A Mechanical Mozart, Ctd

by Patrick Appel

A reader writes:

Ah, something I can talk about intelligently, as a former grad student in AI and as a well-educated Mozart hound. I'm familiar with Cope and with Emmy, and a little bit with Hofstadter as well, from his Pulitzer-winning book on the subject of math, language, and music, Godel Escher Bach. And I dabbled with algorithmic composing back in the 80s, although on nothing like the scale of Cope. So I'm at least familiar with the controversies involved here.

It was one of the premises of Godel Escher Bach [GEB] that our minds recognize the formal language expressed in music. Music is a language, or at least a syntax without any definite semantic meaning. Poetry without words, if you prefer it that way. We are free to impose our own meaning on the language-like phrasings of music. All the tricks that are used to make music interesting, even simple things like repetition and symmetry and transposition, etc., engage the language processing parts of our minds.

Cope is playing an interesting game here. He feigns exasperation over the way conflicts that knowing his music is computer-generated creates in people, their unwillingness to judge it on the same basis as music they believe human-composed. But I think it's the reaction of these people that is most interesting, not the music itself. I don't condemn them, nor feel his exasperation. His real work of art here is in provoking this conflict, much more so than even the wonderful job he has done in algorithmically analyzing music.

I've wondered myself, too many times to count, why it is that this or that piece of music stirs me so much. I think, I'm being communicated with by a deep soul. Here is Mozart or Mahler or Beethoven speaking to me in a voice that I recognize, and it seems so clear that it must have some meaning. But if it is only notes, then there may not really be anything profound there at all, only my own projections. That's a very lonely thought.

Haven't you ever tried to share a piece of music with someone and felt frustrated by the experience? I have. I've given away CDs and dragged people to concerts, given little parlor lectures explaining how sonata form works, tried to transfer my enthusiasm about particular works to people so they would hear something the way I heard it. And yet, they usually don't. How can they react this way when I feel like I'm in the presence of God just listening to this music? And the answer is, the music itself is wonderful, but the feelings that I experience ARE my own projections. Most people may hear something sad and sweet when they listen to Mozart's Sinfonia K.364 second movement (youtube it) but they can't possibly hear it the way I hear it. We are trapped in our own private bubbles projecting feelings and meanings onto patterns of notes and sounds that remind us of things, that trigger feelings.

Music is like a Rorschach test, although a somewhat more reliable, perhaps, in that we assume the composer wanted to convey something that we might have picked up. Cope's programs have no assumption of such intent. That creates a problem for us. And when that music sounds just like something we are used to thinking of as the voice of God, wooWEE, it's cognitive dissonance salad time. You have to give Cope credit for this, whether you care for his music or not.

But now, let's go back to this wonderful question from the article:

"If a machine could write a Mozart sonata every bit as good as the originals, then what was so special about Mozart? And was there really any soul behind the great works, or were Beethoven and his ilk just clever mathematical manipulators of notes?"

Mozart and Beethoven followed a set of rules. Hat's off to Cope for analyzing them so well as to mimic them. But all the computer does is use the rules to create the music, while Mozart and Beethoven CHOSE the rules they used. A lot of it was plagiarism, that's true, although that is stretching the word a bit for sensational effect. The rules and devices of composition that Mozart and Beethoven chose for their compositions were chosen for what they hoped would create beautiful music rather than boring music. A human composer had to hear the music (or imagine it, in Beethoven's case) and decide whether it was effective or not based on what he projected into the music. Mozart's own Rorschach test projection of what the Jupiter Symphony means would be completely different from yours or mine, but he at least found it interesting enough to create THAT symphony rather than some other that followed the same rules.

I'm reminded of how, when I was a kid at the fair, they had booths with something called "Spin-Art." They gave you a cardboard canvas that you randomly squirted some paint onto, and then a machine whirled it at high speed, splattering it away from the center through centrifugal force. I was fascinated by it. After a while, I decided I liked some of them more than others. I took some home to show my Mom and told her the titles I had made up. This one is Battle of the Butterflies, or whatever. She thought I was terribly creative. What else would she think, she was my Mom, right?

But was it art? Was there "soul" in it? A f'ing machine just spun around in a circle for five seconds. Big whoopee. The laws of physics did all the work.

But there was still human input into it which affected the results. The process of choosing which spin-art to keep must have involved its own rules set, subliminal perhaps. I also remember that too much paint of too many different colors just created a big brown slushy mess, for instance, which wasn't very interesting at all. It's possible that with enough representative examples of the Spin-Art that I considered cool, a sophisticated program could determine a rule set that would reliably produce Spin-Art just as "cool" as what I created. Indistinguishable even. Nobody would win a prize for doing it, but I'm just saying, it could be done the same way Cope has analyzed classical music composition.

That program, like Cope's program, would follow rules that I created as a child for creating Spin-Art. It just wouldn't be CHOOSING good, better, best based on the kinds of projections that I made as a child. And it would never be able to grow to add new rules as the old rules became boring. And it would never get my Mom's approval at all.