Iannis Xenakis coined the term “stochastic music.” He used models of mathematics, physics, and statistics to create music that is very different from Western tonal traditions. Many of his scores were written with the help of computers.
In this example for strings, trombones, woodblock, and xylophone, Xenakis uses each instrument as if it is a model of a molecule; the music as a whole represents probability theory applied to the movement of particles of gas. The title Pithoprakta translates to “actions through probability.”
Interactive Music Systems
More recently, Ritwik Banerji, a graduate student at UC Berkeley, has developed a program called Maxine, which listens to its environment and responds improvisationally. Ritwik wants to create software that can be as musical as a person, and he frequently refers to Maxine as his daughter. His ultimate goal is software that is capable of directing itself in a musical and personable manner in an ensemble. Ritwik envisions the process of creating Maxine as parallel to raising a child; in the beginning, parents have to tell their children what to do, but they eventually hope that their children will learn on their own what is wrong and right. You can read more about Ritwik’s insightful ideas regarding Maxine, childrearing, and decolonialism here. You can hear examples of how Maxine sounds here.
Uniform Distributions and The Melody Stochaster
Xenakis’ stochastic music uses complex mathematical and physical phenomena as models for music. For example, Pithoprakta relies on a Gaussian (normal) distribution to model the movement of molecules in a gas. With the Melody Stochaster, I wanted to make music out of a model that is much simpler: the uniform distribution. In a normal distribution, events around one center are more likely than events far from the center. In a uniform distribution, all events are equally probable.
The Melody Stochaster breaks with tradition in complex ways. It rejects Western ideas about music, especially tonal harmony (it’s very difficult to make it output something that would work well in a tonal context, although my stochastic melodies might work in more atonal settings) and tempo (given that the program can randomize over note lengths and rest lengths in any way possible, the only way to have a steady tempo is if all note and rest lengths are integer multiples of a single value; this is difficult in practice). However, it also does not use complex models like the ones in Xenakis’ music. Instead, I opted for the simpler model of a uniform distribution, simply because I could not find any well-known examples of sound made with a uniform distribution in mind. This project is sort of an experiment to see what sound created with uniform distributions sounds like, rather than an attempt to create sound that resembles music or could work in musical contexts. Thus, unlike Ritwik, I am not very interested in making my instrument fit into a pre-existing tradition of human music in an ensemble setting. I also wanted my project to be very transparent (as opposed to fitting into an abstraction such as “this is an artificial human that makes music”), which is why I describe each control with such detail on the User’s Guide and why the project is open source.
However, the Melody Stochaster is not free of all tradition and existing practices. Its use of midi signals restricts it to a 12-note scale with equal temperament. In this way, perhaps it is hard to create any melody that sounds extremely removed from Western music. The reason that I was locked in to using midi in the first place is because the coding library I was using could handle midi signals and not OSC signals, and I had a midi controller but not a controller with OSC capabilities. Other artifacts from bugs or incapabilities of the coding language I was using forced me to use ChucK for sound, which led to additional problems that are detailed in Known Bugs and Errors.
For more information about how the Melody Stochaster works, see the User’s Guide.