Music Evolves from Noise

Isaac Wynn

Scientists at London's Imperial College, headed by Bioinformaticist Robert MacCallum, developed a computer program they call DarwinTunes to show “how cultural dynamics can be explained in terms of competing evolutionary forces.”

The program produces 8-second clips of randomly generated sounds, which, in response to user ratings of likability, either move on to “reproduce” with other successful clips or disappear from the database. Each “generation” of songs begins with 100 clips, then chooses a random selection of twenty to present to the listener.

The remaining 80 as well as the top rated half of the 20 presented songs survive into the next generation, while the lowest rated 10 songs of those presented “die” and are removed from the gene pool. The 10 survivors are randomly paired and produce four offspring to replace themselves as well as the eliminated loops.

The initial study involved 2,513 generations of loops evolved under the preferential pressures of 6,931 consumers rating the loops’ aesthetic qualities, although the program is now available online at www.darwintunes.org and the database continues to evolve. MacCallum and his colleagues hope that by uploading the program to the internet eventually millions of users will participate, dramatically increasing the evolutionary rate and thereby generating increasingly appealing music, as has been the trend so far.

"Evolutionary processes in DarwinTunes" courtesy of Robert M. MacCallum Matthias Mauch, Austin Burt, Armand M. Leroi/Proceedings of the National Academy of Sciences.