Hit music formula

Most interesting. I can certainly see where they're coming from. I have no doubt that there are unique combinations of the criteria they listed (Melody, Chord progression, etc) that are intrinsically pleasing to the human brain, and groups of songs that exhibit these same characteristics may be similar on a mathematical level. Music and math are very closely related, and the human auditory system is very sensitive to this, even if the listener isn't aware of it on a conscious level.

I am also surprised at how much the image of the music universe resembles pictures I've seen of the structure of the actual universe, with superclusters of galaxies with voids between them.

Jam it back in, in the dark.

Quote:

In other words, a machine won't be able to discern whether a song which has no relevant data (a new kind of music) can be a hit or not.

I'm not so sure. In some language theories, it is posited that there is a sort of "metalanguage" that is programmed into the human brain that allows a multitude of languages to be built from this preexisting framework. What if something similar exists for music? (I believe that something like this does exist). If the parameters of the universal musical language could be determined and programmed into a computer, you could indeed have a computer that could not only determine the worth of any type of song, (even songs from genres with small or nonexistent data pools) but could itself compose exceptionally pleasing music, as least as far as things such as instrumentation, melody, and whatnot are concerned.

Perhaps this universal musical framework could even be determined by analyzing and compiling all of the positive mathematical aspects of all of the songs currently in the musical universe.

Quote:

For instance, how would a machine determine that the snare drum is cancelling out the melody in the b section? Or perhaps that the high hats and the bass are too up front in the mix. In regards to fleshing a song together, these are parts of the equation that I simply can't see a computer being very discerning.

The computer could perform a frequency analysis of the song to see if any particular band of frequencies dominate any other. If the computer knows the frequency components of the individual instruments, it can locate the problem and adjust the amplitude and/or frequency response of the instrument in question. Ideally, a song's frequency readout should have a gradually falling characteristic, resembling the readout from a pink noise graph.

There's nowhere I can't reach.