Musicians, take note: An artistic mind isn’t required to create appealing music. Starting with short sound sequences more grating than Muzak, scientists created pleasing tunes simply by letting them evolve through a Pandora-like process of voting thumbs up or thumbs down on each sequence.
Inspired in part by long-running experiments probing the evolution of bacteria, computational biologist Bob MacCallum and colleagues decided to see if pleasant music could evolve from a cacophonous mess when human listeners acted as the force of natural selection. The researchers started with a loop of simple audio wave forms and let it randomly evolve to generate a starter population with variation on which selection could act. Then more than 6,000 people listened to the audio loops and rated how much they liked the sounds on a five-point scale. The audio loops rated more favorably were allowed to mutate or combine with others to make a next-generation clip; the bad ones died off.
By 500 generations, the pieces developed into pleasant little ditties with chord structure and rhythm, MacCallum and his colleagues report online June 18 in the Proceedings of the National Academy of Sciences.
Now the researchers are running experiments with stricter, more realistic sources of variation. They also want to scale the project, called DarwinTunes, up to millions of users. “We may see a leap to a new plateau,” says MacCallum, who spends most of his time investigating mosquito genomics at Imperial College London. “Done properly, we reckon the quality of the music would be pretty much comparable to current man-made electronic and dance music, but a lot more democratic.”
