AI's Struggle to Capture the Rich Tapestry of Gulf Music Highlights a Digital Divide

When I first arrived in Abu Dhabi 13 years ago, I had a vision. As a composer, my prior work in Portugal involved developing computer programs to generate music, including jazz. I was keen to explore if similar technology could be employed to create music inspired by the traditions of the Gulf. However, a significant obstacle emerged: I couldn't locate the necessary recordings. Despite numerous descriptions of the region's rich and diverse musical landscape, much of it was surprisingly difficult to find. This led me to begin recording local musicians myself. The subsequent year, I traveled to Kuwait to document traditional pearl-diving music. What began as an experiment in computer-generated music evolved into a more extensive endeavor to document, comprehend, and enhance the accessibility of these traditions. This experience illuminated a problem that has grown increasingly pertinent as artificial intelligence reshapes our engagement with music. While we now have access to an unprecedented volume of recordings, the algorithms designed to help us navigate them exhibit a marked preference for certain musical traditions over others. Paradoxically, the very digitization of music may be exacerbating the disparity between what gains recognition and what remains largely overlooked.

Music recommendation systems excel due to their ability to leverage vast datasets. They can readily differentiate between artists like Taylor Swift and Ariana Grande, identify sonic similarities between tracks, and suggest songs aligned with individual preferences. Yet, when tasked with recognizing 'Al Ayyala,' a traditional Emirati performing art encompassing music, poetry, and movement, these same systems encounter considerable difficulty. The scarcity of comparable recordings and detailed musical information available for training renders these traditions harder to discover, even when recordings are accessible online. Addressing this imbalance is a key challenge for Music Information Retrieval (MIR), a field that merges musical knowledge, engineering, computer science, and AI to digitally analyze and organize music. Much of the existing technology has been developed around musical traditions with extensive recording histories and established theoretical frameworks. However, Arabic music, for instance, employs 'maqam,' a melodic system that can include intervals beyond Western equal-tempered tuning. Similarly, Indian classical music possesses its own intricate melodic and rhythmic structures. Algorithms primarily trained on Western music may falter in recognizing these distinct characteristics.

At NYU Abu Dhabi, we are tackling this issue by integrating computational methods with the expertise of musicians and researchers, rather than relying solely on computers to learn from recordings. In our research into Carnatic music from southern India, for example, we are developing techniques to distinguish the 13 different types of strokes played on a 'mridangam,' a double-headed drum. While current systems can easily identify instruments in a Western drum kit, recognizing the nuanced playing techniques of a mridangam remains a significant research challenge. The potential applications extend beyond mere instrument or style identification. Centuries of trade and migration across the Indian Ocean have fostered connections between the Gulf's musical traditions and those of East Africa and southern India. In my teaching, I occasionally play students recordings of music from Ghana alongside Gulf pearl-diving songs and Brazilian samba. Despite their diverse origins, their rhythmic characteristics, including hand-clapping and percussion patterns, can exhibit striking similarities. Computational analysis offers a powerful tool to investigate such relationships across extensive recording collections, revealing patterns that would be arduous to discern through listening alone.

We have already applied this methodology in research on 'Candomble,' an Afro-Brazilian religious tradition, examining the connections between its rhythmic patterns and those found in West African music. Similar approaches could deepen our understanding of the Gulf's musical interconnections with other regions. However, interpreting these findings necessitates the collaboration of musicians, historians, and ethnomusicologists alongside computer scientists. In 2016, I co-founded the Music and Sound Cultures (MaSC) research group at NYU Abu Dhabi with ethnomusicologist Andy Eisenberg. Our work encompasses recording traditional Emirati music, researching Carnatic percussion, and developing creative projects centered on traditional crafts. One of our primary goals is to make musical connections more apparent. Imagine a streaming platform introducing you to traditional recordings from Kuwait or Bahrain while you listen to a contemporary musician incorporating Gulf pearl-diving rhythms. For this to happen, the technology must first learn to recognize these intricate relationships. This month, NYU Abu Dhabi is hosting the International Society for Music Information Retrieval Conference for the first time in the Arab world, a significant achievement resulting from a collaborative effort involving researchers from our university, Mohamed bin Zayed University of Artificial Intelligence, and Zayed University. We aspire to foster greater international interest in the region's musical traditions and expand the research community dedicated to their study. Among the presentations is the inaugural Arabic music dataset, created by NYU doctoral student Peter Traver, MaSC members Eisenberg and Robert Rowe, in collaboration with virtuoso violinist Layth Siddiq. This dataset uniquely combines expert musical knowledge with computational methodologies.

As a composer, my interest transcends mere analysis and archiving of music. We are also developing tools that empower contemporary artists to experiment with regional rhythms, melodic systems, and performance practices. Musicians like Ghazi Al-Mulaifi, through his work with Boom.Diwan, exemplify how traditional Gulf music continues to inspire modern artistic expression. This reflects a fundamental aspect of cultural heritage: traditions endure through continued practice, interpretation, and adaptation. While recording music for future generations is crucial, especially where documentation is scarce, we must also cultivate opportunities for contemporary engagement with this heritage. I often liken our approach to having an 'oven' rather than a 'freezer.' Our aim is not only to preserve these traditions but also to inspire their creative evolution. As AI's capacity for generating new music grows, its potential to illuminate and help us understand existing musical forms deserves equal consideration. The musical traditions of the Gulf encapsulate centuries of history and cultural exchange. Enhancing their accessibility will empower future generations to discover this rich history, draw inspiration from it, and ensure its continued vitality.