This post is about computer music composition. I touched on one form of this in an earlier blog where I described using algorithmic techniques to aid sound design and composition, but I want to take a longer look at different kinds of computer music composition in this post.
First, let’s discuss some forms of “computer music.” From the 1950s when Max Matthews programmed a large mainframe to perform “Bicycle Built For Two,” there have been various ways computers have been made to play music. The most common these days, of course, are digital keyboard synthesizers which are really small computers dedicated to making sounds when keys are pressed. They all have a central processing unit, memory, a stored program, etc. just as any other digital computer. There are various ways computers can be used to synthesize musical sounds – additive synthesis, subtractive synthesis, wavetable synthesis, physical modeling, FM synthesis, etc. – but those are not the topic of this post. This is not about computers playing music, either pre-programmed or on demand in real time by a performer pressing keys on a keyboard.
This is also not about computers that record music, single or multiple tracks, sometimes processing those recorded sounds, then playing the recording back either from the computer or via an audio CD. And we must remember, too, that a CD player is also a tiny computer with a processor, memory, etc. But in this form the computer is used simply to store and modify sound.
Again, this post is about computers composing music. In my discussion of algorithmic music composition I was really discussing computer assisted music composition. The composer sets up a process on the computer which then generates music based on the nature of the process. This kind of music is also called “process music” or “generative music.” The amount of direct control from the composer can vary according to the kind of processes used. In my own work I consider the algorithmic program an “intelligent musical instrument.” As the composer I am “performing” on this instrument but the computer is also aiding in the creative process so certain sounds and sequences can be created that I did not expect. Thus I can surprise myself with my own “performance” of this computer assisted composition. But I still consider myself the “author” (composer) of this music. The computer is just a tool of my own creative skills or lack thereof. J
But what about letting the computer do all of the composition, using artificial Intelligence (AI) techniques. There have been experiments with this over many years and, as computers get faster and more powerful, the results can be impressive. In early attempts computers were simply given some basic “rules” of musical composition within some genre of music and then the programmer waited to see what the machine came up with. In very early versions this was not even a real time process – it took the computer a long time to “compose” — and some early computers did not even play the actual musical sounds, but rather printed out coded text files that were hand translated into standard musical notation for human players to play.
Modern computers can do work that is very impressive. I recently heard a report on the BBC from a music festival of computer composed music. In the example they played a computer had been fed all kinds of jazz music written by one of the jazz greats (I am thinking Charlie Parker, but my memory might be wrong in this.) It then created a new piece, sound and all, that sounded like Parker (or whoever) but was actually a new work, not just a collage of the artist’s earlier work. Is this truly computer composition? Advocates will say “yes.” Of course the same processes could be applied to nearly any composer. Feed the computer lots of Mozart and a “new” composition by “Mozart” might be produced.
But I still would call this only computer assisted composition, not true original composition. A human composer still created the template. The computer, at least at this stage, seems not to produce discontinuities. When Arnold Schoenberg created 12-tone music it was something unexpected (and often un-liked) – a sudden departure from the 18th and 19th century music that preceded it. It was a discontinuity. Feeding a computer many 19th century operas would not likely produce a result similar to Philip Glass’s “Einstein on the Beach.” In terms of opera Glass’s work represents a discontinuity. If one wanted a computer to write in the style of Glass, one would need to feed it other works of Glass. In other words, the discontinuity – original idea – needed to be produced by a human composer. Likewise, teaching a computer the entire 20th century history of pop rock music would not result in a hip-hop composition. That discontinuity was also created by humans even though computers are often used to provide some of the beats associated with that genre.
I am impressed by the power a computer can give a composer who is searching for his/her own original composition, but I would still consider most of what computers have “composed” to be more computer assisted composition than real original work. With very powerful computers coming into use, such as IBM’s “Watson.” which was programmed to play Jeopardy on television successfully, this could potentially change. Machine learning is in its infancy. Philip Glass attended the Juilliard School and had a very traditional Western music education. When he did some transcribing work for Ravi Shankar he realized that Indian music had a different aesthetic. He took his Western training and combined it with ideas he had about Indian music to create his own unique style that has made his career. Someday soon a supercomputer like Watson might be able to invent a new genre of music without consciously copying the “style” of an existing composer or genre. That would be exciting though I am not sure we are quite there yet.
But things are evolving rapidly. If someone reading this thinks they have seen, or are working on, a truly innovative program that creates whole new musical genres, then I invite you to reply to this post or even write a guest post for The Audio Penguin.