DarwinTunes Iterates, Mixes And Culls To Create Listenable Music From Noise 53
Posted
by
timothy
from the soon-there-will-be-cover-bands dept.
from the soon-there-will-be-cover-bands dept.
Shipud writes "A collaboration between a group in Imperial College and Media Interaction group in Japan yielded a really cool website: darwintunes.org. The idea is to apply Darwinian-like selection to music. Starting form a garble, after several generations producing something that is actually melodic and listen-able. The selective force being the appeal of the tune to the listener. From the paper published [Monday] (abstract) in the Proceedings of the National Academy of Sciences: 'At any given time, a DarwinTunes population
has 100 loops, each of which is 8 s long. Consumers ratethem on a five-point scale ("I can't stand it" to "I love it") as they are streamed in random order. When 20 loops have been rated,truncation selection is applied whereby the best 10 loops are paired, recombine, and have two daughters each.' Note that in 2009 the creators of darwintunes harnessed the power of Slashdot to help 'evolve' their site."
Silence is golden (Score:4, Interesting)
Musicians also know that musical compositions benefit from the appropriate amount of silence between notes. If this algorithm were tweaked a bit to include some relative silence here and there, I think it would help the "listenable" factor. Take the final tune for example, and imagine a four-count measure that contained only one note or instrument (or even bass drum-like sound) playing eighth notes on beats 1,2,3,4. It'd create some anticipation, I think. This is the electronic equivalent of "white guy syndrome" -- too many notes!