Switched on Bach: David Cope's computer compositions

Professor David Cope speaks in purposeful abstraction, attempting to brace us for what we're about to see. We've been on the road for a while now, I tell him. We've seen a lot of strange and wonderful things -- robots and space shuttles and ghost hunts. "Yes, well," he answers quietly, as we ascend the stairs of his Santa Cruz, Calif., home. "I guarantee you've never seen a laboratory like this."

It's hard to say precisely what we've gotten ourselves into. It's a fairly standard suburban house from the outside, a few blocks from the base of the hill that holds the University of Santa Cruz. In amongst a forest of redwoods, it overlooks the pristine wilderness of the central coast that so famously inspired Kerouac, Miller and Steinbeck. Cope, a lifelong music professor, wears a denim jacket, floral button-up, white stubble and a sly smile. If there exists a walking manifestation of Santa Cruz, it might well be him. It's the perfect uniform for an unassuming computer music pioneer.

There's nothing of particular note to speak of downstairs in the living room, where Cope gives lessons on a grand piano littered with any number of music books. When we first arrive, Cope's wife answers the door slightly confused and momentarily sure that we're there to sell magazines -- the professor, it seems, has forgotten to inform her that our arrival has been pushed up by an hour. From upstairs, Cope suggests we shoot the art lining the walls above the piano as he readies himself for our conversation. "I made them on a computer!" he excitedly exclaims about the planetary orbs and psychedelic swirls -- mathematical, formulaic interactions imprinted into a bronzed-aluminum backing. They're a small selection of a seemingly infinite and diverse collection of Cope's artistic expressions that decorate the house.

Cope opens the door to his "laboratory," the small upstairs office where he creates much of the work that drew us to him in the first place. He's right, of course, there isn't a space robot around that would prepare one for Cope's workspace and its wind chimes -- hundreds of them, by rough estimate, of every shape, size and tone, from all corners of the world. If you've been in the position of having to buy the professor a gift over the past few decades, they are a go-to. And Cope will happily tell you the collection represents far more than just a propensity for dangling porch knickknacks. It speaks to a much deeper theme that runs across several decades of work.

"Like the wind chimes that surround us here, I'm interested in algorithmic things," Cope explains, shortly after taking a seat a few inches below the cluttered overhang. "I believe that the human brain works algorithmically. We blink our eyes, and our breathing and our heartbeat, and the blood going through our veins, the way in which our mind works is all algorithmic."

Cope turned 72 earlier this month, and while the nearby university deemed him a professor emeritus this year, his title is seemingly the only indicator of having slowed down. His laboratory is equally cluttered below the wind chimes, with piles of sheet music, notebook scrawl and stacks of books, many bearing his name on the spine. Most notable among these is New Directions in Music, a tome first published in 1971 and currently in its seventh edition.

"[It was] the only book of its kind at the time, and it was involved with the musical avant-garde," Cope explains, without a note of ego. "At that point there was very little on computers and music composition, but at least I had a small portion of that book involved."

His follow-up, New Music Composition would delve far more deeply into the subject of computer music, right around the same time he began dabbling in the form himself.

"I [composed] a piece using an IBM computer using punch cards," says Cope. "This is 1975, and it took [an] enormous amount of time to program and a very long time to actually get the results out, because the cards that we received an answer, in response to the input, had to be converted from those little holes in the punch cards to actual little physical notes on a page. But it succeeded -- at least in getting performed." The professor pauses and adds, thoughtfully, "It was an awful piece, just truly dreadful." Awful, perhaps, but it was a start.

Early the following decade, Cope began work on an opera, while staring down the barrel of a particularly nasty bout of composer's block. "I couldn't figure out why C# is any better than C to begin with," explains Cope. So, naturally, he did what any composer with a deep fascination in computer-generated music would do: he set out to build a system that could write an opera for him.

"I had a bit of a problem trying to figure out what my style was, so I chose Bach to imitate," Cope says. "I ran aground a few times before I succeeded in a program that uses a lot of data in the form of Bach's music, and a small analysis program, which analyzed the music for its rules, and then used that analysis to create new music in the style from the music in the database, but not actually replicating any actual piece in the database."

Experiments in Musical Intelligence (EMI) began in 1981, a project Cope affectionately nicknamed Emmy, in a tender bit of mechanical personification. It was a slight blurring of the lines that foreshadowed some of the inevitable questions surrounding the nature of creativity that such a program would pose. With the program capable of spitting out convincing Bach, Mozart and even Cope when fed the proper musical metadata, it immediately called into question the role of abstract concepts like creativity and inspiration.

