Written by Matthew Parsons
Around 1980, the American composer David Cope was failing to write an opera. The commission had been hanging over his head for a while, but he just wasn't making any headway. So, he did what any forward-thinking late 20th-century artist would do: he figured out how to make a computer do his work for him.
The software he developed — Experiments in Musical Intelligence, abbreviated as EMI and pronounced "Emmy" — was able to analyze and imitate the work of a specific composer based on a sample of their work. It proved unable to imitate Cope's style, leading him to turn in his completed opera seven years late, but EMI did manage to come close to some of music history's better known names:
If you find that disturbing, you're not alone. This music sits smack in the middle of the uncanny valley. Some classical musicians expressed interest in performing EMI's music, only to be shot down by their frightened agents, who thought that the controversy of those performances could outweigh the novelty.
But, here's the thing: humans have been fascinated by the idea of algorithmic music — music made by some formal process with minimal human intervention — for ages. In fact, this fascination even pre-dates computers.
In the 18th century, there was a brief craze for dice games that allowed people to "compose" their own simple music. The first of these dates from 1757, but the best-known one has been attributed (probably misattributed, actually) to Mozart.
It works like this: Mozart (or whoever it was) wrote 11 possible first measures for a minuet, 11 second measures and so on. So, all you have to do is roll two dice to find your first measure, roll them again to find your second, and keep going until you've got a complete 32-bar piece of music. The system can produce nearly a septillion different minuets, although most of them sound basically the same:
In concept, this isn't so different from EMI. Both the dice game and the computer program use a set of pre-determined parameters — in one case, many distinct measures of many possible minuets; and in the other, a sample of musical works by a specific composer — to produce an output that can differ due to an element of randomness.
In a sense, the key difference between the two is that the dice game's human-determined parameters are on full display, whereas EMI's are hidden from view. Maybe that's why EMI creeps us out so much: because it gives the illusion that a machine is making music without human aid. But it is an illusion, as Cope is swift to point out: "A human built the machine, listens to the output, and chooses what's the best. What's less human about that than if I had taken years to just compose the whole thing myself?"
There have been plenty of other novel and successful musical algorithms since EMI started churning out Vivaldi and Beethoven imitations. Brian Eno got in on this approach with his "generative music." He explained: "Since I have always preferred making plans to executing them, I have gravitated towards situations and systems that, once set into operation, could create music with little or no intervention on my part. That is to say, I tend towards the roles of planner and programmer, and then become an audience to the results."
There are elements of this kind of rule-based, adaptive music in modern video games like Portal 2, as well. That game's music changes depending on the actions of the player.
Even CBC Music's gotten on board, with Scott Tresham and Ben Didier's C.P.E. Bach machine, which allows anybody with a printer and a pair of scissors to generate two-part counterpoint in the style of that famous composer.
But, perhaps unsurprisingly, the most impressive algorithmic music around today is still the handiwork of Cope. He followed up EMI with an even more flexible computer program, one so advanced that Cope even felt it deserved a proper human name: Emily Howell.
The main difference between EMI and Emily Howell is that, where the former was only able to imitate specific composers, the latter is able to synthesize a huge amount of pre-existing music into what Cope argues is "Emily's own style."
This, of course, is what human composers do. They take inspiration from the whole history of music, and use elements of it as ingredients in their own music. So, is this the point where algorithmic music finally becomes truly creepy? Remember, Emily Howell is still beholden to a set of parameters set out by David Cope — but this is a computer that makes music in a unique style.
Listen to her works, ye mighty, and despair. Or don't. I'll let you judge for yourself:
Follow Matthew Parsons on Twitter: @MJRParsons