Duke Listens!: Visit my main blog at MusicMachinery.com

This  Wired article:  Fine Tune Your Music Discoveries  is a quick review of 4 popular music discovery sites: Pandora, Last.Fm. iLike and QLoud.  The author gets it a bit wrong when he says that all of these sites use collaborative filtering to recommend music - Pandora relies on content-based matching of content, not collaborative filtering.  Interestingly enough, the author liked the Pandora recommendations best. The author does point out the problem that collaborative systems have with popularity bias: Oddly, I found the more mainstream an artist is, the less precise your results are likely to be. The recommended artists similar to Neutral Milk Hotel, which is arguably the most left-field pick of the three, were almost universally more accurate across all four services.  I do agree with the article that the QLoud's Ajax-heavy site is a bit of a turn off.