Web Something.0

February 21, 2007 10:53 AM

iLike"provides a buddy-list for your iTunes - it helps you discover new artists based on what you're already listening to, and it helps you browse your friends' music libraries and share music suggestions with each other." Basically, there's an iTunes plugin (OSX only) that automatically sends your iTunes metadata to the iLike site to be shared with the community.
posted by Kwine (19 comments total)

... and if you don't like Ticketmaster, it should be known that the company recently invested a good chunk of change into iLike.

Personally, I like iLike.
posted by pfafflin at 11:04 AM on February 21, 2007

Works on Windows, too.
posted by yerfatma at 11:04 AM on February 21, 2007

So they hope to drive traffic to concert ticket sales? Is that the revenue model? How does the web side of the service compare with last.fm?
posted by grobstein at 11:09 AM on February 21, 2007

Personally, I prefer MOG.
posted by NationalKato at 11:10 AM on February 21, 2007

I never used last.fm, but I cannot get past the web design at MOG. I like my websites clean and MOG feels irrevocably cluttery to me.
posted by Kwine at 11:24 AM on February 21, 2007

I have an iLike account. It is pretty cool. What I like best about it is that I have to do almost nothing -- it gets all of its information from iTunes for me.

They also gave me a free, 4GB red iPod Nano, simply for using their site and the plugin (it was some promotion in December). My niece has it, since I already had an iPod, but that was super cool, nonetheless.

And they do have an iTunes plugin for Windows XP. So you don't need a Mac. You don't even need iTunes, strictly speaking, but the real advantage of the site is gone if you don't let the iTunes plugin work its magic.

They also provide links to free MP3 downloads of songs similar to the one you are listening to, if they are available (the folks that make iLike are affiliated with Garageband.com, IIRC).

I don't really know what their revenue model is. But it's an interesting way to check out music. You can find people that like stuff like yours, see what they have that you don't, and even listen to sample of lots of it.
posted by teece at 12:12 PM on February 21, 2007

I dig iConcertCal myself, although it's not quite related.

Is this collaborative filtering like last.fm as well or is it some other sort of algorithm I wonder?
posted by mkb at 12:13 PM on February 21, 2007

Recording what you actually listen to (Audioscrobbler/Last.fm) seems like a better approach than looking at the content of your library. It appears iLike does both, but I'm not sure how including tracks I never listen to will help generate good similarity data.

Plus, I get an icky feeling letting a corporation read a good chunk of my hard drive.
posted by D.C. at 12:47 PM on February 21, 2007

From their FAQ:
You can now re-order the "artists iLike" to select a different top-12 (some people felt their iTunes listening didn't properly reflect their tastes).

I interpret this as either: (a) They're catering to the myspace crowd that cares more about appearances than sharing accurate likes.

Or, (b) Their method of data doesn't work well, and they're getting user complaints.

I'm guessing it's more A than B.
posted by D.C. at 1:00 PM on February 21, 2007

. . . data collection . . .

gah.
posted by D.C. at 1:02 PM on February 21, 2007

Haven't we done this before? I'm fairly certain I tried this a few months ago after reading it here. (No time to search. I've always wanted to post "Double" dammit.)
posted by effwerd at 2:27 PM on February 21, 2007

It was in AskMeFi, in a comment.
posted by mkb at 3:00 PM on February 21, 2007

And I thought it was a shoutout for the wonderful ilike.
posted by peacay at 3:50 PM on February 21, 2007

I think I was thinking of this thread on audiomap but that wasn't it. Sorry for implying that this thread might be a double, Kwine.

I got rid of iLike, btw. Made my iTunes too unstable. It was pretty nice when I used it though. I found a few cool artists and tunes that have stood out in my collection since. But I found them from exploring other people's playlists, not iLike's suggestions, which never thrilled me.

