Hit Song Science

Hit Song Science is a new technology that allegedly can analyze new songs and tell whether or not they will be hits…

Hit Song Science is a new technology that allegedly can analyze new songs and tell whether or not they will be hits…

HSS is an artificial intelligence program that “can isolate individual patterns in key aspects of the music that humans detect and that help determine whether or not they like a given song.”

The description of how the program works is suspiciously vague, or so general as to be meaningless: “this technology is able to detect what those melody patterns are as well as decipher patterns in other aspects of the music such as beat, harmony, pitch, octave, fullness of sound, brilliance and chord progression.” But it can also “recognize hidden market trends” and extrapolate from them. Hmm, I wonder which is more important for a song’s becoming a hit: chord progressions or hidden market trends?

The debate over artificial intelligence has largely focused on the question of whether, if a machine seems to be conscious, or acts the way that we would expect a conscious being to act, we are then justified in assuming it is conscious. This is the assumption, of course, that we routinely make of other human beings, whose consciousness we cannot access directly. “Strong AI” people argue basically that, if it acts like a duck, it is a duck. Whereas skeptics suggest that a machine might well be able to simulate consciousness, without for all that actually being conscious.

Now, nobody is suggesting that the Hit Song Science program is actually conscious, or that it in any way actually enjoys the music it predicts people will enjoy (or more accurately, rush out and buy). But it strikes me that, precisely for this reason, it’s a kind of answer to the AI debate.

From the point of view of marketing and commerce, HSS is the perfect customer. For the only thing that matters to the music industry and to advertisers is customer behavior–making purchases, enunciating “revealed preferences.” Any actual inner subjective enjoyment is superfluous at best, a perilous distraction at worst. If only we could eliminate it entirely, all that would be left is consumer behavior, and a song that was supposed to be a hit would be a hit, every time. Disappointing album sales would be a thing of the past.

HSS is supposedly able to analyze and unearth the unconscious reasons why people like songs; consciousness is just an unnecessary epiphenomenon of these deep, unconscious “revealed preferences.”

The question for AI, therefore, is not whether computers can approach the human experience of being conscious, but whether humans can approach the ideal computability built into the very structure of every Turing Machine.
Just as, in today’s postmodern mediascape, we no longer have images of real things, but instead a situation where images themselves are the realest things: so, rather than dismissing the simulation of consciousness, we can dismiss whatever in the mind is not a simulation.

As a quantitative social scientist once said: “Culture is just an unexplained variable.” Doubtless the same goes for consciousness. Once we get rid of the irritation of those unexplained variables, we are left with the perfect world of the political economists: one in which autonomous agents act rationally at every moment, so as to maximize their utility, as expressed in the way they allocate their expenditures.