MyMedia – a very deep recommendations engine


The internet is too big. There’s simply too much stuff on it. In 2006, an estimate of how much put the figure at 40 Petabytes, which is about 38000000000 short novels. There’s a lot more than that today, and Microsoft’s innovation labs are trying to work out the best way of sifting through it.

Recommendations engines can run in lots of different ways. You can look at people’s interests and try and match those up with content, which is a bit of a hassle. You can use the wisdom of the crowd to predict based on what other people have consumed next to each other, which is a bit better.

But what Microsoft is doing takes things a step further – they try and work out your mood, see what other stuff you listen to when you’re in that mood, and predict things that way. They also tap into your network of friends – who you presumably have common ground with – to try to work out what stuff you like.

As with any recommendations engine, the more you put in, the more you’ll get out. Microsoft doesn’t claim, therefore, that this is going to work perfectly out of the box. Instead, it’ll get better the more you use it, and ties in with Microsoft’s massive social network around Windows Live Mail (Hotmail) and Messenger (MSN).

The project’s still at an early stage, so it’s difficult to judge whether the company will be able to pull off something as effective as’s music recommendation engine. Let’s hope, though, that something comes off it, and the stranglehold that the media holds over mainstream taste can be eroded.