Generative Artificial Intelligence (AI) is rapidly becoming ubiquitous, with AI systems now capable of producing creative works in fields ranging from visual art and moving image to product design and music. While machine learning techniques dominate much of the generative landscape today, hand-crafted, rule-based generative systems have played a central role in creative digital media for decades. In music, rule-based approaches have supported composition and improvisation with prominent examples arising from the likes of David Cope, Alexandra Cardenas, Francois Pachet, Shelly Knotts, Karlheinz Essl, Andrew Sorensen and Brian Eno. In visual media, similar principles can be seen in the work of artists like Jared Tarbell, Lauren McCarthy, Casey Reas, Olivia Jack, Zach Lieberman, and Joshua Davis, who are renowned for their contributions to creative coding. The popularity of platforms like Processing and p5.js exemplify how rule-based generative systems can be employed to produce unique artistic experiences. Audiovisual outputs from numerous creative coders demonstrate the breadth of this approach, from sophisticated systems to highly interactive, co-creative works.
In the current era of generative media, where machine learning processes dominate, the value of rule-based approaches may seem diminished. However, I argue that human-crafted algorithms retain significant value, even as machine learning models make algorithmic processes more accessible. The key distinction lies in the personal connection between the artist and the algorithm. Rule-based systems are carefully constructed by the artist, reflecting their creative vision through a set of designed algorithms that shape the generative process. This direct involvement can enable intimate and personal expression, something that machine learning systems, which often operate on vast datasets, struggle to replicate with the same level of creative control.
While machine learning produces impressive and innovative results, it can lack the nuanced human touch and authorial presence that rule-based methods afford. The handcrafted nature of rule-based systems allows for a closer alignment between the artist’s intentions and the generated output, ensuring a stronger connection between the creator and the work. Rule-based systems also offer dynamic and intuitive adjustment of parameters in real time, and often have minimal cost of execution compared to running AI models.
However, the distinction between rule-based and machine learning approaches is not absolute but complementary. Both involve coding to express creative intentions, much like how traditional lithographic printing coexists with modern digital printing techniques. Similarly, I believe that hand-coded generative systems will continue to play an essential role in digital artistic practices alongside machine learning approaches, especially in creating work that is not only technically impressive but also culturally relevant and emotionally resonant. In this way, co-creative processes that combine human intuition, cultural knowledge, and emotional insight with automated systems of pattern recognition and generation will remain crucial for the future of generative media.
This post is based on a couple of academic publications:
- Brown, Andrew R. 2024. “Rule-Based Algorithmic Music in an Age of Generative AI.” In Proceedings of the Australiasian Computer Music Conference. Sydney / Melbourne: ACMA.
- Brown, Andrew R., Nick Coleman, and Ben Silver. 2024. “Endless Moments: Generative Music in the Browser.” In Proceedings of the 4th International Conference on Digital Creation in Arts, Media and Technology. Macao, China.