Musical taste is an inscrutable thing. Some songs are immediately pleasing to the ear, while others demand a lot of work and repeated listens. Some are hailed by critics and spurned by the public, while others are panned by critics as they ascend to the top of the charts. On a much smaller scale, two people could have the same reaction to nine bands, but then diverge greatly on the tenth.
There are a number of bands that I have failed to like, even though I have tried quite hard to like them, and even though the critics adore them—even though my friends whose musical tastes I respect have recommended them. At the risk of disappointing some of my friends (Michael), here are a few of the acclaimed bands and musicians that I want to like but somehow can’t:
- The Arcade Fire
- Bright Eyes
- Elliott Smith
- The Rolling Stones
- Ryan Adams
- The White Stripes
It all makes me wonder, can we really recommend music to others with any degree of confidence? Can we say, “if you like this, you’ll love that”? What determines what we like anyway? Certainly cultural forces are at work. Familiarity must help, too, as long as it doesn’t turn into over-saturation. A little peer influence may nudge us in a certain direction, but it can only take us so far. Genetics? Who knows.
Some companies think they have it all figured out. Siren Systems Inc. has a website called Soundflavor that “marries objective (and unbiased) metadata, artificial intelligence, and listener ratings to yield unparalleled accuracy” in recommending songs that match a user’s individual tastes. Polyphonic HMI claims to have “developed proprietary music analysis technologies capable of identifying music preferences of a user or the whole current recorded music market and intelligently selecting music to recommend to the user or to release as a single.”
A few weeks ago on NPR’s Weekend Edition Saturday, Scott Simon interviewed the founder of Savage Beast Technologies. The company has created the Music Genome Project, “an effort to break music down to its most elemental forms, using a computer program that evaluates each song by 400 distinct musical attributes.” Another company, calling itself MusicGenome, apparently has the same idea.
Audioscrobbler is a community-based website that matches users with others who have similar musical interests. It also has a streaming music feature, in collaboration with LastFM, that lets people listen to a personalized online radio station of sorts based on a user’s profile. I use it occasionally at work, and it works quite well.
It’ll be interesting to see how these music recommendation programs develop over the next few years as the demand from music conglomerates for sophisticated analytical tools increases. After all, the chief goal of the record companies is to remove all risk—and spontaneity—from the business of selling music. Makes for happy shareholders.