the scientists who mistook music for darwinism

There’s been a good bit of coverage of and interest in a study published in the Proceedings of the National Academy of Sciences with the provocative title “Evolution of Music by Public Choice.” I’m always on the lookout for research like this because I’m fascinated with what music is and how people’s tastes are shaped, but the more I looked into this effort by scientists from the Life Sciences department at Imperial College London, the more upset I became – with its premise, its methods and its sloppy use of language. The authors are enjoying their moment in the media, but they’re spreading a conception of music that’s inimical to creativity, artistry and human connectivity. In other words, to music itself.

The basic facts are HERE and the study itself HERE. In brief, computers generated random sounds in 8-second loops. Thousands of on-line volunteers voted for the more appealing of pairs of loops in A/B listening comparisons. The nicer loops were “mated” so that the appealing qualities were passed on generationally. And the noise became something resembling music. Hence the name of the project: Darwin Tunes. A sampler of the changing sounds generated in response to voting feedback and recombination of winning loops is at the bottom of this post, and they are marginally interesting.

The study may have some limited findings and applications, but the media is making it sound far more significant than it is. Here’s one lead from a serious science website: “Do away with the DJ and scrap the composer. A computer program powered by Darwinian natural selection and the musical tastes of 7,000 website users may be on the way to creating a perfect pop tune.” Other stories used the same rhetoric, implying that music creation is on some kind of inevitable arc toward composing by Artificial Intelligence. It’s not, and this study doesn’t help us get anywhere worth going.

The study’s authors didn’t write the stories of course, but they are contributing to the hype. Again from Science Daily:

Armand Leroi, co-author of the research and Professor of Evolutionary Developmental Biology from the Department of Life Sciences at Imperial College London, said: "Everyone 'knows' that music is made by traditions of musical geniuses. Bach handed the torch to Beethoven who gave it to Brahms; Lennon and McCartney gave it to the Gallaghers who gave it to Chris Martin. But is that really what drives musical evolution? We wondered whether consumer choice is the real force behind the relentless march of pop. Every time someone downloads one track rather than another they are exercising a choice, and a million choices is a million creative acts. After all, that's how natural selection created all of life on earth, and if blind variation and selection can do that, then we reckoned it should be able to make a pop tune. So we set up an experiment to explain it."

We know we’re in perilous territory when somebody elevates Oasis (The Gallaghers) to the level of the Beatles. But more to the point, what’s a scientist doing speaking in these really vague metaphors? What is this “torch” he speaks of?  Of course some bodies of work influences others, but in a dance so complex and non-directional that he should be ashamed for telling people that there’s a “relentless march” through time of music popularity. Beethoven preceded “How Much Is That Doggie In The Window,” which was a big hit. And that preceded Radiohead’s entire body of work. And if there’s a thread or a marching direction from one to another in that sequence, I’m Louis Armstrong.

Elsewhere Mr. Leroi is quoted as saying “We don’t often think of music as evolving.” Beg your pardon, but yes we do. Folks like me use that term all the time. A Google search for the exact phrase phrase“evolution of music” turns up 1.5 million hits. But it’s another metaphor, with vast limitations. Of course music changes over time in a dialogue between creators and audiences. But Darwinian evolution? Natural selection? That’s an outlandish stretch. Musical works don't breed or recombine their inner code. And beyond that, music that’s less popular doesn’t die. It may not sell great, but new music, from the innovative to the derivative, continues to get made in all genres heedless of commercial success. Build on the Leroi premise and there should be no jazz being made today and no great new composers making new jazz, because new jazz comprises about 1.5% of the music market and almost nobody on the street could tell you who Christian McBride or Brad Mehldau is. Yet jazz clearly continues to evolve, despite its fragmented and modest audience. 

(It's also worth pointing out that popular music has been influenced by myriad factors far more consequential than audience feedback loops over the past 100 years, including electric microphones, the Sears catalog, the Great Depression, World War II, the guitar pickup and multi-track recording. Just to get started.)

So I categorically reject the study’s core inference that music’s “evolution” is led by audiences. Every artist I’ve ever interviewed said the same thing - that they make the music they hear in their head and hope people respond to it. Certainly the shifting tastes of the general population contributes to music trends, some ephemeral and some long-lasting. But they play almost no role in the creation of novel sounds. The world did not know it wanted The Beatles before the Beatles arrived. The only people who had any kind of plausible role in the emergence of Radiohead or Sigur Ros comprised a minuscule percentage of the human race who were actively interested in avant-garde rock. So let’s be clear: composers lead. Audiences follow, in ways so complex, subjective, granular and ambivalent that the supercomputers of the year 2100 won’t be able to model them, if I may be granted my own bit of wishful thinking.

Think about it from the flip side. What’s the surest path to the creation of BAD pop music? Well it happens all the time: a team of songwriters or an artist/producer try to jam whatever qualities they can discern from the current Top 10 into their music in hopes of achieving a hit, paint-by-numbers style. Conversely, great works of music have been utterly misunderstood or rejected by the public at the outset,  yet that didn’t stop the onrush of modern music in its myriad forms. If anything, mass audiences are resistant to change and vote to favor the familiar over the unfamiliar – not the novelty the study purports to “explain.”

