Letter to the Editor, Computer Music Journal, Vol. 11, #3, Fall, 1987
A followup to material recently published by CMJ

A Short History of Intelligent Instruments

Regarding the historical public availability of intelligent instruments:

It was erroneously stated in the Winter 1986-7 issue of CMJ that the release of 2 programs ("M" and "The Jam Factory" for Apple Macintosh computers by Intelligent Music, Inc.) was "historically important because it marks the first time that intelligent musical instruments are available to the general public." For the record, as well as for general information for your readers, IM's programs were not the first available. There are several historical precedents to the IM programs.

Though the Intelligent Music programs have only just become available, they were preceded by a variety of intelligent musical performance products which, though quite different in character instance to instance, all constitute "intelligent instruments". In inverse chronological order, most recently preceding IM's products to earliest before, these include:

"Instant Music" by Robert Campbell, published for the Amiga computer by Electronic Arts, Inc., in summer or fall 1986. This program acts much like a stencil superimposed on mouse movement in one dimension, producing pitches in real time which are subsets of predetermined chord progressions or scales provided with the program as components of supplied pieces.

"Music Mouse - An Intelligent Instrument" by Laurie Spiegel, now available from OpCode Systems for the Macintosh and Amiga, first publically made available for sale in Jan. 1986 by Knowledge Engineering, Inc. It derives 4 voice harmony from 2-dimensional mouse movement in real time while reading 44 realtime keyboard controls which affect harmony, voicing, orchestration, etc. It also adapts the harmony type, transposition, scale degree, chord quality, and melodic inversion of predefined 4-voice patterns in realtime to match mouse-selected pitches as they are played.

Prior to that, a number of music products were publically available which functioned as intelligent instruments in various ways but did not explicitly represent, describe, or think of themselves as such. The distinction between tools for "algorithmic composition" (or for interactive composition in general) and "intelligent instruments" is somewhat blurry, having to do with the creator's personal sense of role, with the balance between realtime interactivity and predetermination, and with the nature, complexity, and inherent interest of the relationship between user input and musical output. These precedents include:

The "Sequencer" by Dave Oppenheim of OpCode Systems, publically available since July, 1985, which permits realtime interactive change of the transformations to which recorded materials are subjected while replayed. Several types of transformation, involving transposition, rhythmic and durational variation, and melodic permutation are changable on-the-fly in realtime. The Sequencer also allows generation of new musical sequences from others through various realtime-changable transformations.

The algorithmic and compositional programs available from 1984 onwards as "Dr. T's" music software, written by Emile Tobenfeld, Jack Deckhard, and Jim Johnson, for Commodore and other small computers. While intended for composition rather than performance, these are still relevant precedents. Amongst the "Dr. T." group of programs occur the abilities to change interactively in realtime such parameters as transposition, time delay, orchestration, and the random seeds used in realtime algorithmic music generation.

The "Algorithmic Music Language" by Ray Jurgens, published in 1981 by Electronic Arts Research (not to be confused with Electronic Arts, Inc. above) for 8080 and Z80 S-100 computer systems. Though intended primarily as a medium of algorithmic definition for non-realtime (compiled) compositional use, it was modified and updated in 1982 to accept and use keyboard console input interactively during realtime play. This software has been used for realtime-modifiable algorithmic generation in live performance by Jeff Rona and other members of its user community.

Though not in the low-cost personal computer price range of the above, the McLeyvier (publically available 1981-2) functioned as an intelligent instrument via its interpretive macro language ("McLeyvier Command Language" by David McLey et al). Individual macros were triggered by specific user actions, allowing, for example, complete reorchestration during live performance in different ways depending on the specific pitches played, selective retrieval in realtime of materials from disk for playback as accompaniment depending on keyboard performance content, random pitch generation, and macro sets for music education which changed the instrument's configuration and response depending on individual students' abilities to match pitches or accomplish other musical tasks.

Another genuine precedent to IM's programs in the realm of low cost intelligent instruments available to the public was Marvin Minsky's "Muse" (available from Triadex, Inc. for $300, I don't remember exactly when but sometime in the early 1970's), a self-contained "computer" instrument for melodic generation and thematic developement.

