Rob Young

Apple recently announced some musical updates at their September keynote – most notable were some updates to the current range of iPods, and also the unveiling of a new version of one of their most popular pieces of software, iTunes 8.

iTunes

I used to be very opposed to iTunes three or four years ago (as was I also opposed to Apple computers in general) on the grounds that I thought that it took it upon itself to convert all your music to Apple’s .aac format. For this reason I ended up using software like MusicMatch Jukebox, and also even not using any software to “manage” my music, just software like WinAmp or foobar2000 to listen to it.

When I did finally switch and started using OSX instead of Windows, it was almost inevitable that I’d have to at least take a look at iTunes, if not use it full time. Not only did I discover that my concerns about having all my music converted against my will were completely unfounded, but also that actually, it did a very good job of ripping, tagging, playing and managing the file structure of my music behind the scenes. So I’ve gone from shunning iTunes like a bad disease to proactively recommending it to people regardless of whether they use Windows or Mac OS. It’s fair to say that I can’t really recommend it to many though, as most people have already got it given the popularity explosion of the iPod, and in reality I’m the one who’s arriving late to the party! (albeit having heard the dull noise of the music from the party from the end of the road ).

Musical Annoyances

There are some things that have always annoyed me slightly about my music collection. The main thing is that I’m really rubbish with music genres. People will ask me what genres of music I like, and I can respond only with a list of artists I like, I just seem completely inept at drawing associations between the two! This inability to be a music critic destroys any hope of creating any playlists in any music player grouped by musical genre, which is just something I’ve grown to sigh at and forget.

Thinking “Are my albums tagged okay?” and “Have I made sure I’ve ripped music to my computer in a sufficient quality that I may differentiate the output to that of a shower radio?” is all well and good, but when all the organisational compulsions have been satisfied, what you really want is a great music player. I want to walk up to my computer, give it a minimal amount of information, and have it do a really good job of DJing, intelligently choosing things I want to listen to.

This raises the issue of how your music manager gets to learn what you like. Until now you’ve had to rely entirely on the input of which songs you listen to often, devising a complicated system of playlists in iTunes saying “don’t play me this, I’ve heard it a lot this week” coupled with “I clearly like this, and I’ve not played it in a while” in order to get an effort-free (hit and miss) DJ. This all falls down if you’ve not yet used the software for the amount of time for it to have amassed listening data to the level where you get any form of variation in your music.

The second thing which has annoyed me about my music is that I’ve noticed that there is an absolute mass of music in my library that I really just don’t find myself listening to that often. This makes weeding out the stale rubbish really hard, and also renders the task of the computer doing a good job of DJing your collection for you much harder… until now!

Apple’s Genius

With the release of iTunes 8, Apple have introduced a feature named “Genius”, which automatically generates a playlist from your library of songs that “go well” with a given song. When you install it and activate the service, it gathers some data about your library, submits this to Apple for processing, and returns with information of associations within your collection! All that remains for you to do is to select a song from your library and select Start Genius.

I was really sceptical about how well this would work, but on the whole it does a great job of finding music of a similar mood to the seed song you give it. This solves the problem of my computer DJing me similar music and also the problem of areas of my library being left to stagnate! If I genuinely don’t like something it’s picked me from the Genius playlist I’ve just generated, I’ll skip it. More skips mean less listens, which leads me to flag it as something I probably shouldn’t have in my library and delete it!

Because it worked so well and I’m a geek, I spent quite a lot of time scratching my head about how it might work behind the scenes. I thought that they couldn’t have had much sample data to work from, and they couldn’t possibly guarantee listening habits for the amount of users they have with no gathering of sample data… Of course they have the data needed to drive this thing! The iTunes store has over 8 million songs and as of June 2008 it had made over 5 billion sales… !

The new iPod Nano (and the updated iPod Classic) can create genius playlists on the fly which is pretty cool, but you can create them as regular playlists from iTunes to use on older iPods. I wouldn’t be surprised if Apple released a software update for older iPods to enable this, where the hardware was right.

The only potential downsides may be that you require an iTunes store account to use genius playlists, and also that it did take quite a while to let iTunes do its thing with regard to analysing my library of 8000 songs and download the 80MB file it needs to perform the magic.

last.fm

Of course, iTunes is not the first time the internet has seen a music recommendation service. Pandora was the first one I heard of and used, but it was very restricted by the record companies – and indeed now I can’t even get into their website listen to their online player due to restrictions.

If you asked someone for a music recommendation service today, the name most people would blurt is last.fm. What we see today of last.fm was produced as a result of the merging of two websites, Audioscrobbler and last.fm. Audioscrobbler started as a computer science project by a student, and allows people to upload data about what music they’re listening to. Over time, this data can be drawn together from multiple users and common tastes analysed, and predictions made as to what music they may like based on common tastes with other users. Pandora and last.fm differ from this approach in that they obtain the input that drives their recommendations from users listening to recommended playlists on the web, and providing an “I love this, play me more things like this!” or “Nope, I hate that” response.

When they merged in 2005, they became a very powerful musical and social revolution, connecting people with similar tastes together, and providing people with a great way to connect with artists they would probably like but didn’t know existed! It’s also pretty cool to take a look at your listening statistics, and also listen to your very own radio station of recommendations wherever you are, based on your listening data!

Apple is clearly now in a similar position of being able to, over 5 billion song downloads later, offer a good recommendation service in its iTunes genius playlists and store. So is last.fm, with years of listening data having been scrobbled from people’s iPods, iTunes and other music players and software. The only real difference I can imagine between the two is that one is based on purchases and the other on actual listening data, making the safe assumption iTunes isn’t gathering data about listening habits. This means that iTunes also disregards music that is being listened to, but not necessarily obtained through them or any other legal means, for that matter.

I’m not really sure which service provides more accurate recommendations, but clearly both have quite good data on it. As illegal music downloading is something that nobody can really dismiss as non-existent,
it’s a real shame that iTunes haven’t got an API to play with; a service combining the power of both would be almost unstoppable, even if a bit slow given last.fm’s 10 requests per second limit.

LastGraph

LastGraph is one of the coolest things I’ve come across in a while – it takes your scrobbled last.fm data and lets you graph your changing musical tastes over time. This makes a lot of sense – it’s how people discover and listen to music and to find a way to represent it visually is pretty good!

It’s the creation of Andrew Godwin as a personal project, written in Django, Python (with his own Graphication library drawing the graphs) and using Amazon S3 for the storage of generated music posters.

Aside from the main attraction (for me, anyway) of creating a timeline poster of your musical taste changes, you can also export all the data from last.fm and do some other nice stuff

You can find me on last.fm here

Happy graphing…!