Rule-based Algorithmic Music in an Age of Generative AI
Andrew R. Brown – Griffith University
Generative Artificial Intelligence (AI) techniques are becoming increasingly pervasive, encompassing creative AI systems that generate media such as visual art, moving images, product designs, and music. For decades prior, hand-crafted rule-based generative systems were successfully used in creative digital media work. In music, rule-based generative systems supported composition (Taube, 2004) and improvisation (Dean, 2003). Examples range from the sophisticated—Experiments in Music Intelligence (Cope, 1992)—to the elegantly simple—The Continuator (Pachet, 2002)—to the highly interactive and co-creative—Lexicon Sonate (Essl, 2016).
Human crafting of algorithmic processes still holds significant value, even as machine learning models make algorithmic generative processes more accessible. The value of rule-based work lies in its ability to reflect the artist’s unique perspective and creative choices. Each algorithm is a carefully constructed set of rules that guide the generative process, resulting in a structured and dynamic work. Techniques developed for these generative practices help articulate knowledge about musical practices (Sorensen and Brown, 2007; Dahlstedt, 2009). In hand-coded algorithms like those used in my works, the authorial voice seems closer to the surface. This proximity allows for a more direct and personal expression of the artist’s vision, as the rules and structures are meticulously designed and implemented by the artists themselves.
Tightly defined hand-crafted rule development contrasts with machine learning models, which often rely on vast amounts of data and complex neural networks to generate content. While machine learning can produce impressive and innovative results, the handcrafted nature of rule-based algorithms offers a different kind of creative control and intimacy. Nevertheless, rule-based and learning-based methods both involve crafting code to express creative intentions and are likely to be complementary approaches in modern generative musical systems.
References:
Cope, David. 1992. “Computer Modelling of Musical Intelligence in EMI.” Computer Music Journal 16 (2): 69–83.
Dean, Roger. 2003. Hyperimprovisation: Computer-Interactive Sound Improvisation. Middleton: A-R Editions.
Dahlstedt, Palle. 2009. “Thoughts on Creative Evolution: A Meta-Generative Approach to Composition.” Contemporary Music Review 28 (1): 43–55.
Essl, Karlheinz. 2016. Lexikon-Sonate (1992–2010). New Interfaces for Music Expression, Brisbane, Australia. https://vimeo.com/176422741
Pachet, François. 2002. “The Continuator: Musical Interaction with Style.” In International Computer Music Conference, 211–18. Göteborg, Sweden: ICMA.
Sorensen, Andrew, and Andrew R. Brown. 2007. “Aa-Cell in Practice: An Approach to Musical Live Coding.” In Proceedings of the International Computer Music Conference, 292–99. Copenhagen: ICMA.
Taube, Heinrich. 2004. Notes from the Metalevel: Introduction to Algorithmic Music Composition. London: Taylor & Francis.