Crap, crap. Crap, crap, craaap...Ahhh. Now here's good work. --The Joker, Batman
George Orwell imagined in 1984 that novels would be produced mechanically, subject to orders from Planning Committees, edited by Rewrite Committees, every new one as dull and designed to specifications as the ones before. The process would go like so:
'Oh, ghastly rubbish. They're boring, really. They only have six plots, but they swap them round a bit. Of course I was only on the kaleidoscopes. I was never in the Rewrite Squad. I'm not literary, dear -- not even enough for that.'
SC used to think this was satire, or at least an overly pessimistic view of the future. But then one thinks of Lou Pearlman, who made a fortune cranking out boy bands by formula ([technically, it was his second fortune -- ed.]), or the paint-by-numbers approach to cranking out books under the V.C. Andrews (who wrote at most 7 novels in her lifetime; "her" 40th comes out in March) or James Patterson (the subject of an admiring profile in the Harvard Business Review) names, and it's hard not to think that music and publishing executives see Orwell's vision as a beautiful dream.
So your host was very interested when Mrs. SC relayed news that she heard on the radio yesterday, of a company specializing in computer-based analysis of music to determine its "hit" potential. That company, Polyphonic Human Media Interface, promises to ensure that our future will be an unbroken chain of Britneys and Backstreet Boys for eternity.
Actually, the PHMI folks would take considerable umbrage at that statement. Here's their view of what they're up to, extracted from various parts of their FAQ:
"[E]very new style of music that has come into being: country, rock, punk, grunge etc. have all had similar mathematical patterns and the hits in those genres have all come from the same hit clusters that exist today...artistic integrity and creativity are the lifeblood of the music industry and are of paramount importance to our business...Our computers have not invented anything, rather they've only detected patterns and parameters that already existed...Hopefully by identifying them [the patterns] musicians can become better composers and more insightful."
But all this hemming and hawing about the importance of the human element is pretty quickly laid to rest by examining their explanation of the technology. While PHMI is fairly cagey about the exact clustering algorithms used, they acknowledge attempting to categorize song features like "melody, harmony, tempo, pitch, octave, beat, rhythm, fullness of sound, noise, brilliance, and chord progression" (SC doesn't claim to get all of this; what exactly could they mean by "brilliance"?). The associated graphs indicate that their analyses attempt to find how similar new songs are to previous commercial successes. Aside from a point regarding the inability of their system to account for lyrics, it is very difficult to see how this analysis is anything but an attempt to prove that new music is more of the same. If two songs have similar melodies, harmonies, tempos, etc., then it stands to reason that they'll sound alike.
Looking at a sample report provided by PHMI gives some room for optimism, albeit of a very limited sort. Assuming the data is not entirely fictional (although your host couldn't find any song called "Wild Party" by an artist named "Hal"), their software claims that a tested song is very similar to both a raucous rap hit by 50 Cent, and a rather more laid back R&B hit by R. Kelly. Admittedly, most of what sounds similar about them are the intonations of Messrs. Cent and Kelly; the background music sounds very different to SC's ears. But then, he had to hear the Black-Eyed Peas' Where Is the Love? dozens of times before coming to the stunning realization that it was in fact the old Nestle Crunch jingle (all clips courtesy of Tower Records' website). So the common features don't necessarily have to beat you over the head. But there's also no denying that their list of hits all fall very strongly into the hip-hop/R&B genre and are very much examples of conventional genre thinking.
The PHMI folks are of course attuned to the concern that their work will be used to crank out "soulless digi-hits". But they also promise to cut back on the number of songs on CDs that consumers won't like. And they're sure that they're onto a "a highly accurate and scientific tool", which means nothing if not repeatable. SC sees no reason not to take them at their word on this point, nor on their claim that major labels are already trusting them to help make decisions. Anything that cuts down on risk is music to an executive's ears. The question is whether it will be to ours.
Comments