21 David Cope Makes Music That Is “More Bach than Bach”
Good music requires no further justification, regardless of its composer.
—David Cope74
David Cope has made a life’s work out of creating Bach that out-Bachs Bach and Mozart that out-Mozarts Mozart, with the help of increasingly sophisticated algorithmic “collaborators.” His work has attracted fury from the musical establishment, which accuses it of having “no heart” or simply ignores it.
Douglas Hofstadter, however, author of a key work on computer science and cognition, Gödel, Escher, Bach, disagrees. He writes, “In twenty years of working in artificial intelligence, I have run across nothing more thought-provoking than David Cope’s Experiments in Music Intelligence. What is the essence of musical style, indeed of music itself? Can great new music emerge from the extraction and recombination of patterns in earlier music? Are the deepest of human emotions triggerable by computer patterns of notes?”75 Cope’s work stirs all these profound questions and more.
Cope started as a highly acclaimed young composer whose music was performed at Carnegie Hall. But he was also interested in computer composition and set out to write a piece for a behemoth IBM mainframe. The work “was a bunch of crap,” as he puts it, because the machine couldn’t process musical data in a viable manner.76
Then in 1980 he was commissioned to write an opera and came up against a severe case of writer’s block. He would sit for hours at the piano with a blank score sheet in front of him, falling ever deeper into despair. Finally, he decided to have another go at writing software to generate music, using a new generation of machines—laptops. He wanted to examine what constitutes style, particularly his own.
For Cope, the great revelation is that, as in art and science, music progresses through the accumulation of knowledge. In essence, it is inspired plagiarism. Picasso built on the work of Cézanne and Ingres, Einstein on the research of Maxwell and Lorentz. Joyce provided hints for Beckett, who inspired Pinter. No one starts with a blank slate. Everyone begins by composing “in the style of.” We all have a vast foundation—an internal database—of musical references. Composers are the people who have the ability to transform that foundation into distinctive new patterns. As Steve Jobs said, in words he attributed to Picasso, “Good artists copy; great artists steal.” If Mozart were alive today, Cope asserts, he for one would certainly be using computers to augment his work.
For his music-generating software, Cope built a database of hundreds of scores by Bach, Chopin, and other masters. He broke them down into notes, coding each by pitch, duration, and volume, when the note occurs in the piece, and the voice or instrument that makes it. This is the only database he uses, to make sure he is working with clean data, free of errors. His pattern-matching software lifts patterns from the data and from these infers rules for composition. To check on his software, he took chunks of notes from Bach’s scores and used Markov chains to reassemble them according to the patterns the software had discovered—which turned out to be the rules for composing baroque music. Using Markov chains enabled him also to factor in the randomness that is a crucial part of the creative process.
He called his program Experiments in Musical Intelligence (EMI), and it was soon churning out enormous amounts of music. It seemed to take on a personality of its own, and so it became Emmy. “The processes Emmy uses are the same as composers use,” Cope says.77 Like any composer, it has a database—a vast inbuilt knowledge of music—with rules for composing. But it’s not as simple as that. What distinguishes great composers is the particular notes, motifs, and themes they choose and how they choose to apply the rules for composition—rules they may sometimes break.
Algorithmic theory—the kind of mathematics he uses—is of no interest to Cope. “I don’t give a damn what the algorithm is as long as it produces the output I want,” he says.78 He devises his own algorithms rather than using already existing ones. “My algorithms are my own.”
Eight years down the line, he came to the conclusion that there was no such thing as his own style distinct from everyone else’s. Inevitably, it had to be based on the styles of others. Emmy brought this home to him when he noticed that the style patterns in his own works contained those of others. He finally conquered his composer’s block and, in 1987, completed his opera, which opened to modest acclaim. The Richmond Times Dispatch described one passage as “a supreme dramatic moment, punctuated by the captivating beat of drums.”79
By now, Emmy was pouring out thousands of scores. Cope curated them in the same way that composers keep some of their creations and discard others. By 1993 he was ready to release an album, Emmy’s first, Bach by Design. But record companies refused to produce it, saying that it was not contemporary music. Cope’s reply was, “Then what is it?”80 Eventually Centaur Records agreed to bring it out, but the musicians claimed that the fingerings were too difficult. In the end the selections had to be generated on a MIDI-controlled Disklavier (a piano capable of being hooked up to a computer). It was, Cope says proudly, the first album of music neither composed nor played by human musicians.
