Article published in the VIRTUALIA supplement, of the journal La Jornada. December 2, 1997.
The need to create new methods of composition, as well as the formalization of these methods, opens a new field of research within music theory, a rather broad field where the tool, and sometimes the main instrument, is the computer.
This need, mixed with scientific progress in the fields of genetics, chaos theory, artificial intelligence, dynamical systems, and a great development in computer simulations, has led musicians of recent generations to take a look at the various scientific disciplines. From that moment on, the musical perspective changed radically, uncovering a new and unexplored world, a world in which every idea can be developed to the maximum, and its possibilities can be unsuspected.
The first attempts to make music by computer were made by musicians who took the idea of random music on a dodecaphonic scale, implementing a program that would choose the next note of a melody at random, thus obtaining a chain of notes that were then interpreted by musicians, obtaining rather poor results.
The subsequent goal was to find a method that could express a certain coherence within these chains of notes, a kind of rule that would allow the computer to generate better melodies.
In the 1950s, the Greek musician and architect Iannis Xenakis began to use Markovian processes (Markov chains) for his compositions. The idea of using Markov chains is very simple: on an already written work, it is calculated how many times a note is played during the whole work and how many times, being on that note, the melody travels to each of the other notes. With this data a probability table is calculated, which is used by a random function to decide which is the next note to be played. At the end of this process we obtain a melody characterized by the original work.
We should note that the two previous processes are based exclusively on probability and chance. However, the deterministic current also has its own proposal, based mainly on the theory of dynamical systems. In general terms, a dynamical system is one that has an initial state and a rule of change, and given these two characteristics, it is possible to know the state of the system at any time. Some of the most common applications of dynamical systems in composition are deterministic automata.
An automaton is a system consisting of a set of states, a set of tansitions or actions and a set of characters to be recognized by it. The definition of an automaton is somewhat abstract, but we could try to give a simple example of how it works.
Suppose we have a spotlight and we have an optical sensor connected to a circuit that controls the spotlight; let’s suppose now that the following actions occur: if the sensor sees a light, the spotlight is turned off, and if the sensor sees no light, the spotlight is turned on. The model that regulates this system is called automaton.
In this example the states are: light bulb on and light bulb off; the transitions or actions are: turn off the light bulb and turn on the light bulb; and the set of characters it recognizes is light and dark. In fact, the automaton illustrated here is a deterministic automaton, that is, for each character it reads, it takes one and only one action.
The applications used in musical composition have the same mechanics; we have a set of states and a set of rules or actions; the set of characters recognized by each automaton built for musical composition is left to the free choice of the programmer.
This system has the disadvantage that the “tunes” obtained, given an automaton, depend entirely on the character strings read; if the character string is the same, every time the program is executed we will obtain the same results.
To solve this problem, some researchers proposed hybrid systems that contemplate both the probabilistic and random part and the deterministic part, using precisely the two previous methods (Markov chains and deterministic automata). At the Music Informatics Laboratory of CIMAT, in Guanajuato, some music composition programs have been developed using these methods.
Other examples of music generated with dynamic systems are those that take chaos theory as a tool, as well as some properties of fractal geometry to generate what has been called “fractal music”.
Artificial intelligence uses neural networks and genetic algorithms to try to reproduce animal behavior, based on brain analysis.
Thus, we may have intelligent machines with the ability to compose and create musical works as good as our own. Several questions arise here: Are we willing to accept such music as art? To what extent can we conceive of these new technologies as creations of our own brains? We can refine more and more the details with which a computer piece is generated, getting as close as possible to our concepts of music, aesthetics and art, but &is this really the end we are pursuing by making computer music?
Perhaps this problem is not for this branch of musical research to solve, and despite the criticisms and bad comments, it will continue its work in this field.