There is such a thing as too much choice.
Wouldn't it be nice, considering the near infinite possibilities of music on the Web, from downloads to radio, if someone just gave you what you wanted?
While not quite perfect, Pandora wants to be that special someone. It's a website that creates personalized "stations" based on your input. You pick an artist or song as a starting point, and then Pandora, drawing on its 400,000-song database, offers selections that it thinks will match your taste. A simple thumbs-up or thumbs-down for each selection allows further customization, which, hopefully, results in a playlist that's exactly what you want to hear.
Pandora ( www.pandora.com ) has its roots in the Music Genome Project, an experiment in classification founded in 2000 by Tim Westergren, a composer and musical researcher. Together with a team of like-minded musical scientists, Westergren categorized the works of over 10,000 artists, identifying hundreds of different musical attributes or "genes" that could make up a song.
According to the site, Pandora began through conversations they had with their friends looking to use the database's powers to get an insight into their own tastes. They liken the resulting website to having a music-geek pal who can introduce you to your next favourite song.
Since it takes one to know one, I called up one of the experts I turn to when looking for advice. Omar Majeed is a film editor, music critic and aspiring DJ from Montreal who had already played with Pandora on his own. He wasn't impressed.
"I input all my favourite bands and it always ends up playing me something by Styx or Journey," Majeed says.
"Not that I have a big problem with that, but my cheesy taste is not what needs beefing up. I want to discover cool new indie bands, and Pandora just ain't providing."
As of last month, Pandora's pro-grammers started to address the fact that despite their analytical deconstructions, some songs just don't match up with listeners' preferences.
"We evaluate Pandora through listener feedback and our own exhaustive testing," said Westergren, "For some time we've noticed that occasionally we miss certain things. Even with 400 musical attributes, some songs escape our approach."
My first go at Pandora used Feist as a starting point. I was expecting the occasional saccharine singer/songwriter, not the power hair-ballad majesty of Cher's If I Could Turn Back Time.
To respond to this sort of concern, Pandora's database is now taking input from the way listeners choose songs. If enough people click thumbs-down for a given song on a particular radio station, its chances of reappearing on the same station for someone else are radically reduced.
Because of its roots in musicology, Pandora can always tell you why it makes each decision. For example, an offering by Morcheeba came up for me because it had "basic rock song structures, use of a string ensemble, mixed acoustic and electric instrumentation, and minor key tonality," which it turns out is just the kind of thing I go for.
However, this doesn't always work out.
"I think that in being overly scientific they miss an important aspect of music geekness namely, that the band's persona and approach are just as important as the distinct musical elements," says Majeed. "I.e., I like crunchy, fuzzy guitar in a cool band like White Stripes, but not in a band like Def Leppard. Pandora can't really quantify what's cool about music by breaking it down into such distinct parts."
In contrast to Pandora's approach, consider the All Music Guide ( www. allmusic.com ), a comprehensive encyclopedia of artists and songs that actually predated the Web as a guide for music shops in the early 90s. Its editors classify songs not just by genre, but also by mood and theme, in what we in the world of Web 2.0 (the second stage of the World Wide Web that is more community-driven) now call tags. Songs can be labelled anything from "boisterous" and "visceral" to "wry" or "bittersweet," and somewhere in that matrix of adjectives you get an idea of a sound. It's often these elusive qualities that can be the most useful in trying to express what music evokes.
So far, All Music Guide doesn't have its own station-building capability.
"I think each has its merits," says Westergren. "Different people like browsing music in different ways. Some like tags, some like the Amazon-style collaborative filtering. We have a very specific objective, and we think a musicological approach is the best way to do that. Over time, we'll incorporate those other features, like tagging and community feedback."
Now, if only there were a way to get a custom station that matches the tech talk of Pandora with the music-geek-speak of AMG. That would be a box worth opening.