Bieber de novo

June 19, 2012 7:15 AM

Evolution of music by public choice. Results are out from the DarwinTunes experiment (previously) - a Darwinian music engine working with mutating short audio loops that underwent evolutionary selection for 2,513 generations under the influence of thousands of listeners. Article in the Proc. Nat'l Acad. Sci. You can participate at the DarwinTunes web site.
posted by exogenous (41 comments total)

Sounds like background music for an ad, or interstitial music for a training video.
posted by leotrotsky at 7:37 AM on June 19, 2012 [2 favorites]

The autechrishness of gen zero was way better than the pop-quasi-minimalist crap of gen 3060.
posted by symbioid at 7:38 AM on June 19, 2012 [7 favorites]

Haven't listened yet, but pretty sure it all converges into the opening chords of "September Gurls".
posted by BitterOldPunk at 7:42 AM on June 19, 2012 [1 favorite]

Pretty sure I am gonna loop this all day while banging out work.
posted by beaucoupkevin at 7:43 AM on June 19, 2012 [3 favorites]

Those testers are going to be very disappointed when they discover 10000 generations of natural selection eventually all converge in Auto-Tune.
posted by wolfdreams01 at 7:45 AM on June 19, 2012 [1 favorite]

generation 100,000
posted by East Manitoba Regional Junior Kabaddi Champion '94 at 7:47 AM on June 19, 2012 [4 favorites]

It's interesting how it all seemed to converge to the pentatonic scale, the simplest type of consonance. It seems maybe that the software inherently doesn't really do "melody" or melodic lines - the structure of the music is all very vertical, so all of the selection pressure seems to just be acting to make it more and more consonant without, e.g., giving it any forward direction or tension/resolution. I'd love to see more of the intermediate steps, too. Still, this is a cool project.
posted by en forme de poire at 7:49 AM on June 19, 2012 [1 favorite]

interstitial music for a training video

OH GOD NO MAKE IT STOP PLEASE FOR THE LOVE OF GOD MAKE IT STOP.

(Aaaah. Better.)
posted by maudlin at 7:49 AM on June 19, 2012 [2 favorites]

I'd like to hear the version that evolves away from what people choose.
posted by swift at 7:52 AM on June 19, 2012 [3 favorites]

maudlin: "interstitial music for a training video

OH GOD NO MAKE IT STOP PLEASE FOR THE LOVE OF GOD MAKE IT STOP.

(Aaaah. Better.)"

I had to stop listening to the Reich stuff - it was doing ... weird shit... to my brain. Damn, that's powerful.
posted by symbioid at 7:52 AM on June 19, 2012

Instead of self-report ratings of how nice the music sounds, they should have participants do an unrelated task with these tunes as background music.

The tunes associated with highest performance levels "survive" to the next generation.

Over thousands of generations you will produce pure scientifically derived motivation in musical form.

The natural hypothesis is that it will sound like this, but science has been known to make mistakes from time to time.
posted by saturday_morning at 7:59 AM on June 19, 2012 [2 favorites]

"It's interesting how it all seemed to converge to the pentatonic scale"


From the full text of the study:
"Each genome/program specifies note placement, instrumentation, and performance parameters; however, tempo, meter, and tuning system are fixed for all loops"

The scale was fixed and did not vary.
posted by idiopath at 7:59 AM on June 19, 2012

I realize this probably sounds untrue, but I genuinely prefer generation 0 to generation 3060.

I'd kinda like to see what DarwinTubes would do with Merzbow.
posted by namewithoutwords at 8:04 AM on June 19, 2012 [2 favorites]

That's sure not Bieber. But it does remind me a lot of Yanni.
posted by benito.strauss at 8:13 AM on June 19, 2012

idiopath, I think "tuning system" just means they left it in 12-tone equal temperament, right? I just meant the notes that actually get selected. Though actually re-listening to the evolved ones, they aren't totally pentatonic either - I was just fooled by all the relentless consonance.
posted by en forme de poire at 8:14 AM on June 19, 2012

Muzak for robots. I like it.
posted by freakazoid at 8:20 AM on June 19, 2012

"It was one of countless similar songs published for the benefit of the proles by a sub-section of the Music Department. The words of these songs were composed without any human intervention whatever on an instrument known as a versificator. But the woman sang so tunefully as to turn the dreadful rubbish into an almost pleasant sound."

