Daisy, Daisy
March 12, 2009 1:01 AM   Subscribe

Thanks, I'm going to have nightmares now.
posted by shii at 1:06 AM on March 12, 2009

Jesus! They didn't mention it was 2,000 people in the fucking bowels of hell!
posted by lattiboy at 1:15 AM on March 12, 2009 [7 favorites]

Hehe. The main page lets you listen to the voices in 12 separate tracks as well (as well the as whole thing mashed together as in the video), just use the People drop down menu.
posted by bjrn at 1:22 AM on March 12, 2009

And on the main page, as the notes scroll by, you can click on them to see and hear the original waveform and the attempts people made to replicate it. Some people are way off in pitch & duration - one person even just submitted a recording of a guitar string for a spoken syllable. It would be cool to be able to hear the song with only the best sample for each note.
posted by scrowdid at 1:33 AM on March 12, 2009

That was orders of magnitude creepier than I expected. I once did a cover of this song, and it just ended up being really creepy, so I think this song just lends itself to being either charmingly cute, or freakishly frightening.
posted by spiderskull at 1:45 AM on March 12, 2009

The computer wins.
posted by NekulturnY at 2:27 AM on March 12, 2009

Wow. Turns out the Uncanny Valley is made of people.
posted by Happy Dave at 2:43 AM on March 12, 2009 [19 favorites]

I guess it seemed like a good idea ahead of time.
posted by digsrus at 3:37 AM on March 12, 2009

I think one of the problems is that a lot of the participants tried to imitate the robotic sound, so you have a bunch of people imitating Stephen Hawking, singing the song. It's interesting, and it really works, but I have to wonder if it wouldn't come out better telling people to use their "normal" voice.

Now, if you want to hear a real robot sing Daisy Bell, there's always that episode of Futurama where Bender dates Planet Express Ship.
posted by explosion at 3:55 AM on March 12, 2009 [1 favorite]

That bicycle is going to kill me in my sleep.
posted by louche mustachio at 4:26 AM on March 12, 2009 [1 favorite]

If you click on a "note" you can hear what individual submissions sound like compared to what they were asked to produce. Some of those people are obviously deaf. Like a "shhhhhh" sound is reproduced as "eeeeeeeeeeeeeeeeeeee". What.

He says he captured the sound using "a Processing applet" but I assume it was just a sound recorder of some kind. Maybe it should have some feedback or filtering to try to keep people somewhere in a reasonable range of what they should be doing. Like, they only get paid if the frequency response is within X units of the required value or whatever.
posted by DU at 4:45 AM on March 12, 2009

Why, hello there nightmare fuel.
Mechanical Turk needs to be used to make some kind of ridiculous Rube Goldberg device to counteract the horror.
posted by incompressible at 4:50 AM on March 12, 2009 [1 favorite]

Although now that I think about it, maybe the poor quality is kind of the point. It's the kind of thing that should work in theory but doesn't in practice. You can't get a good song by asking 2000 people each to sing one note. And you can't get a good piece of software by asking 2000 people each to write one function. And you can't get good health by asking 2000 doctors to each handle one of your medical problems. Hyperspecialization doesn't work because there's nobody overseeing the larger picture (or at least a larger subsegment of the enormous picture).
posted by DU at 4:51 AM on March 12, 2009 [2 favorites]

Wow, sibilance is a real problem here, and not just in the usual audio sense, where the 'S' sound is too sharp and prominent, but also in the rhythm of the song. The sibilance totally fucks up the rhythm: this thing drags like crazy! Also interesting how the 'G' sound in "carriage" and "marriage" is so prominent, and also messes with the timing.

As has been noted in the past, there was one aspect of the famous "Daisy" scene in 2001, A Space Odyssey that didn't ring true: the way that the computer voice gradually slowed down and dropped in pitch only happens with analog technologies like audio tape and vinyl. Of course, HAL wouldn't have slowed down, he'd have stuttered and/or dropped out here and there, with perhaps a few odd digital sonic artifacts popping up here and there. I read somewhere a long time ago that Kubrick knew this, but went with the slowing down anyway, for the powerful effect.
posted by flapjax at midnite at 4:59 AM on March 12, 2009 [3 favorites]

People have a lot more time to spend on meaningless stuff these days. Thanks computer age!
posted by Pollomacho at 5:05 AM on March 12, 2009 [3 favorites]

Finally, we have been able to harness the power of crowdsourcing and digital media to replicate the long-lost sound of an LP melting.

