The basic idea behind Goldberg is to provide a computer composition assistant. Goldberg will use various kinds of musical operations, based on simple musical ideas and rules of thumb, to assist a composer at work. The basic metaphor is the idea of variations on a theme (as in, for example, Bach's Goldberg Variations). At each stage of a session with Goldberg, there will be one or several current themes, and Goldberg will generate variations on those themes by applying operations such as transposition, reversing, inversion, randomization, or any Rube-Goldbergian sequence of manipulations.
The basic Goldberg idea was inspired by a system for developing interesting pictures. Using a "plant grammar" and genetic algorithms, the program would present a number of pictures to a user, who would then choose or rate the pictures. Then the next generation would be created based on the user's choices (in genetic algorithm terms, the user acts as the selection function).
Goldberg is potentially aimed at composers of many different kinds: from people with relatively little musical knowledge who want to listen to a number of choices and choose their favorite, to musical experts who know exactly what they want and just desire some assistance auditioning and transcribing it.
How should themes be constructed? How many themes and variations should be available at any one time? How much control should the composer have over Goldberg's variation choices, both in terms of which musical passages get changed and how Goldberg modifies them? All of these questions could be answered in various ways to produce extremely different incarnations of the basic Goldberg idea. How these questions should be answered depends on which portion of the potential audience we particularly want to target.
For the purposes of this report, we loosely refer to composers as being "traditional" or "nontraditional". The former tend to be classically influenced, highly musically trained, and used to composing with a pencil and staff paper. The latter are more likely to use a variety of tools, including algorithmic and computer composition methods, and are more oriented toward experimental and electronic music. (Note that this split reflects a difference between Mike and Kevin and, relatedly, between two basic kinds of composers we knew and were able to test and survey.)
