Last.fm uses collaborative filtering… in other words as you listen to stuff they keep track of all the artists. Then they match your listening to that of their other users and fill in the gaps. This is an interesting approach because you can get playlists that have vastly different sounding artists in them but that statistically appeal to the same audiences.

We’re a bit of a hybrid… we do a lot of data mining of user playlists, we also have categorized artists by ear (but certainly not to the degree that pandora does)… our data is not perfect yet but it gets better every hour of every day as people listen and make playlists.

I’m much more inclined to check out a band that I keep seeing/hearing on my friends playlists then something I hear unprompted because it has an upbeat tempo and deep male vocals.

Ultimately I think a better measure of how much you like something is how often do you listen to it… not whether or not you give it a thumbs up or down. There are plenty of things I think I like but skip over almost every time they pop up in a playlist.

Hope you come back and give us a shot.