Electronic/Computer MUSIC NEWS by Jim Horton

The invention of philosophies and techniques for automated symbol manipulation, using computers, seems to me to be an event in the history of human ingenuity as important as the discovery of fire, the wheel, the plow, the various metals or even the introduction of spoken and written language.

This "visionary" notion derives its support from the embryonic science of Artificial Intelligence, whose goal is to design machines which exhibit mind-like behavior. Its eventual aim is to discover general principles of all mental activity whether specifically human or not. The search can be expected to generate strong implications for how we view ourselves and our civilizations.

Since aesthetic factors are crucial to understanding the mind, it can be expected that music will make an important contribution. We can watch the development of this line of thought in the Computer Music Journal edited by Curtis Roads and published by the People's Computer Company in Menlo Park. Issue 9 contains several relevant articles.

In "Grammars as Representations for Music," Curt Roads gives a valuable review of modern thinking about whether music can be thought of as a language. Sophisticated software has developed around formalized grammars and these tools can be applied where music is analyzed or perceived as segmented and the techniques of editing can be used in composing. His conclusion is worth quoting at length:

"Grammars may lead beyond unified composing and analysis models and toward intelligent musical devices. An intelligent musical device will be able to convert the iconic musical signal into symbolic form, and be able to recognize, for example, not only frequency, amplitude, and duration (as analog devices do today) but also larger syntactic forms such as phrases and other macrostructures as well as extra-syntactical aspects of the music. Acting from a base of programmed or even acquired grammatical knowledge, such a device will be able to listen and respond intelligently not just to sound, but to music."

"A Microcomputer-controlled Synthesis System for Live Performance" by the well-known Canadian composer Martin Bartlett describes, in a wonderful manner, how he connected a Kim-I microcomputer to his handmade analog synthesizer, the "black box." Martin composes for the synthesizer by designing patches and playing the machine by small adjustments to its parameters using potentiometers and switches. Because he doesn't use a keyboard, it has always been difficult to generate stable complex sequences. "The computer offered the possibility of transcending this problem." Martin has built an eight channel interface that outputs control voltages from the computer to four oscillators and pulses to envelope generators. It also converts voltages coming from control modules into digital words.

The programs are organized around tables containing data that represents frequency and duration. Duration values go to a timer which, when it counts down to zero, indicates that an updated frequency value is to be output to an oscillator. The choice among tables and table scanning algorithms can be made by inputs coming from the synthesizer. The tables are read either sequentially or by using various "random walk" subroutines. Probability distributions are generated by variably repeating data values in the tables.

Of special interest to designers of musically intelligent instruments is the implementation of a program that generates numbers coning to a 1/F statistical distribution. This type of randomness was discovered to be integral to music by the physicist Richard Voss, who plugged a radio into an otherwise underutilized computer and had it make a long term analysis of Bay Area rock, classical, jazz and talk shows. 1/F randomness is in between overdetermined Brownian motion and truly random white noise and has been conjectured to be a signature of much activity that is relevant to intelligent behavior. Martin tells us that "...the results are musically interesting, having that peculiar combination of volition and purposelessness which seems so characteristic of computers when they are feeling playful."

Issue 9 also contains two other articles pertaining to Artificial Intelligence in Music. Thomas Blum reviews Herbert Brun's Project Sawdust, an automated system of composition, and Paul Berg's "PILE--a Language for Sound Synthesis" describes a radically new way "to hear that which without the computer could not be heard." p.9