When it comes to ways that technology can ruin music, it looks like one bad idea deserves another.
Two weeks ago, in this space, we saw how the mathematical magic behind Auto-Tune has polluted modern popular songs in ways its algorithmic ancestors could never have anticipated. But the march of technology is relentless, and scientists are ever on the look-out for new ways to ruin music.
At least, one could be forgiven for coming to that conclusion based on a report in mid-July that EMI Music, the company originally founded back in the late 19th century by the inventor of the gramophone, Emile Berliner, was teaming with a nonprofit British scientific organization called Data Science London to invent a hit predictor.
To be more precise, they are attempting to invent an algorithm which will "predict if a listener will love a new song." To that end, they hosted a 24-hour "hackathon" two weeks ago, in which participating data scientists competed for a £6,500 prize (that's a little over 10,000 Yankee dollars) for coming up with the best means of predicting "a listener's level of appreciation for songs and artists, based on the listener's demographics, word associations and the past interviews contained in the EMI Million Interview Dataset."
"The EMI Million Whozit?" I can hear you ask. The EMI Million Interview Dataset is a special database purported to catalog the interests, biases, attitudes and musical knowledge of pop aficionados from around the globe. It is a massive, heaping gob of data EMI has been building for some time now, and with it, the company believes it will succeed in quantizing and decoding the secrets behind what constitutes a hit.
Now, the phrase which comes to Craven's mind upon hearing this is unprintable but involves a surplus of guano and a deficit of sanity. I strongly suspect the neural and cultural complexities which, in their tangle, conspire to make a song popular will stymie the data miners, just as true artificial intelligence remains distant and out-of-reach years after we were assured that robots would be reasoning by now. How can scientists cajole themselves into believing you can encode within software the special spark which music ignites in the human mind when we cannot yet teach a computer to understand a joke?
Of the 138 teams which competed in the hackathon, the winner was a group out of Shanghai, whose algorithm will reportedly be made public soon. We'll see then if coders can succeed where so many executives, producers and artists throughout pop history have failed.
Here's Craven's prediction: The technology will not live up to its promise, but will appeal to desperate record labels and publishing groups, which nowadays are controlled by bean counters. (EMI itself is now a division of Citigroup, the banking conglomerate.) Label heads will incorporate it into their planning and signing, thereby causing pop music to further constrict and conform and homogenize beyond its current state of overproduced blandness. Because however large EMI's dataset of interviews grows, it will always be distorted by the assumptions of the interviewers, the questions left unasked, the inability to express the ineffable. Even the earliest computer programmers understood a principle they called GIGO: Garbage in, garbage out.
Notes is supported by the Gay and Lesbian Fund, helping the Girl Scouts build leaders throughout Colorado.
Craven Lovelace produces Notes, a daily cultural history of popular music, for KAFM 88.1 Community Radio, kafmradio.org. You can visit cravenlovelace.com for more of his musings on the world of popular culture.