Features | Jericho

Jericho / VI :: Crowd-Sourcing

By Joel Elliott | 6 July 2012

The BBC News covers it with the headline “Is this the End of the Composer?” and creator/evolutionary biologist Armand Leroi insists that it is. This is DarwinTunes, an attempt to apply sweeping evolutionary principles to the creation of music. The concept is incredibly simple: a computer generates random noises in eight-second loops, like Stockhausen’s mutant children, quickly killed off by the blunt options given to the various users (“I love it!” “I hate it!” “It’s okay!”). Sooner or later, after a few hundred generations of random strains, the computer inputs something with enough of a pattern to be tolerable to the general public and the loop is allowed to breed. After a few thousand generations, the result is something Leroi compares to “chill-out electronica,” or more humbly “something a little better than a nice ringtone.”

While there may be hope for the program yet, no one will confuse the results for something created by a human being in 2012; ironically the music doesn’t reflect any of the technology that presumably made it possible. Think the kosmische experiments of the ’70s with the sense of minimalist precision and momentum replaced by loops which succeed each other with casual indifference; admirably complex but completely empty. Co-creator Robert MacCallum even admits as much: “The evolution led to pleasant, jingly tunes that didn’t offend anyone but didn’t really move anyone, either.”

Nevertheless it’s an impressive accomplishment, at the very least illustrating that rhythm and musical complexity isn’t solely an intuitive process. But it’s probably more interesting for its limits: in the study the authors document a plateau wherein the new generations of loops cease to become more aesthetically pleasing as they mutate (in pairing the newer loops with the slightly older, the more “advanced” generations didn’t score any higher in listener response). The authors chalk it up to a complicated biological process that more or less translates to: the more complicated, and finely-tuned the loop combinations become, the more vulnerable their appeal is to being split apart through “mating.” Increasing mutation ruins the appeal of more complex constructions.

That conclusion at least hints at a fact that common sense would already suggest: the lovechild of Bach and John Coltrane, for example, would not necessarily be beautiful, and one can assume the same is true for their imperfect, embryonic forms as well. Cross-pollination may provide the seeds of innovation, but selectively, which is why music discovery programs like Pandora always seem frustrating to dedicated crate-diggers (and why former CMG writer David Greenwald created a service dedicated to restoring the human element in music curating).

A more likely scenario, since we’re not dealing with Bach or Coltrane, but always and forever “chill-out electronica” amen, is that rather than become more catholic in its reach the program will conversely become more limited in an attempt to satisfy (if only slightly) the greatest number of listeners. At the moment this may be because of a limitation on the number of sounds and parameters, but it also makes sense on a basic, conceptual level, which is why creating a real masterpiece is probably more than just a matter of making the program more sophisticated. Even most pop music is too divisive to please everyone—especially the undergraduate students who were the initial test subjects of the study—more than something decidedly less remarkable but difficult to object to.

The authors do show some awareness towards the end of the paper that music (or culture as a whole) doesn’t work exactly like evolution, anymore than consumer choice determines music production in any kind of straightforward way. They acknowledge that listeners are affected by other factors, including the preferences of others. That last point seems like it belongs somewhere other than a final caveat, and that’s not even primarily me speaking as an embittered tastemaker. What about the role of producers who decide what should be popular before it even is? Or the entire top-down v. bottom-up argument that rages in all histories of technological and cultural innovation? What about the extent to which listeners adapt and even grow to love certain kinds of sounds over time that they might otherwise find irritating in an eight-second loop? What about musicians who actually create the music and are free to respond both to and against popular taste? What about how fringe art affects the mainstream but bypasses the ordinary listener? Was it necessary for the broader public to approve of Stockhausen before they approved of the Beatles? Did millions of people click an “I Love It” button on Messiaen before they bought Kid A (2000)?

On one level, DarwinTunes is of a piece with other recent new media sound works—acknowledging a certain relinquishing of authorial control in the face of technology that ultimately determines the endpoint. And yet works like Jem Finer’s brilliant 1000-year composition Longplayer or Ryoji Ikeda’s “data” works find the inherent patterns and quasi-biological growth within technology and information itself. In comparison, Darwin seems bound to merely be a metaphor (albeit a fascinating one) for an evolutionary process which is itself restricted by the vast differences between the reproduction of organisms and the development of culture. Even in “real life” the role of the audience in the growth of music is far more multifaceted and complex.

As an interactive work, DarwinTunes abstracts the user’s input. We’re contributing something to the end product but it’s impossible to really perceive what, making the process seem frustratingly anonymous at times. I wrote previously in this column about Zack Settel and Mike Wozniewski’s Audio Graffiti (the “tagging” of sound by participants in a virtual space), a work that sits on the other extreme. That piece was seemingly boundless, embracing the ephemeral, celebrating dissonance and placing the user’s contribution front and centre, at least for a brief period. DarwinTunes, by comparison, erases difference and reigns in the more chaotic strains of individual taste.

Crowd-sourcing can be a powerful tool though, especially in cases where rare ingenuity can create measurable results. Medical researchers recently created a video game where users construct realistic proteins out of amino acids in order to brainstorm potential breakthroughs in fighting various diseases, using both the ability of technology to facilitate change in accessible ways, as well as the human ability to manipulate one’s environment in ways beyond the scope of any software. Likewise, Avaaz and Moveon.org provide future models for participatory democracy, bypassing bureaucracy and channeling political will in constructive ways.

Nobody expects a program like DarwinTunes to be that useful, and yet the flip-side is simply market research. At its worst it reinforces an alienated populace, like the “soma” of Huxley’s Brave New World, creating an illusion of infinite liberty that precludes conflict or action. At its most harmless, it’s like the occasionally baffling “targeted” ads on Facebook. George Prochnik, in his book In Pursuit of Silence, profiles the world of retail sound-design, an industry which might, if it hasn’t already, soon catch on to the developments of DarwinTunes. Both are less emblematic of popular music per se than they are of a general, positive, uplifting feeling, a quasi-utopian comfort that, in the case of retail outlets, encourages impulsive shopping. Prochnik calls it a “means of creating a connection,” but rather than one between people, it is one “between the individual and a state of group ecstasy.”

Consequently, it may be fortunate that eventually DarwinTunes always plateaus, never infinitely becoming more pleasurable (if such a thing is even possible). In its current incarnation, a few thousand generations old, it’s a bit of an uncanny valley: too close to perfect to be beautiful, like a slightly overdone plastic surgery job. In that sense it reminds me a bit of James Ferraro’s Far Side Virtual, although Ferraro (coming from a background in noise music) is far more keenly aware of how to engineer an almost nauseous level of pleasurability. Like that album, DarwinTunes is far too complex to fade into the background, but too innocuous and anonymous to command complete attention, elevator music for a hyper-accelerated culture. It’s as much noise as those original raw, abrasive loops, both by-products of the rush of information, consumption, and technology.