Dave sent me a link to a long New York Times magazine article on Pandora.com and the Music Genome Project, The Song Decoders. You’ll need a lot of time to read it, but it’s a fascinating article. The founder of the project and Pandora, Tim Westergren, is my hero for doing something using actual humans—actual trained musicians in many genres, y’know, experts— that everyone else would let either happen by itself (e.g. indexing of websites using keywords) or “crowdsourcing” (e.g., Last.fm and so many others). The underlying philosophy is that what other people like is not the most important factor in identifying new music one might like. It could be starting point, maybe, but other people just don’t know everything, y’know? They know what they like, but sometimes that’s not terribly relevant. If your mission is to discover new music that appeals to you, the opinions of the crowd would only lead you to stuff you already know and connections that have been made countless times already. Meanwhile, if you can get people together to classify recordings according to dozens of criteria, then use powerful computing technology to explore the resulting web of connections… well, I believe it works.
When I first learned of the Music Genome Project, it was one of those few times I actually thought “Wouldn’t it be GREAT to work there?”
Westergren is hosting a “town hall” on Pandora in metro Detroit tonight. I was tempted to go, but I doubt I’ll get out of work in time.