When Genius Isn't So Sharp

I'm no expert on artificial intelligence ... heck, I'm not even sure that's the applicable field here.  But one of the greatest challenges for technology that I can think of is matching items that are full of subjective "data": music, people, etc.  (I would have placed visual art into that category too - but it's amazing how far image recognition has come in the last few years.)

For the sake of my current profession, it's a good thing that "people" will likely be the last nut that matching technology cracks.

Getting back to music, I've been playing with the "Genius" feature of iTunes this morning.  Bottom line: there is still a way to go here.  There have been services out in the market for far longer - iLike comes to mind - and none have really completely solved the problem.  For example: let's take one of my cheesy seventies tunes currently on heavy rotation: "Let It Ride" by Bachman-Turner Overdrive.  Another blogger described this music as "big dumb rock" that is undeniably "catchy".  I can't describe it any better than that.  In fact, I'd love to have nine more songs just like it.  But no service to date has really given me good answers.

The wisdom of the crowds - social data that is generated from other people's playlists and listening habits - can be helpful, but at the end of the day, music tastes are so subjective that the match is not completely precise.  I get a lot of recommendations for random seventies hits this way, but that's not necessarily what I am looking for.

Technical data - the actual notes - could also be useful, but the problem can be summed up this way: any band could learn to play the music of "Let It Ride" and it still wouldn't quite capture the actual "feel" of the original.

So I am left playing "Let It Ride" on repeat - neighbors be damned - until something better comes along.

Technorati Tags: music recommendation, iTunes, iLike, Bachman Turner Overdrive

Comments

[Bryan Scott]

Have you tried Pandora? Their recommendation system is guided by music professionals who categorize songs based on hundreds of fundamental musical attributes (e.g. use of minor / major key tonality, verse / chorus relationship). You simply sit back and give a thumbs-up or thumbs-down to each song. Pandora identifies your listening preferences and presents similar music.

I'm a heavy user of Pandora and quite impressed with some of the recommendations it's delivered. I've legitimately discovered new artists via Pandora.

http://www.pandora.com/