We describe the conceptual background and implementation of various biologically inspired computational models for creative decision making in the realm of computer music.This includes the use of evolution as an alternative to explicit design as well as using emergent functionality in simulated societies of interacting software agents.We aim to explore complex behavior in interactive systems using self-organisation: global, overall complexity is a side effect of the application of simple local rules in a distributed system.
The current paper documents three specific approaches. First, cellular automata for which the lookup table rules are seen as genotypes and manipulated using genetic operators.This provides for a huge genetic space to be explored although the structure of the rule does not itself evolve. A second method deals with this limitation and views the rewrite rules of Lindenmayer Systems as geno¬types.These rules are represented as nested Lisp structures of arbitrary complexity. The user selects interesting rules by interactive inspection of their resulting phenotypes i.e. their realization to the midi domain. Evolution is guided by applying cross-over and mutation operators to selected rules and the cycle repeats. This method is an example of true exploration for it allows for growing complex artefacts without any initial formal specification. In addition, goals are mobile and dynamic; they are identified while engaged in the act of searching itself.
Finally, a third class of programs accommodate complex social behavior in a collective of agents equipped with sensors and effectors. Agents move in a two-dimensional space. Sensors capture nearby activity from fellow agents as well as external midi input from a human wetware agent. Effectors control sensitivities, how the agent moves and how musical responses are created. Agents express a personal character yet they also wish to integrate external input from a human performer.A behavioural wealth issues from the conflicting forces of integration and expression. These programs provide real-time audio-visual feedback and are implemented in HMSL.
In summary, this paper offers examples of systems where man and machine mutually contribute to a climate which favours invention and collaboration. The results would not be obtainable by either man or machine when acting in isolation.
- Peter Beyls, Belgium, has been exploring computer programming as a medium for artistic expression since the early Seventies. His approach views computers as cognitive partners in the process of artistic creation and borrows methods from the science of artificial intelligence. His educational background includes music studies at the Royal Music Conservatory, Brussels and computer graphics at University College London. Active as a composer/performer and visual artist. Beyls has published extensively and lectured at various institutions in the US and Japan. He currently teaches computer graphics and aesthetics of the digital medium at the St Lukas College in Brussels. He is a member of the ISEA association (Inter-Society for the Electronic Arts) Board of directors.