And at the end of the day, it's not even newsworthy that computers are autonomously generating “music.” They’ve been doing this, with banal results, for decades. The novelty of this study is that it got lay people to teach the computer the rules of acceptable-sounding music, rather than an expert using algorithms. Yes, their computer improved the musicality of its little loops with outside input. But even leaving aside the fact that brief, cycling loops do not come remotely close to “composing” a true piece of music with inner logic and import, somewhere out there, a real composer is right now making far more nuanced, interesting and important computer-assisted music using synthesized sounds that would be perceived by most in the study to be noise. Moreover, computers have proven really good at far more difficult tasks than generating ambient audio treacle. They can design pharmaceuticals and even write convincing journalism. Now that may give ME cause for concern. But composers and musicians, fear not. Your job is precious, uniquely human, and here to stay. 

 

 

 

Comments

[Jon]

Yah, this whole story struck me as either really pointless or really badly conveyed. It appears that the conclusion - that audience choices influence compositional results - comes from an experiment that is built precisely to have audience choices influence compositional results. Big surprise.

[Bobmaccallum]

@Jon Sure, we (the authors) had a hunch that it would work with a "large" audience but we still wanted to test it as rigorously as possible and of course find out what kind of music emerges. I'd say, subjectively, that the results of the last 3000 generations (since the PNAS paper + associated publicity) have been the most interesting. We addressed a technical issue with the genetic mechanisms that was probably causing the plateau in the published results and now it seems we have proper melodies, counterpoint (occasionally) and more.

@stringtheorymedia of course we know we're not making a nuanced symphony (and probably never will in a purely selection-driven approach) - loops are pretty the only way we can go at the moment. It's an experiment - you don't have to have everything exactly like the real world, as long as the controls are there.

Furthermore, with one (or two) evolving populations driven by effectively the whole of internet-connected humanity, we know it's going to turn out rather bland and unoffensive. You state that niche "markets" such as jazz exist in "real" music - there's no reason why they can't in a future evolutionary music ecosystem which would sit happily alongside all other forms of music. I haven't stopped listening to human composed and performed music and Armand Leroi is on record saying the same thing (http://www.guardian.co.uk/science/audio/2012/jun/25/science-weekly-podcast-darwintunes-naturally-selected-music at around 30mins).

Finally, we think musical works do breed and recombine their inner code - not to the total exclusion of human composition - but enough to let audience selection have some effect. How much effect? We didn't address that question. You'd have to look at the corpus of "real" music for that. We've just shown that audience selection can be a reasonably powerful force. Listening to the stream right now, I think it's quite amazing to be listening to listeners and not a composer.

[Alec K]

You took this too far to heart. They aren't attacking the institution that is your music they are just adding to it. You can argue that this music is not pleasing you in some way but that would be further validating that it is music. No one is seriously suggesting that this is how all future music will be written. It's just a tag line to generate interest in the story. I'll admit one that is not needed because the real story should focus on how this is one of the first times people who have no connection in real life are coming together to democratically make music....doesn't really mean anymore than that and if you had made this point instead of bluntly attacking a phenomena that is actually pretty cool maybe you could get more people to read. Don't want to lose YOUR job to a computer? Then boil a story down to its bones. Just the straight facts about what is going on...that's more interesting than your opinion about how these people are "wasting" their time.

[Adam Criswell]

Good points ... I am almost always on the side of nature/humanity/complexity over technology/artificial intelligence/mastery ... I get in heated discussions about this a lot with people who claim that humans will be physically augmented by technology in the near future. We have such a limited understanding of the human conscious, neurology, and higher order processing like music, and it reeks of hubris for people to claim we do (such as in this article). I'm in medicine, and all you have to do is look at the tiny (relatively speaking) dent we've put in cancer over the last 100 years of modern medicine to understand the ways we have to go. Let's figure out a way to deal with small cell lung cancer before we plant computer chips in people's brains.

[stringtheorymedia]

I left for vacation and didn't check in here to see several great comments waiting in the approval queue (an unwanted setting made necessary by a big spam comment problem). I'm very pleased a study author thought enough to reply and I appreciate your notes. I think I'd have been writing about this study with enthusiasm if the coverage and commentary from Leroi and co. had been more clear about its limits. I see Darwin Tunes more like a user interface than a source of insight about how music "evolves" in arts history. It's a way of teaching a computer a few basic rules, chiefly that consonance beats dissonance and rhythmic beats chaotic. (Of course those rules were made to be broken and are all the time to powerful musical effect.) I also think the general spirit of the findings use language that implies that if a composer took two successful works and actively blended them they could make something successful, even more successful than the original works. But believe me, I hear that happen all the time on Music Row, and that's why we critics treat the word "derivative" as a pejorative and not a compliment. All that said, all research is good, and heavens, maybe one day your hive mind/machine will "compose" something that truly amazes me. I've been surprised by computers before!