Computer Models of Musical Creativity , MIT

Music and computers are natural partners. For centuries, music theory has been closely aligned with (and an important source of inspiration for) science and mathematics. Music compositions, however, are too large and complex to model mathematically, at least in practice. For the most part, science and mathematics diverged from music composition and developed independently—but then computers came along. As early as 1955, Lejaren Hiller and Leonard Isaacson composed the “Illiac Suite”, one of the first pieces of music to be written by computer. In spite of this promising beginning, Computer Music as a field has been dominated by studies of sound synthesis and control, leaving the composition up to humans. However, the idea of automated composition has attracted a steady stream of composers and researchers over the years. One branch of this work on automated music composition focuses on “algorithmic composition”, where the emphasis is on abstract sequence and pattern generation that happens to make interesting music. So far, this approach has had the greatest impact on music, but most computer scientists and at least some composers find this disappointing. With powerful computers and AI techniques, it seems that we should be able to build sophisticated models of the music composition process. This requires not only sophisticated representations of music and music structure, but also a model and approach to musical creativity. For many musicians creativity lies at the core of music making and is therefore the most human of skills. Computer Models of Musical Creativity describes its author’s many-faceted approach to music composition by computer, emphasizing the nature of creativity, how it can be modeled, and how these models are applied. I find this book especially interesting because David Cope is one of the few composer/researchers to attempt to model traditional musical styles of composers such as Bach, Mozart, and Beethoven. (See the figure.) While the musical quality of his computer-generated works has been the subject of much debate, the fact that there is any debate at all is quite an accomplishment. Just as computers and AI have raised many difficult questions about the nature of intelligence, Cope’s work and this book raise questions about “creativity”—where does it come from?, can computers be creative?, and is creativity different from intelligence? While most work in computer generated music ultimately relies on human creativity or possibly serendipitous discovery filtered through human perception and selection, Cope’s work is much more concerned with the construction of creative processes. Part I of the book covers definitions, background, and current models of musical creativity. This is not an exhaustive review, but it covers a wide range of interesting topics relating to machine creativity including the difference between music and language, definitions of creativity, originality vs. creativity, and tests for creativity. Cope introduces the reader to many theories on creativity, ranging from Freud, who viewed creativity as a “defense mechanism and an unconscious protection against unpleasant thoughts” (p. 40), to Chris Thorton’s recent proposal that creativity is a form of “run-away learning” in which objectivity is relaxed. Cope rejects the notion that creativity results in entirely original ideas or work; thus, “originality” is different from “creativity”. Instead, Cope defines creativity as “The initialization of connections between two or more multifaceted things, ideas, or phenomena hitherto not otherwise