Computers vs. Composers

Recently, a super-computer—Google’s “Deep Mind” artificial-intelligence network—defeated the three-time European champion in the “Chinese-chess” game Go, five matches in succession, which sounds pretty definitive. Go is the deceptively simple-looking game of black and white stones of which tic-tac-toe is a primitive shadow. It is notoriously complex, and notoriously susceptible of dominating the human mind in an all-consuming manner: my late friend the Korean-American composer Donald Sur claimed to have lost three or four years of his life to Go, until he finally went cold-turkey.

Nonetheless, score one for the hominids. While computer algorithms, given adequate input, have generated more-or-less recognizable Bach counterpoint in two and even three voices, the synthesis of an entire, musically coherent two-part invention, let alone a form as complex as a Bach-style Fugue, appears to remain beyond them…for now. As to anything along the lines of a Schubert lieder, a Beethoven quartet, or even a Beatles song, fuggetabahtit.

People have been trying to structuralize music composition since Pythagoras, with varying degrees of success. Bach himself used the letters of his name (“h” is b-flat in German) as pitches in more than one theme, and Schumann followed his example, while John Cage famously threw the I-Ching to generate sequences of pitch, duration, rhythm, and so on. But these are examples only of “input,” raw materials for themes of motifs; the art and complexity of composition still was performed by humans—in these instances, exceptional ones.

Of course, computers are getting closer; search “algorithmic composition examples” on YouTube or SoundCloud and you’d find thousands, representing every degree of accomplishment. Most are in the “contemporary” vein: what I like to call “academic music,” since nobody but academics (and me) listens to it. Interestingly, the best algo-sitions realized by humans or machines playing traditional acoustic instruments, as opposed to computer-synthesized sounds, come close to matching stylistically similar work by meat-bag composers. Certainly close enough that lay people, i.e.nearly all of us, would be hard-pressed to pick ‘em. Ironically, some of the most advanced adaptive/real-time algo-sition work is being done by video-game developers. (Though come to think of it, I suppose the irony has been mostly leached out of video-game references by the drone wars.)

All of this may say more about the place in which “serious” music composition finds itself today than it does about the state of algorithmic composition. If this makes me sound reactionary, so be it. (I’m not. Really.) But either way, it can only be a matter of time. When computers first started beating world-class players of Western chess a decade ago, it was said that Go, mathematically thousands of times more complex, would withstand the challenge for decades.

Looks like the old saw may really be true: Give an infinite number of monkeys an infinite number of typewriters, and sooner or later you get “Hamlet.”

What’s that you say? “What’s a ‘typewriter’? Who’s ‘Hamlet’?”

Broberto's picture

Hamlet=a small ham.
Typewriter=geez, no freekin' IDEA!
(I am a are a Monk-Key!)