This is how our robot overlords will control us.
posted by tempestuoso at 8:23 AM on June 19, 2012 [1 favorite]

This is neat.

And boy are the cookie buttons annoying. "Yes, I'll accept your cookies" and "No, I'd like to learn more". How about just "No"? And then clicking any link is an implicit agreement? WTF?
posted by DU at 8:31 AM on June 19, 2012 [1 favorite]

Feh. I liked them better before they sold out got popular evolved.
posted by Devils Rancher at 8:38 AM on June 19, 2012 [1 favorite]

From the first link,

"The evolution led to pleasant, jingly tunes that didn't offend anyone but didn't really move anyone, either."

So, like Paul McCartney + Wings.
posted by philip-random at 8:45 AM on June 19, 2012

At the other end of the spectrum: Scientific Attempt To Create Most Annoying Song Ever.

But just like gen 0 is better than 3060, so too, is this song mos def NOT the most annoying song ever... kinda catchy.
posted by symbioid at 8:49 AM on June 19, 2012 [3 favorites]

The autechrishness of gen zero was way better than the pop-quasi-minimalist crap of gen 3060.

Generation zero sounds almost like a less reverbed Oversteps.
posted by TimeStove at 8:51 AM on June 19, 2012 [2 favorites]

I was listening to Oversteps when I posted that ;)
posted by symbioid at 8:52 AM on June 19, 2012 [1 favorite]

Came here for the Komar and Melamid link. Leaving satisfied.
posted by spitefulcrow at 8:57 AM on June 19, 2012 [2 favorites]

I'm surprised that they're framing this as providing insight as to how musical language evolves, especially since the study converges towards one result. It seems to me that it's more functional as a means of getting a picture of what our baseline expectations for music are in the aggregate -- a static picture, not a dynamic one. My intuition tells me that a study of stylistic evolution would be better focussed on a tree or network of inter-related styles, though of course in that situation you would have to take into account the fact that the "mutations" that are introduced are not independent from other factors as they are here. You could use the same genetic algorithm they use here, but with a different distribution strategy -- what would happen if you gave the same first generation clip to several groups on different continents? Then what would happen if you redistributed the clips at generation 1000, versus letting the groups take their original group to its convergence point?

It's interesting how it all seemed to converge to the pentatonic scale, the simplest type of consonance.

I get what you're getting at, but I can't think of any sense in which the pentatonic scale is the simplest type of consonance (that distinction belongs to the unison, or the octave* if starting with the trivial case weirds you out). It's true that the three most acute dissonances (minor second, major seventh, and tritone) are not contained by the pentatonic scale, but you've still got the major second, minor seventh, and (depending on your aural proclivities) the perfect fourth in there, which are treated as dissonances in a lot of musical contexts. If we're approaching it from one of those contexts, the pentatonic scale is actually dissonant, since the consonance of a group of more than two notes implies pairwise consonance for all pairs of notes in the group.

It's true that the pentatonic scale doesn't have the unpleasant or astringent character that we tend to associate with "dissonance," and my personal pet theory for that is that we hear a harmonic major second as two stacked perfect fifths with the middle fifth omitted (e.g., C-D is heard as an elided C-G-D), and indeed the pentatonic scale can be constructed by taking the first five notes of a sequence of rising fifths (e.g., C-G-D-A-E -> C-D-E-G-A). That may have something to do with its apparent mildness, and there are certainly contexts in which that pitch collection could function as a consonance, but I wouldn't call it consonant full-stop.

But, yeah, there's a pretty prominent fa-mi figure going on through most of the final product, so I guess this is all technically moot.

* This all assumes harmonic overtones. In more exotic contexts, dissonant octaves are certainly possible.
posted by invitapriore at 9:11 AM on June 19, 2012

At the other end of the spectrum: Scientific Attempt To Create Most Annoying Song Ever.

