The insignificance of the wunderkomputer

Much ado has been blogged about Emily Howell, the computer program that writes Bach chorales , Mozart sonatas, and other kinds of pieces in the style of various composers. You can listen to two samples of the program’s work here.

The samples are impressive, but there are three considerations that temper the achievement:

First: the program itself is not simple. The programmer added innumerable tweaks in order to get the output to sound right. Presumably he did this with a clear idea of what the output should sound like. Why he should add tweaks to produce certain kinds of output (beautiful output) and not other outputs preserves the basic mystery of why we find the particularities of classical composition beautiful. The algorithm does nothing to explain this mystery.

Second: the pieces are based on statistical analyses of past composers. These analyses extracted methods, then, which were created not by the program but by human beings. Presumably the program cannot create new musical idioms as composers have done.

Third: the samples, while surprising and occasionally beguiling, are not as good as the originals. Not even close. The first sounds like it’s supposed to imitate Faure, though the article doesn’t say, and it’s quite successful at capturing his characteristic harmonic moves, but try listening to it five times in a row, as I have just done. Then listen to a piece by the real Faure (an amateur musician playing on a wretched piano, which I hope evens the score somewhat, since the computer’s music was generated by a player piano).

The second sample‘s inferiority is harder to understand, for several reasons. It is a perfectly correct fugue, and the mere form of a fugue is beautiful (I, for one, would love to listen to, say, a 64-voice fugue in Bach’s style written by Miss Howell). Also, many of Bach’s ardent admirers miss what is exciting about him. As grouchy Adorno wrote:

“his influence…no longer results from the musical substance of his music but rather from its style and play, from its form and play, from formula and symmetry, from the mere gesture of recognition.”

Bach is himself partly to blame. His greatness is particularly difficult to grasp, among the really greats. DB Hart wrote a piece in First Things this morning that describes the poetry of T’ao Ch’ien in terms that can be applied to Bach:

“The greatest writers of China’s poetic golden age regarded him as the absolute virtuoso of ‘the natural voice,’ almost magically able to combine the subtle and the simple in verse that was most lyrical precisely where it appeared least adorned. ‘On the outside,’ remarked the great Sung poet Su Shi (Su Tung-p’o), ‘it is withered, but on the inside abundant. It seems plain, and yet is truly beautiful.'”

With that in mind, listen to the most hellish performance I could find of a Bach fugue.

All this is not to say that the program doesn’t mean anything. As with many technologies, it neither upends the way we’ve been doing things for millennia nor leaves everything the same. Specifically, I no longer see any use for the industry that composes music to accompany movies. If I were a studio executive, I’d already be inquiring about how much it might cost to licensed a copy of Emily Howell.