Tech Blog

[#1]

Ever wonder how iTunes Genius recommendations work? Their suggestions and playlists are a black box no more! Well, let's call it a "dark gray" box, because we don't have all the secrets, but some insight into how the system works. Essentially it works through sharing and multivariate statistics.

The first part of Genius is when you send your info to Apple. The Genius algorithms look at the types of songs, the artists, how often they are played, etc and then compare them with other users' data sets. As you can imagine, as you start to poll lots and lots of people, patterns emerge. For example if someone has a lot of Eric Clapton and plays it in heavy rotation, what do other users who play music similar to that also play? The stats come in (and believe it or not, they are using a lot of the same techniques I used to reconstruct past climates an ecosystems in my Master's work) by giving songs and genre's value in a multi-dimensional spatial plane. This is really how the patterns start to emerge. You see people who have this mix of music are similar to other people, and even better who is not similar at all.

It's all about the patterns.

I don't know about you, but I rarely make a playlist from scratch in iTunes any more. I just pick a song and make a Genius mix. Hey it might not be perfect, and sometimes odd songs sneak in there, but I get a fast 100 songs to sync into my devices to listen to. Oh and at first the Genius mixes for classical music were rather bad, now (because of so much additional user data) I find them very good.

What do you think? Making a playlist with iTunes Genius, awesome or awful?

For more in-depth (and the math behind it) info Technology Review has a good explanation of what's really going on .

[#2]

I just wish it wouldn't pick Christmas songs in the middle of summer

[#3]

No kidding! I have BLue Christmas by Diana Krall stuck in a Jazz mix. You can manually remove them by unchecking the check box next to the song.