(But what it really reminds me of is Dictionaraoke!)
posted by kittyprecious at 5:54 AM on March 12, 2009 [1 favorite]

And here I was hoping for an actual pedal-powered conveyance built for 2,000 people! I demand justice!
posted by Xoder at 5:59 AM on March 12, 2009 [2 favorites]

This is what "Thunder Road" would sound like at a zombie Springsteen show.
posted by stargell at 6:01 AM on March 12, 2009 [1 favorite]

Quite similar to this in many ways. Both are pretty cool, this kind of thing always gives me a smile.
posted by Thoth at 6:03 AM on March 12, 2009 [1 favorite]

I found it charming, if a bit discombobulated.

Nobody felt any sort of malevolence in making the sounds, nor was the overall structure in any way creepy, but people seem to still hear creepy or threatening sounds in this (not me, I have had my sonic emotional receptors mutated by listening to years of Merzbow and David Tudor and Iannis Xenakis - they all get much better when you stop thinking they sound evil or threatening or angry, by the way). I am tempted to name this "sonic xenophobia" and make it a metaphor or symptom of human intolerance, but that would be silly.

The Max Matthews who made the original computer sing this song is one of the namesakes of Max/MSP (Miller S. Puckette being the other).
posted by idiopath at 6:27 AM on March 12, 2009 [1 favorite]

It seems to me that one or two people could have made an equally terrible version of this song and saved 1,999 other people from wasting their time.

I get that distrubuted, collaborative online projects can be very cool and some are surely going to create impressive results, but not this one. All that planning, testing, merging, editing and the final product is drivel.
posted by Paid In Full at 6:40 AM on March 12, 2009

...the final product is drivel.

posted by flapjax at midnite at 6:56 AM on March 12, 2009 [1 favorite]

I'm not usually one to question others' hobbies, but I'm really having a difficult time seeing the point of this exercise.
posted by Fleebnork at 7:12 AM on March 12, 2009

Can anyone help me get a possible mp3 made of this? I want to use it as part of my promo for the radio show I do.
posted by Lipstick Thespian at 7:21 AM on March 12, 2009

Okay, I listened to track 1. "Can't" sounded like a sample of Barney Gumble burping, and the second half of "carriage" was distinctly a dude saying "nooch." So my question is, are there really that many people on Mechanical Turk just to mess with projects? Was it a reaction to the oddness of the project? Or are the creators messing with us?
posted by No-sword at 7:21 AM on March 12, 2009

I love that track.

Now, can anyone tell me where I can get the version of "A Bicycle Built for Two" that played during the drunken tricycle race in Revenge of the Nerds?
posted by JBennett at 7:31 AM on March 12, 2009 [1 favorite]

Both collaborators are interested in the combined efforts of many to make something larger, if you check their other work. Aaron Koblin was one of two people behind Thoth's Ten Thousand Cents link (the other being Takashi Kawashima) and the guy behind Radiohead's House of Cards 3D "open-source" video, discussed previously, as was his Flight Patterns project (now housed here), and The Sheep Market. Daniel Massey's work is more installation-based, but also playing with multiple points of input/output.
posted by filthy light thief at 7:33 AM on March 12, 2009

Lipstick Thespian - they've already done it for you (download this - retrieved with the ever-useful Download Helper firefox extension). It's only 128kbps, but probably fine for a radio show, and suitable given the input material.
posted by filthy light thief at 7:36 AM on March 12, 2009

(And they seem to have ALL of the source files available, if you root around their server from that download link - sorry for the series of comments, I'm a bit slow on the uptake right now.)
posted by filthy light thief at 7:38 AM on March 12, 2009

HAL wouldn't have slowed down, he'd have stuttered and/or dropped out here and there, with perhaps a few odd digital sonic artifacts popping up here and there

You're right about the pitch, but a slowdown could have been possible depending on how HAL's software worked. If shutting down HAL in steps as shown in the film gradually reduced HAL's processing power, and HAL's software was set to cope with this by doing the same amount of work but doing it slower than real-time, it would result in a slowdown effect (but no drop in pitch). The closest I've seen to something like this happening in real life was being able to speed up and slow down games (including the sounds and music) on my old 486 computer by toggling the TURBO button.

I'm not usually one to question others' hobbies, but I'm really having a difficult time seeing the point of this exercise.

Writing computer programs is mostly about taking a complex task and breaking it up into small enough steps that a computer can do them. Computers don't have any common sense, and are actually fundamentally only able to perform simple mathematical operations, so telling a computer to do something is harder than it sounds.

