The likes of McCartney, Springsteen and Dylan could soon be composers of the past as a team from the Imperial College London have programmed a computer to write the ‘perfect pop song’ by adopting the laws of natural selection. The system, known as DarwinTunes, writes tunes from random noises and would seem to indicate that you don’t need a composer to make music—it is all down to mathematics.
The “evolutionary music engine” works by generating sound loops and then responding to listener feedback. The team from Imperial College London began with 100 loops of music, each of which lasted eight seconds. An audience of over 7,000 online participants were then invited to score the loops in batches of 20 on a five-point scale from “can’t stand it” to “love it”.
From the feedback, the computer program then combined the top 10 loops, pairing them and mingling their musical elements to create 20 more “offspring” tunes. This entire process was repeated again and again, until the worst rated loops became ‘extinct.’ In true evolutionary style, the most popular loops were then allowed to reproduce.
“We don’t often think of music as evolving, but everybody knows it has a history and it has traditions. But if you think about it, it really has evolved, it is changing continuously.”
“There are all the same forces of change, variation, selection and recombination as different musical traditions join together, transmute and fuse and divide again. This is all the stuff that is familiar from our understanding of the biological world, but we see it here in music as well.”
“We believe music evolves by a fundamentally Darwinian process—so we wanted to test that idea.”
According to reports, DarwinTunes has evolved through 2,513 generations with surprising results. DarwinTunes programmer Dr MacCullum said they started to notice “a kick drum or a bass drum, and that just spontaneously came, we didn’t put any drum sounds into the algorithm.”
MacCullum has noted the program’s limitations however, saying it doesn’t take into account the influence of peers, but believes it could have a huge impact on the music industry.
“I’ve no doubt that if we ran this experiment for longer, using bigger, faster computers, and millions of people rather than thousands, and for years, instead of months, we could evolve fantastic music.”
The entire report has been published in the journal Proceedings of the National Academy of Science (PNAS).
So what do you think? The next evolution of music or computers sticking their noses into what should be a purely creative endeavor?