Ramadan
Ramadan
Lot's of praying, with no breakfast
Ramadan
So much fun
Do all your shopping....AT WAL-MART

Come on, that's awesome.
posted by teekat at 9:27 AM on June 19, 2012 [3 favorites]

invitapriore, you're totally right. I was trying to get at something more like, if you bang around on a pentatonic scale you tend to get stuff that's more consonant than if you did the same thing with a major Ionian scale, because you're so likely to get fifths and thirds (and fourths, which I guess are sort of consonant) and you never get, e.g., a minor second or a tritone. But maybe that's kind of an imprecise thing to say.
posted by en forme de poire at 10:18 AM on June 19, 2012 [1 favorite]

BTW, there's another DarwinTunes experiment going on with just percussion right now. Again, generation 3 sounds pretty interesting and generation 202 sounds a lot more commercial - but I don't hate the evolved version either since it seems less insipid than the non-percussion version.

(You can also see earlier generations of the DarwinTunes at that first link if you scroll down, btw. I thought generation 150 was pretty cool, in particular.)
posted by en forme de poire at 10:29 AM on June 19, 2012 [1 favorite]

Full of sound and fury and signifying nothing.
posted by Seekerofsplendor at 10:29 AM on June 19, 2012 [1 favorite]

BTW, there's another DarwinTunes experiment going on with just percussion right now.

Holy shit, look at the waveforms between generations. More than anything it seems that people love compression.
posted by invitapriore at 12:02 PM on June 19, 2012

It's an interesting experiment. I really enjoyed the late generation loops at first, but they do get repetitious after awhile. And after fifteen minutes, everything starts to sound the same!

Dum da dum da dum da da.

Dum da dum da dum da da...

Dum da dum da dum da da......
posted by Kevin Street at 2:33 PM on June 19, 2012

By which I mean, the late generation loops are very ear-wormy, but they don't hold up to repeated listening. And it's kind of neat to take part in the experiment rating earlier generation loops, but after fifteen minutes I could barely tell them apart.
posted by Kevin Street at 2:43 PM on June 19, 2012

I guess I was confused in assuming that music has always been evolved by public choice. What do computers have to do, have to do with it?
posted by Twang at 3:14 PM on June 19, 2012

The computer just speeds things up, and removes the composer from the process. It's more like a measure of public taste, without any artistic vision at all.
posted by Kevin Street at 3:24 PM on June 19, 2012

"the experiment starts with loops where the notes and sounds are randomly assigned by the computer. The sounds are built from combinations of sinusodal waveforms. They sound pretty awful."

Huh. I thought they sounded pretty cool.
posted by delmoi at 10:31 PM on June 19, 2012

Yeah... It's like taking something wild and untamed and chaotic and truing it to something boring. They must have some kind of seeding going on because it does sound 'musical', not just a collection of random noises.
posted by delmoi at 10:42 PM on June 19, 2012

They should do this experiment again but with fans of different genres trying to match their favorite kind of music.
posted by delmoi at 10:44 PM on June 19, 2012

this is pretty cool
posted by delmoi at 10:47 PM on June 19, 2012

Somewhat reminds me of a few of the Exhibitions at the Museum of Techno.

Start here:

joseph kelly and the deep structure of techno

followed by this:

restoration of basslines to c20 techno

and then this:

the search for perfect techno

Note: "the first 9 iterations are omitted because they all sounded a bit rubbish."

###

While you're there, don't miss the description of "Aafke van Hoovernaars' steam-powered drum-machine."
posted by illongruci at 10:41 AM on June 20, 2012

Argh, what happened to my links? Starts here, anyway:

http://archive.museumoftechno.org/exhibition_detail.php?id=4
posted by illongruci at 11:41 AM on June 20, 2012 [1 favorite]

A similar visual project which is years old. Electric Sheep is a collaborative abstract artwork founded by Scott Draves. [previously]
posted by Monkey0nCrack at 5:12 PM on June 22, 2012