Virtual Music: Computer Synthesis of Musical Style (review)

Virtual Music: Computer Synthesis of Musical Style. By David Cope. Cambridge, Mass.: MIT Press, 2002. [xiii, 564 p. + 1 CD. ISBN 0-262-03283-X. $45.] Music examples, illustrations, bibliography, index, compact disc.

David Cope is well known for his stunning, successful research in the algorithmic replication of musical style. Most significantly, Cope has coded computer programs that compose convincing musical works across a broad spectrum of musical styles. In listening sessions related to Cope's work, even accomplished professionals have found it difficult to determine whether specific pieces were composed algorithmically by computer or traditionally and directly by human.

In Virtual Music: Computer Synthesis of Musical Style, Cope and several distinguished scholars discuss questions that his research raises about human/machine creativity. What does it mean to compose with assistance from a machine? What differentiates human-composed and machine-composed music? What is art and what is craft? What constitutes musical style? How can musical style be recognized and replicated by a computer? Who or what is the author of a musical work composed by a machine?

Virtual Music is Cope's fourth book on computer-assisted composing. Its conceptual focus and accessibility to a wide range of readers differentiate it from his previous three books on computer-assisted composing—Computers and Musical Style (1992), Experiments in Musical Intelligence (1996), and The Algorithmic Composer (2000; all Madison, Wisc.: A-R Editions). In those works, Cope wrote primarily for specialists about technical aspects of computer- assisted composing such as artificial intelligence techniques, including pattern matching and inference engines.

Cope divides Virtual Music into four parts. Together, they give a comprehensive overview of his research. In the first section, he defines virtual music as a "broad category of machine-created composition which attempts to replicate the style but not the actual notes of existing music" (p. 1). Among the historical precedents that Cope cites are baroque figured bass, eighteenth- century Musikalisches Würfelspiel (musical dice), pop music, and computer-based projects by Lejaren Hiller, Leonard Isaacs, Iannis Xenakis, and Charles Ames.

Cope's definition of virtual music is pragmatic and modern; it affirms and extends the musical canon. As such, Cope's definition does not build upon the fascinating body of interdisciplinary postmodern critical theory on all things digital and virtual. Nor does it look to texts such as Digital Mantras (Cambridge, Mass.: MIT Press, 1994) by Stephen Holtzman, who advocates for virtual worlds that derive from the computer-assisted manipulation of abstract structures. The virtual possibilities and the corresponding digital aesthetic that Holtzman describes are not tied directly to the style of existing music, as is Cope's definition.

In the second section of Virtual Music, Cope describes the process that his software follows when composing. Cope explains how his programs first look for patterns in a database of compositions, then he shows how patterns, once identified, become the basis for new works that inherit stylistic attributes from repertoire in the database. Cope sees the computer-composed works that emerge from this process as "just as much ours as the music created by the greatest of our personal inspirations" (p. 140). For Cope, machine-based creativity extends human capability.

Through commentary from musicologist Eleanor Selfridge-Field, software engineer Bernard Greenberg, music theorist Steve Larson, composer Jonathon Berger, and cognitive scientists Daniel Dennett and Douglas Hofstadter, the third section of Virtual Music further explores the dynamic between human and machine creativity. Hofstadter, the Pulitzer prize-winning author of Gödel, Escher, Bach (New York: Basic Books, 1979) poses a difficult question: Does the success of Cope's computer-assisted composing project show that writing music is a craft and not an art?

This interesting question generalizes to other projects that explore creativity through algorithmic means. Two such examples are IBM's Deep Blue chess-playing machine, which is discussed in Virtual Music (p. 34-36, 40, 344-45), and Harold Cohen's well-known Aaron, an intelligent painting program, which the book does not mention. (A...