Cope happily demos the 30-year-old program for us on an aluminum PowerBook, a far cry from the ancient beige Apple desktop that sits in the corner of the room, an ever-present reminder of Emmy's origins. The choir, Cope adds, wouldn't fit in the room, so we're stuck with listening to computer-generated sounds for the time being; MIDI-like music, in the place of beautiful choral singing. It's not the ideal way to hear a computer-generated Bach chorale, but given our budgetary restraints, it will do.

"I want to prepare you for the fact that it's not going to hit the top 10 any minute based on that," the composer adds with a smile. "But if you can be liberal in your thinking about it, you can hear the music as it would sound in a concert hall with a 16-voice, 32-voice choir singing it."

He enters the number of chorales he wants Emmy to compose (just one at the moment, though he notes that, at one time, he asked her to create 5,000, all of which can be downloaded from his site) and then, fittingly, clicks "compose." A few moments later, she begins, with her human creator listening intently along, conducting with his right hand.

"I would rate that one 'fair' at best," Cope begins, after a pause. "It was a good progression. It was very interesting. The cadences coming seemingly too far apart, and it ended incorrectly with, in this case, an E on top instead of a C. So I would rate it as maybe a C+ for output, and that's pretty bad."

Cope and Emmy try again. Another C+. On the final go-round, things are looking slightly up. "The ranges were a bit off," Cope explains, now fully in professorial mode. "I would say that was more like a B. Still not the best work at all, but not bad."

But while Cope grades the work on a sliding scale, all of the chorales are derived from Bach, and for the most part, they sound like the composer's work. "They just have these little deviations which come from mistakes," he adds. "Bach would do these one time; this maybe does it twice, and therefore it doesn't sound quite right. The program is not statistically based, and therefore it will sometimes be very creative with these things." It's those "mistakes" that further blur the line between human and machine. Anomalies in composers' pre-defined rules are what allow music and musicians to evolve -- they're also what keep Cope's programs from becoming too repetitious.

"The mistakes we make often are useless, but occasionally are extraordinary in making changes in the way music evolves and the way the human race evolves," explains Cope. "You can actually see that in music history. You can actually see a composer compose something that seems so bizarre; it's almost like they've made a mistake and then it becomes a major feature of their style." Computers, on the other hand, are designed to eliminate such aberrations. "In a way that's unfortunate," says Cope. "In my process, mistakes are included. They come out in the output, which makes it sound more human than otherwise."

In the years since designing EMI, Cope has further pursued the mysterious line that separates man and machine, fittingly continuing his human-based naming conventions with Alice, Sarah and, in 2003, Emily Howell, a nod to the program that started it all. "Howell is my father's first name and my middle name," explains the inventor. "It was a neat pairing of words because this program is interactive. I play an important role of interacting back and forth with the program."

Cope fires up Emily Howell, eager to show off his more recent creation. But where generating chorales with her predecessor was as simple as clicking "compose," Howell requires far more input. Through a series of dialog boxes, Cope slowly teaches her to speak in front of us. He enters phrases like "hello" and "hello, how are you?" and receives the sorts of responses one would expect from someone attempting to learn our language. "Are how hello," answers Emily Howell.

"Every time you input a word, or a note, or a chord or anything you'd like to define as a unit, it is placed within a kind of virtual node," Cope explains. "These nodes are placed in the order in which they're received, and they're connected to every other node with a series of arbitrary, randomized weights." The language in this case, is arbitrary. The dialog, instead, is intended to create associations inside the program; Cope has also done with music what he demonstrated with language for our benefit. It's a much more involved process that can prove far more frustrating, but the results speak for themselves. Cope has had the A+ results performed by full choirs. You can buy them on iTunes and stream them on YouTube.

Of course, Emily Howell will never be J.S. Bach, no matter how many iTunes downloads she may move, but her A+ compositions have been more than enough to reignite the sometimes troubling questions posed by her predecessor. And in a sense, that's precisely what has maintained Cope's attention for all these decades.

"Part of the teacher's job is to challenge students; to make them angry; to make them think for themselves," says Cope. "And I saw the program actually doing this, making people think about these things, at least some of the people. Some had closed minds immediately, and that was that, and I felt that was really good, very healthy."

Cope retired this year, after 46 years of working as a college professor. But he continues to teach, both in the classroom and through programs like Emily Howell. But as long as there are still questions to be answered, neither Professor Cope's nor Emily's jobs will ever be done.

"I'm working on another program now which will remain nameless because I don't want people looking for it necessarily, because I don't know if it's going to be successful or not," says Cope. "But I'm enjoying working on it, nonetheless."

This video originally appeared on Engadget Show episode 43.