage with the models.

The workshop presenters will draw upon their contributions in developing, testing, and using AI models for Arab music, with the aim of translating their technical research for a humanities audience. Al-Ghawanmeh will address the challenge of implementing generative AI for Arab maqam music, a non-Western musical idiom that is beyond the scope of current AI models. He will discuss the significance of using pre-trained LMs for Arab musical practices, which have limited curated datasets available for training models. Jensenius will elaborate on the potential and current ability of motion-capture technologies to detect real-time audience feedback. Nsairat will discuss the potential of AI music tools for early music education in Jordan. Scott will reflect on using AI music models in a “world music” university course to examine conventional and emerging models of musical ecstasy and listening practices (Racy 2003).

Biographies of Participants

Fadi Al-Ghawanmeh is a Ph.D. candidate in music technology and computer science at the University of Oslo (Norway) and University of Lorraine (France), and previously held the position of Lecturer in Music Technology at the University of Jordan. His research focuses on automatic music composition based on natural language processing with control, via motion capture, to allow for individualized customization.

Alexander Refsum Jensenius is Professor of Music Technology at the University of Oslo (Norway), and Director of RITMO Centre for Interdisciplinary Studies in Rhythm, Time, and Motion. He studies how and why people move to music in the fourMs Lab and uses this knowledge to create new music with untraditional instruments. He is widely published, including the books Sound Actions and A NIME Reader.

Nedal Nsairat is Professor of Music Education at the University of Jordan, where he is the Vice Dean for Academic Affairs and Graduate Studies in the Faculty of Arts and Design. His recent research publications address issues regarding distance learning for music education.

Melissa J. Scott is a postdoctoral fellow at Carleton College, and received her Ph.D. in ethnomusicology from the University of California, Berkeley. Her work examines the role of music in displacement and humanitarianism in the Arab world. She is also an oudist and previously collaborated on user experience research focused on AI applications for Arab music.