A few weeks ago I read an article in the New York Times about a computer program that can “optimize” faces for beauty. It measures and alters symmetry, shape and average distance between different facial features. The program is really remarkable because the improved faces were unmistakably similar and arguably more attractive.
Interestingly, it mentions that there is no perfect face, but I assume this is only because individual faces can only be modified to a certain point before they lose semblance to the original. The generation of a perfect face should be possible because this program proves there are universally accepted characteristics of beauty.
Following the concept of this program, I’m almost certain that “perfect” (universally appealing) songs can be generated. After all, music is nothing but patterns, there are probably some similarities in the top songs, all of which can be input into a program of some sort. After inputting the characteristics of more songs, the program reaches a point where it can randomly generate music that is “perfect”, or universally appealing to the majority of people, say 99%.
Some current variables I can think of right now are just the basics such as genre, tempo, instruments, time signature, syncopation, vocals, vocal gender. Pandora, Zune Mixview and Apple Genius all suggest music, but none of them generat music.
When I have the time I will definitely pursue this idea in the future. At the moment, there’s too much physics to study.