I was telling a friend about Pandora, a magical Internet music service that can tell what kind of music you like.
You have to give it a little bit of a push to get it going — you tell it an artist or a song you like, and then it starts playing more songs like that. As you go, you can tell it which songs you really like or don’t like, and it quickly adapts to keep you listening to music you like. Pandora knows which songs are similar because they’ve had professional musicians score each song on a wide variety of variables to fingerprint it. The method is much more precise than grouping by genres, or even by artists — sometimes an artist will play something out of genre, and Pandora’s method easily picks that up.
When I first tried it, I was amazed for good number of hours — it’s very cool. I was sure I was going to subscribe.
Pandora’s music database is pretty extensive, but I did start to hit the edges of it. The real killer, though, was that while it did a good job of playing a certain kind of music, it didn’t know how to vary the mood (what Muzak used to call “stimulus progression”) — which ended up being very fatiguing.
I abandoned Pandora, and looking for something with similar gratification qualities, played with last.fm for a while. Not quite as amazing, but it was still really cool to type something like “jpop” into the keyword search and just listen to a certain kind of music. I heard some things that were just okay, and some things I really loved, but in any case, music I would have never heard before.
I’m looking forward to when these services get a little smarter, like a human DJ, and can shape a progression of mood, and also know how to surprise and delight me with things that aren’t the same as before, but that I still like.
(So I can keep track: adapted from a post to Words-L, “Re: Have we talked about…”, 2006-01-24.)