Traditional channels for introducing intelligence in computer music systems are firmly rooted in the knowledge-based approach; methods and computational strategies borrowed from the field of artificial intelligence. Expert systems for composition and pattern-directed inference systems for real-time man-machine improvisation are exemplary. In general, the aim is to introduce independent creative decision making through computer simulation of human creativity. Impressive statements have been produced along these lines, in music as well as rule-based computer graphics. Two observations have led to the consideration of a totally different method.
First, expert systems become problematic if situations occur that were not anticipated by the programmer and sooner or later, the programmer is faced with a complexity barrier. Second, appreciation of the pattern making potential of nature led to the study of concepts like self-organization.
Complex dynamical systems are an alternative to the constructivist approach in composition, i.e. the critical assembly of architectures of time according to some explicit scenario. Complex dynamical systems, on the other hand, consist of many elements interacting according to very simple laws but giving rise to surprisingly complex overall behaviour: Composition becomes experimenting with attractors — instead of creating a rule-base — as well as designing tools that allow the topology of the composer to interact with the system’s internal activity. The idea is to critically push the system out of equilibrium using tactile motor control as to explore the various degrees of freedom of a given system. The implicit behaviour is then mapped to the musical problem domain. Improvisation becomes navigation in a hypothetical world of which the composer is both inventor and explorer: Strange and intricate imagery, in both space and time, is found in physics, biochemistry, fluid dynamics, ecology and nonlinear mathematics. We have implemented and evaluated various models for spontaneous pattern formation, including one-dimensional cellular automata, direct computer simulation of chemical instabilities as witnessed in the BZ reaction and a spatial model exploring equilibrium behaviour in a society of interacting agents moving in 2D space.
The present paper outlines a connectionist-like model, a regular structure of agents engaged in local interaction, using forces of activation and inhibition between neighboring agents. Randomness/determinism and chance/necessity seem at the heart of creativity and happen to be central to the music of our time. We propose to view emergent properties from initial random configurations as a subtle alternative for both constraint-based, reductionist handling of randomness as well as rule-based composition by way of some generative grammar; complex dynamics as a creative, generative principle and a channel toward higher levels of man-machine interaction.
- Peter Beyls (Belgium)