ITunes 8’s Genius Playlists Aren’t as Smart as They Could Be
The newest version of iTunes was released Tuesday. Considering I bought in to Apple’s music software/hardware ecosystem long ago, I downloaded iTunes 8 right away and started messing around with it last night. I recommend going straight to Apple’s iTunes download page rather than relying on iTunes’ or Mac OS X’s software updater.
The OMG feature for most users is the inclusion of a new music recommendation engine. Not unlike what we’ve seen before in services like Pandora and iLike, the new iTunes Genius playlist feature promises to construct playlists around songs you like, selecting music from your library that best compliments any single track of your choosing. You just select a song in your library, click the Genius button in the lower right corner, and a new playlist is generated.
This recommendation engine is curious to me, since I listen to a lot of very, very obscure stuff. I’m not trying to brag or anything, that’s just the way it is — music recommendation engines haven’t ever worked as advertised for my tastes. Pandora comes closer than most, but I find myself switching to something entirely of my choosing within a half hour or so.
Check out Eliot Van Buskirk’s extensive iTunes 8 review over on the Wired blog Listening Post. An excerpt:
Once you opt in to the service, which sends your music library and usage data to Apple anonymously according to the company, iTunes can generate a list of music in your library that sounds similar to any song in it, from which you can make a playlist. The Genius sidebar performs essentially the same function, except that it draws recommendations from the iTunes music store. Those recommendations include similar songs, top albums by the artist, songs by the artist missing from your collection and iTunes Essentials collections that include the song.
Genius relies on information and song data that’s partially culled from the iTunes Store. So, if a song isn’t sold through iTunes, will it still lead to any close matches?
Eliot liked the recommendations. I trust him, but I read up a little bit more on Apple’s Genius feature and found this experiment by Tim Moynihan at PC World. Tim managed some close matches with songs by rock bands like Wilco and The Pixies. But consider his results when he plugs in a few songs by the Beatles, who are famously absent from the iTunes Store — he gets a string of errors. Genius also hiccups when he tries a Kool & The Gang song that was actually purchased through iTunes.
My own tests, on my work machine with about 12GB of music, showed some equally confounding results. ITunes paired “Nude” by Radiohead with three tracks from the new Coldplay disc. Appropriate, but nothing else made sense: some rocking Steely Dan, The Specials and no less than three Rush tracks, including the epic 12-minute masterpiece, “Xanadu.” Similarly, suggestions to go with one of Robyn Hitchcock’s latest were split about 50/50 between hits and misses — including another appearance of “Xanadu.” I love me some Rush, but how does that equate with the poppy British singer-songwriter vibe?
As expected, a great deal of my music threw errors. If Apple doesn’t sell it, Genius can’t suggest a match. Sorry, Sun Ra.
So the Genius system is not perfect. It’s young, and its reasoning relies on the collaborative contributions of the masses. So, one can assume it’s going to get better over time. Maybe.
What are your experiences with Genius playlists? How does Apple’s implementation stack up to the other music recommendation software you’ve used? Let us know in the comments.
