Peter Gabriel, Unfiltered

Back in the day, I was a rabid Peter Gabriel fan. Through his Genesis days and his first five solo albums, he ruled; his concerts were a transcendent experience. His music tailed off after that, as did my interest, though his Oscar-nominated collaboration with Randy (I Write A Song For Every Second Movie) Newman on That’ll Do still gives me chills. Look, I’m a big enough fan to absolve him of the responsibility in the creation of the tribute to the paucity of taste in the world that is Phil Collins.

So I was thrilled to get an excuse to write about him in the form of the launch of The Filter, his online project that uses complex algorithms and your input to recommend music, movies and television programs that’ll appeal to your taste.

(Although, I’m sure an interview would have gone badly:

“Peter Gabriel here.”
“Peter Gabriel? You’re really Peter Gabriel?”
“Yes. Mr. Webb, I understand you wanted to learn a little more ab-”
“OMIGOD!!! IT’S PETER GABRIEL!!!”
“Yes. Now, let me tell you a little bit about Bayesian mathe-”
“EVERYBODY!!! PETER GABRIEL’S ON THE PHONE!!!”)

The Filter, as Mr. Gabriel was tring to tell us before he was so rudely interrupted, uses Bayesian mathematics to sort through the tripe and find the stuff you’ll like. According to the Web site, “the engine uses an evidence model (which includes purchase, consumption and browsing data) to derive the similarity of items.” Given an item of interest, it produces a pick-list of that are relevant.

As you rate the recommendations, the engine has more and more information from which to derive its lists, ultimately, in theory, knowing what you’ll like before you even know it exists. Once you’re signed up, you can fine-tune by rating other genres and artists.

All good in theory (even if the theory is a little beyond me). In practice, though …

The sign-up process is quick and painless (although the simulated parquet background on the home page shocked my monkey): Choose three genres of music. You’re then presented with three artists in each genre to rate; a handy preview function lets you have a listen if you’re not familiar. Do the same for movies (or don’t). And you’re away.

The recommendations, on the whole, were spot-on to a degree. I’m a picky eater when it comes to music (no Phil Collins, please; I’m allergic). Many of the CDs recommended were already in my collection. Most of the preview music I heard suited me fine. And over time, one supposes, as I rated more and more tracks, the recommendation would have been even more finely honed. With the optional downloadable program that tracks what you’ve played in iTunes or Windows Media Player, the results should be even more relevant, unless you like listening to music that you don’t, um, like.

But then what? The Filter has no record company deals, so it can’t sell you music online. The previews are 15- to 20-second clips; by the time you’ve heard enough to rate the song, it’s moved on to the next. The streaming quality was iffy (probably high traffic in the wake of the launch) and the sound quality and volume varied wildly (but then again, you have to use the volume control on any stereo when switching from Bill Evans to The Killers). And I can’t abide the privacy implications of the download — though the company promises they don’t do anything else with the information, that’s no promise that a court order couldn’t break.

If it actually functioned as a streaming radio player that tunes itself to your taste, it would be worth the buzz. But as it stands right now, it’s not terribly useful. While it’s a different (and arguably more accurate) engine to play arbiter of your musical taste from the “other-people-who-like-what-you-like-liked-this) model of, say, Yahoo Radio, at least that plays full songs.

Sorry, Mr. Gabriel. I wanted to like it.

But to be fair, it is in beta, and surely more features are coming. Keep it on the radar.