[snarfy]

I always thought this idea could be greatly expanded upon.

I've seen it used to guess the native language of a text file based on the compressed input. I always believed this could be used as a sort of universal translator. You could compress the audio sounds of birds, throw this algorithm at it, and extract meaningful content.

[woliveirajr]

To tell the truth, it already has been proposed and there are some papers about it. Search for "NCD - Normalized compression distance" and you'll get some results. It's even used to verify who is the author of some document (authorship attribution). Very interesting.