> Looks like if we describe musical work with high dimensional vectors, I like to find good vectors that are not too close to each other.

It's the classic problem of "Ok, but how do I sort these by which ones are actually any good?!"