Cope was immediately inundated with hostile reviews. Some critics dismissed his music out of hand simply because it was produced by a computer, which, they said, by definition could not display creativity, something that only people possess. Douglas Hofstadter wrote of how disturbing he found it to be moved by what was in effect twenty thousand lines of code: the “composer’s soul is irrelevant to the music [and that] would be a tragedy, because my entire life has been moved by music.”81
According to Cope, the view that “technology represents another world, alien to truth and beauty, is a basic trope for the technophobe.”82 Google’s Blaise Agüera y Arcas sees it as a C. P. Snow “two cultures” scenario with two opposite poles, the “technophobic humanist and the inhuman technologist.”83
Some reviewers asked why computers should try to do something that we have been doing successfully for years—namely, writing music. Says Cope, “Good music requires no further justification, regardless of its composer.” He adds that the “computer composed—not generated—music. … Listeners must set aside any notions about creativity being unique to humans.”84
Cope and Emmy’s next recording, Classical Music Composed by the Computer, included selections in the style of Bach, Beethoven, Joplin, Mozart, Rachmaninov, and Stravinsky. It was performed by musicians and received better reviews.85 Hofstadter, by now a fan of Cope, was quoted in the New York Times as saying, “Nothing I’ve seen in artificial intelligence has done this so well.”86
Some critics claim that when Cope’s algorithm assembles phrases, what emerges is merely pastiche.87 He responds, “The term pastiche is cruel and unfair. The way things fit together is in a way to make musical sense.” He adds, “Emmy’s output has often been mistaken for music written by musicians of high caliber; in fact, Emmy’s Chopin is more Chopin than Chopin. The processes Emmy uses are similar to what human composers use. [Is not] everything a pastiche?”88
Another criticism is that Cope’s music is most successful only when its scores are played by musicians. The critic Stephen W. Smoliar wrote, “Whether or not it was music when it came out of the computer, it was certainly music” when it was played by a musician.89
In the following years, Emmy produced some 1,500 symphonies, 1,000 piano sonatas, 5,000 Bach chorales, and 1,000 string quartets, as well as operas based on the music of Mozart, Schumann, and Mahler, with librettos taken from the composers’ letters. “It’s the life of the composer through the composer’s words,” Cope explains.90
But he was becoming bored with producing music “in the style of,” “tired of people wanting to see Beethoven’s Tenth Symphony and a finished version of Schubert’s Unfinished Symphony.”91 He felt that what made a composer fully understandable, fully affecting, truly human was the fact of mortality. Composers have to die, and the ending makes sense of the entire oeuvre. So in 2004, he turned Emmy off.
He then set out to look for ways to create a computer program that could develop its own musical style. In 2004, Emily Howell was born—Emily after Emmy and Howell, which is Cope’s father’s name and his own middle name.
Emily Howell has a memory that involves an intimate understanding of the works of thirty-six composers, from Palestrina to Cope himself. Cope’s relationship with it is far more collaborative than it was with Emmy. Whereas he programmed Emmy very extensively, he communicates with Emily Howell via dialogue boxes, providing feedback to prompt it to evolve its own style. First he feeds in a musical phrase, then it responds, and Cope either accepts or declines the result, fine-tuning its knowledge base more and more.
Emily Howell’s first album was released in February 2009 by Centaur Records. Entitled From Darkness, Light, it consists of Opus 1, 2, and 3 compositions for chamber orchestra and multiple pianos. A later work, Breathless, was released in 2012.
In a review, Mark Howell of the Guardian mentioned an academic who “warmed” to From Darkness, Light but “then rapidly chilled when advised where it came from,” focusing on the music’s source rather than the music itself—a common response, unfortunately, to AI-produced music.92 Says Cope, “Regarding stylistic machine composition, I am surprised it took so long for the world to catch up with the idea that it has merit at all. … Now it’s caught up—and I am truly delighted with that. In the next twenty to thirty years magical things will happen in the arts. … Things are changing.”93
Cope does not believe in genius. People achieve great things just by dint of hard work, he says, with some native intelligence thrown in. Some people just think faster than others. He claims that with enough practice he could beat champions at chess, provided there was no clock. In fact, he continues, real-life chess geniuses are like computers, terrifyingly fast and accurate.
Cope feels computers can be creative and that to enjoy his music listeners need to set aside their prejudices about creativity being unique to people. But he also believes that computers are only machines and lack human qualities such as emotions. To reconcile his conflicting views of a creative machine that nevertheless has no human qualities, he offers the following rather gnomic definition of creativity: “The ability of the universe and of human beings that do not seem to be logically associable.”94 This seems to mean that human beings have a oneness with the universe that transcends logic and is thus beyond anything machines are capable of. A way out of these conflicting views, he says, would be the development of a “truly chemical machine,” possessing human qualities such as emotions.