You won't have to pick your music because you think that Britney Spears is hot, predicts Paul Lamere, a principal investigator with Sun Labs. But you still might get it because your computer know you think her music is good.

Lamere developed a technology that makes recommendations for music you might like, based on an analysis of the existing music in you library. Instead of using buying patterns (Eg: shoppers who bought Britney spears also bought Beethoven) the software bases its analysis on the actual song, looking at features such as rhythm and beat strength. It will then find songs with similar characteristics.

The software for instance lets you pick two songs you like and then tracks a path from point A to B. Or it will just find songs similar to that one favorite of yours.

In the picture below the software plotted a path from rock band Metallica to Jazz artist Dave Brubeck in 15 steps.

The demo used a collection of 10,000 songs, but Lamere expects that our music collections will grow to about 1 million songs in the future as storage technologies continue to advance.

The second picture shows a map of all the music in the demo, mapped out on a grid. The prepresents the path that the software took to get from Metallica to Brubaker.

29 Apr 2005