An additional very important category of precedent which is often overlooked because of its non-digital logic was the modular analog synthesizer (Don Buchla, publically available circa 1964, with its random generators, control voltage processors, etc., Robert Moog's modular analog systems of the same period, and other subsequent instruments which provided comparators, sample and hold, and other functions which could be configured to provide complex musically interesting transformations of player input). Though it's often assumed that "intelligent systems" must be digital, the analog computer, which dealt with relationships among informational components "by analogy" (as analogues of each other) has much in its logic and design which should not be overlooked or forgotten. The complex realtime interactive generative and transformational processes which these instruments permitted place them as the very first (to the best of my knowledge), and in some ways still the most powerful, "intelligent instruments" ever to be made publically available.

Through quite a different type of availability, other precendents came from individuals who chose to make our interactive musical software publically available in the early days of personal computers by publishing source listings and algorithms in such places as Creative Computing (1977-83?), the Proceedings of the Symposium on Small Computers in the Arts (1978 onwards), and via other consumer-level computer magazines and telecom networks. Throughout the decade since the advent of home computers, many individuals completely obscure, long forgotten, and/or otherwise unbeknownst to the majority of CMJ readership developed individual generative and/or interactive music programs and made them publically avaiable as text. (I won't bother to bibliographize but I could.)

There may be other precedents to Intelligent Music's products which I don't know of personally or which have not sprung to mind during the writing of this letter (and to whose creators I therefore apologize), but the above list is enough to show that it is incorrect and misleading to CMJ readers to write that IM's programs are "historically important" because they are the first publically available intelligent instruments.

The ways in which IM's programs are unprecedented are the ways in which they are unique among intelligent instruments by virtue of their individual natures and specific aesthetic and structural properties, not by their being intelligent instruments per se or by their availability. They belong to extended standing traditions not only of exploration and invention but of efforts to make the results of such work comprehensible and available to others (which, as I've discovered by distributing my own software, may involve comparable effort in itself).

I am reluctant to end this letter without mentioning that the Ultimate Intelligent Instrument, the sound-producing general purpose computer itself, has been available to all at reasonable cost for several years now. Anyone reading this who is interested should be encouraged to learn to program and to extend this new area of musical creativity in their own individual ways, whether by developing intelligent instruments to be made available for sale to others, or by doing so just for use in their own music, as have such software-author-composers as David Behrman, George Lewis, Ron Kuivila, Paul DeMarinas, and others who have never issued products and for whom instrumental intelligence has always been just a musical means, not an end in itself.

Since this letter has turned into more of a history than expected, I may as well mention 2 additional historical referents, though neither was ever a publicaly available product:

My own first memory of the term "intelligent instrument" dates from about 1973, when Max Mathews, Emmanual Ghent, Dick Moore and I used it in discussing our work. Well before realtime digital synthesis had become established (from 1977 onwards), using hybrid (computer controlled analog) technology, we were among the very first to be able to hear the results of high level music-generating software in realtime and therefore to be able to interact with software processes while they were actually computing the music we heard, instead of listening to the stored results later. Because of the realtime nature of the system we used, this Bell Labs group may well have been the first origin of the concept, the practise of software based "intelligent instruments" and also of the expression. (Note: Realtime audio is necessary for "intelligent instrument" creation, and prior to the development of realtime digital synthesis, apart from hybrid systems, computer music was pretty much limited to non-realtime synthesis and pre-compiled non-interactive algorithmic composition. I mention this because of the large number of new readers new to the field who may not realize how rapidly we have come to take completely for granted technology which a mere decade ago was still only hypothetical.)

The very first instance of any automated intelligent instrument / algorithmic composer may have been Leonardo da Vinci's canon generator. After (or while) working on modifications to the Hurdy Gurdy, Leonardo worked out "a cylinder with pegs that can perform a canon simply by being turned ... the apparatus with frets for dividing the string [or] to open in order, at the will of the person setting the pegs, the stops of organ pipes ... to perform a canon with variable time intervals." (Codex Arundel folio 137v and Windsor drawing #12697 as cited in Leonardo da Vinci, Reynal & Co., NY, n.d., pp. 229-230). (To get an idea, picture a musicbox-like drum with easily changable pins to twang notes, and 2 or more reeds at adjustable angles to each other for time delayed repetition.) Though computer-based instruments are a very recent thing, the desire to explore the kinds of things they can do, to amplify the power of performing and composing musicians through technology, is much much older.

End of Brief History Tutorial on Intelligent Instruments. Thanks. Sincerely,

- Laurie Spiegel



Copyright ©1987 Laurie Spiegel. All rights reserved.
back to Laurie Spiegel's writings page | go to Laurie Spiegel's home page