I liked but also kinda hated that it reported everything I played. I didn't like that it would report repeated plays if I stopped and restarted playback. And I was making a few tunes in GarageBand at the time and thought it funny that my most played tunes were mine, played 25 times in a row!

I also tend to listen to new tracks I'm interested in a lot at first. That might not have been so bad had it been able to actually get real play counts from my library and use that to determine my favorites (or I gave it more time). I've done a lot of work on trying to find a good algorithm for determining my favorites in iTunes (via AppleScript) and I think what I've got works really well (at that, but it blows at keeping my site updated). iLike was a little frustrating in this respect.

Still, I might try it again in the future, though I'll give it time to clear up the bugs. Like I mention in the audiomap thread, Pandora has been the best for me. iLike introduced me to a few tunes I like. Pandora introduced me to music I love, and a lot more of it.
posted by effwerd at 3:57 PM on February 21, 2007

I really like last.fm. It's great for finding new music.
posted by cell divide at 4:02 PM on February 21, 2007

effwerd, I'd be interested in learning about your AppleScript iTunes favoriting algorithm. How does it work, generally?
posted by Kwine at 4:22 PM on February 21, 2007

effwerd, I'd be interested in learning about your AppleScript iTunes favoriting algorithm. How does it work, generally?

Heh. I worked so hard on it I have trouble understanding it myself (and I evidently suck at commenting my scripts). That week was a blur. Here's what I remember...

I use a simple points system and weight it. The most important factors are play count and track ranking since the basic question is how much do I enjoy each artist. Weight is added for being played in the last thirty days to (usually) handicap more recent artists over long established ones. If iTunes had a play date history, I could add some nuance by factoring in the number of times played in the last thirty. (I realize this would actually be silly to implement and maintain in the track data but I want what I want.) Tracks ranked above three stars get weight added to their play count.

Weight is added to the artist for having a high five star to total number of artist tracks ratio. The four star ratio is given much less weight but it gets a little something, as does five star to total number of library tracks. The zero star ratio is counted as a negative but not as negative as the one star ratio (since zero star shows a passive lack of interest while one star shows an active dislike). The disabled track ratio is negative, too, of course. Oh, and five star to total recently played tracks, that's in there as a positive.

The number of tracks an artist has in the library is given a little weight. It's already represented, for the most part, in all of the above but a large number of tracks by an artist tends to demonstrates, in and of itself, an active interest in that artist above the average so I give it a little extra. (This may not work for people who just collect and keep a wide variety of material regardless of any interest at all.)

I used to factor in the skip count and if the track was skipped in the last thirty days as negatives but then I got an iPod shuffle. I never bothered to see how much this changes the assessment. I didn't think it had enough weight either way and I wasn't really interested since I skip songs on the iPod for completely different reasons than I do at home in front of my computer, and it correlates not at all with how much I like the artist.

Standard disclaimers: IANAProgrammer. YMMV.
posted by effwerd at 6:56 PM on February 21, 2007 [1 favorite has favorites]

Interesting. I muck about with the 5-star system to create favorites lists in a brute force way now and again-I'll zero out the stars for each track, get 10 random tracks and pick five to advance to the next round by getting one star, repeat the process with all the one star tracks to two stars, and repeat again until I get a five-star list. I've often thought about creating an algorithm, but I actually enjoy the brute force decision-making. But one day I'll cook up something similar to your scheme.
posted by Kwine at 12:06 PM on February 23, 2007

A kind of musical playoffs? Nice. I like that idea.

What I described was just for determining a list of favorite artists. I haven't gotten around to using it to develop random playlist scripts, which is actually why I started it -- so I could weight the pool of artists to pseudo-randomly choose from (to then get a random or weighted track from that artist and so on). Ultimately, I'd like to wrap it all up in nifty little HUD palette but at my current leisurely rate, plus teaching myself Cocoa, that'll take about five years. Though writing up that last comment kinda got me excited about it again.
posted by effwerd at 5:15 PM on February 23, 2007