The development of computers, and devices that have computers embedded within them, has encouraged the exploration of machines that perform human actions. This exploration includes software that performs intellectual tasks, such as playing chess and composing music, and software-hardware systems that control machines robotically to perform physical tasks.

In the early years of modern computer development, mathematician and computer scientist Alan Turing developed ideas about computational algorithms and artificial intelligence. He hypothesized about the fundamental definition and nature of intelligence in his 1950 article “Computing Machinery and Intelligence“. In that article he proposed what eventually came to be known as the Turing test, which, if passed by a computer would qualify that machine as exhibiting intelligence. His premise can be paraphrased as implying that the appearance of human intelligence, if it is indistinguishable from real human behavior, is equivalent to real intelligence, because we recognize intelligence only by witnessing its manifestation. (This assertion was interestingly disputed by John Searle in his 1980 article “Minds, Brains, and Programs“.)

How can musical intelligence, as manifested in music composition, be emulated by a computer? To the extent that musical composition is a rational act, one can describe the methodology employed, and perhaps can even define it in terms of a logical series of steps, an algorithm.

An early example of a music composition algorithm is a composition usually attributed to Wolfgang Amadeus Mozart, the Musikalisches Würfelspiel (musical dice game), which describes a method for generating a unique piece in the form of a waltz. You can read the score, and you can hear a computer generated realization of the music. This is actually a method for quasi-randomly choosing appropriate measures of music from amongst a large database of possibilities composed by a human. Thus, the algorithm is for making selections from among human-composed excerpts—composing a formal structure using human-composed content—not for actually generating notes.

In the 1950s two professors at the University of Illinois, Lejaren Hiller and Leonard Isaacson, wrote a program that implemented simple rules of tonal counterpoint to compose music. They demonstrated their experiments in a composition called the Illiac Suite, named after the Illiac I computer for which they wrote the program. The information output of the computer program was transcribed by hand into musical notation, to be played by a (human-performed) string quartet.

Another important figure in the study of algorithmic music composition is David Cope, a professor from the University of California, Santa Cruz. An instrumental composer in the 1970s, he turned his attention to writing computer programs for algorithmic composition in the 1980s. He has focused mostly on programs that compose music “in the style of” famous classical composers. His methods bear some resemblance to the musical dice game of Mozart, insofar as he uses databases of musical information from  the actual compositions of famous composers, and his algorithm recombines fragmented ideas from thos compositions. As did Hiller and Isaacson for the Illiac Suite, he transcribes the the output of the program into standard musical notations so that it can be played by human performers. Eventually he applied his software to his own previously composed music to generate more music “in the style of” Cope, and thus produced many more original compositions of his own. He has published several books about his work, which he collectively calls Experiments in Musical Intelligence (which is also the title of his first book on the subject). You can hear the results of his EMI program on his page of musical examples.