Written for the Computer Music Journal, Summer 1996.

As technologists and/or musicians smitten with technology we find computers fascinating,  but the fact is that most of what computers do is essentially boring. Their utility for bookkeeping tasks has made the biggest impact: The ability to keep track of large amounts of information, to handle repeated similar tasks which have minor variations, to quickly access previously assembled data, to control machinery with spatial and temporal precision. 

It's largely been no different in music. Most of the current musical uses of computers are of a bookkeeping nature. Sequencers and digital audio workstations don't do anything really new: they just make things that were difficult before somewhat easier.(And in art, this is not necessarily a virtue: when it took weeks or months to assemble a four minute musique concrete composition, the composer made quite certain that the quality of the material justified the labor, and the quality of attention poured into the works was remarkable. One side effect of making the job easier is affording the ability to create works without sufficient attention.)

In the twentieth century, among many musicians who identify themselves as composers, music has been largely redefined away from its social definition as an activity of players and listeners, and towards an emphasis on compositions as sound objects. These sound objects have become ever more complicated, ever more data-intensive in their specification, from Schoenberg's  serial tone row, to the more extensive parameterization of Messiaen and Boulez, through to the complete sample level specification of computer based works.

Indeed, modern life in general is increasingly complicated; the computer is of course both the child of this way of thinking, and the solution to the problems that this unceasing complication give rise to. As we develop tools to handle more and more information, we also develop ways of working which demand that more and more information be handled.  Here the computer's nature as a bookkeeping tool gives us greater control over materials, but really stands apart from any direct participation in the creative aspect of art-making.

The work I and others in the experimental tradition have been doing grows out of a different conception of computer technology, that in a sense sidesteps the problems of control specification.  One can conceive of a computer system as a framework for embodying systems offering complexity and surprise, rather than as a tool which performs the bookkeeping tasks associated with recording and organizing one's compositional materials.  This work inherits from the somewhat contradictory but complementary insights of Cage and Xenakis, both relying on mechanisms open to chance operations to lighten the burden of specifying details of complex musical structures. Under this paradigm, composition is the design of a complex, even wild, system whose behavior leaves a trace: this trace is the music. This work reflects an intent quite different than that inspiring the crafting of highly polished sound objects. 

Both of these ways of working --what I would call the Crafting School and the Wild System School-- have inherent failings and difficulties. The Crafters, laboring under the burden of specifying so much, often produce works suffering from a certain lifelessness; The Wild System types are often hard pressed to come up with systems whose behavior provides a close enough match to human perceptual abilities for listeners to perceive any pattern at all. These two ways of working represent two responses to complexity, two approaches to the problem of design. Neither approach is sufficient to work in the increasingly complex world engendered by our computer systems, a network of interacting human artifacts which is approaching biological levels of complexity.

What I see happening now and in the immediate future is the closing of the gap between these two ways of working, brought about by increased understanding of the nature of complex systems, and the increasing sophistication of machine learning and modeling algorithms such as genetic algorithms, neural networks, Lindenmayer systems and other methods based on strong analyses or analogies with biological systems.  If the Crafted Object way of working has meant Getting What You Want (and specifying that in great detail); and the experimental way of working has meant Wanting What You Get ( being interested in whatever arises from ill-understood systems of interaction), the emergent ways of working based on evolutionary paradigms means Getting What You Didn't Know You Wanted. 

The inherent problems of both ways of working are addressed in this convergence. Genetic algorithms, for example, provide a way to deal with the problem of having too much data to specify in a much more sophisticated way than dealing with matters statistically. They represent a family of machine learning algorithms for exploring huge problem spaces, for finding a path through an astronomical profusion of possibilities. More importantly, in these systems computers are actually creative, generating solutions or material  which is shaped by constraints imposed by the human engineer / designer / composer. Composers can work with systems which incorporate their judgements into a machine based generative process. 

This move to a new relationship between human and machine, in which the machine takes a more active or creative role in specifying structure, is an unavoidable step on our road to dealing with ever higher levels of complexity. In these systems human-machine interaction  really becomes something more akin to social interaction than the old relation between human and tool.   I think that in the future evolutionary design systems of this nature will become increasingly important in all areas of human design. As always, music is prophetic.