Sunday, February 05, 2006

Pandora In The Hot Seat

A week or so ago, Lefsetz sparked a very lively debate about MySpace. This week, he set his sights on the Music Genome Project’s Pandora, a site that I recently compared with the slightly more human powered To those late to the table , both of these sites use collaborative filtering and advanced algorithms to recommend new music to a listener.

A recent addition to the whole “discovery through mathematics” set is Muiso, who uses an Audioscrobbler like software plugin to track listening habits on a users machine to create a personal playlist that it compares with that of your musical neighbors. (As it doesn’t run on a Mac yet, I have yet to actually test it, but I’ll get to a Windows machine at some point this week.)

We can apply Bob’s critique of Pandora to the other sites as well. Here’s the crux of his argument.
  • Pandora is a scam and not about turning you on to new music, it’s about advertising.
  • Pandora does not give you useful results..
  • Machines can never replace the good ol’human being.
I disagree on all of these points.

First, Broadcast radio isn’t about the money???? A certain Mr. Spitzer would disagree. Who cares if I have to look at ads if I get quality recommendations? If it worked well, I’d even pay for it (as I have with There is no doubt in my mind that these types of services are ultimately going to drive traditional music radio to an even more marginal status. In fact, if Infinity, Clearchannel and other media giants had any brains whatsoever, they’d be already offering big money to get into this market.

It's no secret I’m a huge fan of Last.Fm. Their model (and seemingly that of Muiso) isn’t based on a “music genome” like Pandora. Their little web machines look at what other people who are already listening to similar music are listening to, and presents an individual with some new recommendations. I’ve actually made several purchases based on what I’ve heard.

I’ve had a decent discovery experience on Pandora. The problem Bob found with it is that they focus only on sound, and not any sociological reasons why people get turned onto music. There may be some stylistic similarities between The Banner and Kreator, but the social scenes (hardcore vs. metal) where these two bands come from, are pretty different when you look at them from within. I also don’t think Bob gave their internal filters enough of a chance to hone his taste to it’s maximum ability. Finally, the Pandora music database isn’t developed enough yet, and as they add more music there will be more viable options to recommend.

I concede, there are indeed duds in the mix. But far less than you’ll find with traditional radio. As more data is added, and the internal filtration / artificial intelligence is refined- near perfect niche, personal radio will be achieved.

As someone who never found new music from the radio, I welcome these services with open arms, and look forward to seeing how they develop.

1 comment:

steve said...

i found out about pandora very early on, and almost no bands i requested matched up with decent suggestions. i think now that the database must've grown a little bit, i'm going to revisit it, and see if it gets any better. i've heard some pretty good success stories, so maybe i'll find a few hidden gems.

as far as goes, i agree. it's a great site, even if it's just to check out my own personal stats every now and again. i've yet to really find myself getting lost on it, though, it is interesting to see a few people's pages who end up being your neighbors and making certain connections. when you see someone else's recent plays and get confused about why they'd even be related to you, and then you look at their broader stats, and it starts to make sense in ways you never would've considered.