The Mechanical Turk is similar in that it allows complex tasks to be broken down into very simple steps, with the added bonus that since people are performing the steps, they can easily do things that computers are bad at (like singing a song without sounding like a robot). I think what this particular project shows is that, at least in the form of the Mechanical Turk, it's difficult to both break a problem down into small steps and still take advantage of a human's innate abilities. When the song is broken down into this small of chunks, humans are pretty bad at matching the sound accurately, even though paradoxically most of them would probably be able to sing a decent version of the song if they had to reproduce the whole thing. It shows that although humans are relatively good at solving complex tasks individually, breaking down that task into small pieces and handing them out to many different humans doesn't necessarily produce as good of results.
posted by burnmp3s at 7:40 AM on March 12, 2009 [5 favorites]

I keep expecting Mechanical Turk to reveal it's sinister plan or something. After all, the best agent is someone who doesn't know they're an agent, right? Some kind of attack or plan put in motion by the mindless factory-actions of bored college students the world over, not knowing of what they do.
posted by The Whelk at 7:54 AM on March 12, 2009

...it would result in a slowdown effect (but no drop in pitch).

But a drop in pitch is just another slowdown. If you imagine HAL outsourcing the notes to an external device ("first do an A, then a G, then a D") etc then each note is the same but the spacing is wider. But if you imagine him actually doing the pulse-width modulation himself (turn the speaker on, then off, then on, then off 440 times per second) then a general slowdown could well translate into a pitch drop.
posted by DU at 7:55 AM on March 12, 2009 [1 favorite]

The internet never ceases to amaze me. I would never have imagined such a project existing until I went to that page. Part of me says, "why?" and part of me says, "I'm clearly not imaginative enough."

Still, I appreciate something new and different!
posted by dios at 9:15 AM on March 12, 2009

it really does sound like what i'd imagine a bicycle 2000 people were riding would look like. wobbly and slow, pretty fucked up.
posted by fuzzypantalones at 9:21 AM on March 12, 2009

Neat concept, but it's an audio Frankenstein('s monster) in more ways than one. It probably could produce some cool results if executed properly.
posted by Challahtronix at 9:22 AM on March 12, 2009

this is one of the original computer recordings out of bell labs...containing the 'daisy' track...judging from what google is feeding me, apparently this was released as both a magazine insert floppy record, and as a stand-alone 7". but...there's another version...an earlier one that doesn't have the 'musical' accompaniment...it's just the computer voice. and very spooky. i downloaded the recording from an article about how a copy was found in kubrick's archives (i believe it was specified as a floppy record version), so it's no accident that that was the song hal was singing in the movie...i'll see if i can dig up the other one...brb
posted by sexyrobot at 10:03 AM on March 12, 2009

ok, so that version was the 7"(1963), the floppy version (1961) is the version i have...discussed here (but don't bother with the link to the track...busted) and then (haHA), there's this crazy third version that has both the solo and accompanied version of 'daisy' ...which i found on this album of computer music from 1978...enjoy!

i really dig this new crazy version, but shouldnt it be 2000 AND ONE voices?
posted by sexyrobot at 10:35 AM on March 12, 2009

The Mechanical Turk is... (snipped)

Ok, now I can understand it a bit better. I had no idea what the implications of the exercise were, based on the information in the link. And, well, I wasn't interested in googling around to try to weed out that information.

Maybe that's lazy of me. I blame internets.
posted by Fleebnork at 11:00 AM on March 12, 2009

wobbly and slow, pretty fucked up.

you guys seem to be complaining mostly that the sound quality is all fucked in this version, so slow, wrong pitch, etc...but actually, its a damn good match of the original version (which is what these guys were given by the mechanical turk), all things considered...its the original thats all off key, slow, strangely sibilant, etc...listen to that first, then compare...i've had the original in my itunes for a few years, now, so it's totally recognisable to me...i don't think the process rates a FAIL.
i'd love to hear something else done this way, but without the music being separated from the vocals...so you'd end up trying to replicate a noise that's part human, part instrument. 'smells like teen spirit' would probably be AWESOME....ditto on 'edge of seventeen'
posted by sexyrobot at 11:28 AM on March 12, 2009

part human, part instrument

Like, the body is human, and the head is a tuba bell?

No, no, wait... the torso and head are human but the limbs are bassoons and soprano saxophones!

Or, the arms and legs are human but the torso is a bagpipe and the head is a pair of hi-hat cymbals opening and closing!

Ooh, this is fun!
posted by flapjax at midnite at 4:14 PM on March 12, 2009

part buttocks, part high-hat
posted by sexyrobot at 4:44 PM on March 12, 2009

I talked to a friend about this, and he knows the person who made it. It seems he works for the "creative" division of google labs, and he likes to use mechanical turk for just about everything. For example that mechanical turk reproduction of a hundred dollar bill was reportedly his doing.
posted by idiopath at 11:00 AM on March 13, 2009 [1 favorite]

« Older Cramer vs. Cramer   |   Time paradoxes and alternate universes Newer »

This thread has been archived and is closed to new comments