Clock Evolution Model

December 26, 2007 12:01 AM

Natural selection and evolution in clocks(youtube) - Video of the details and results of a program written to model the evolution of clocks (if they were alive).

It seems that it was done to disprove a strawman argument used to argue in favor of creationism by some (an argument I hadn't personally heard used before). Found here, and explained in more detail than I can offer.

posted by Stunt (46 comments total)

I just found this rather fascinating. Not the bit trying to prove an argument one way or another, mind you, but the fact that he apparently wrote a program to model the evolution and mutation clocks; I'm a geek. My apologies if this is a double and I missed it somehow, I just thought it was too neat not to share.
posted by Stunt at 12:01 AM on December 26, 2007

Not that I actually want to read his code (much), but I wonder how he explained to the program that 'able to tell time' --> 'more fit to survive.' How is 'able to tell time' defined biologically?
posted by shakespeherian at 12:13 AM on December 26, 2007

So...this video invalidates the ID theory through sheer confusion?
posted by zardoz at 12:14 AM on December 26, 2007

How is 'able to tell time' defined biologically?

As far as I can tell: swinging pendulum < 1 hand < 2 hands etc, with "accuracy" defined by gear ratios.
posted by anarcation at 12:23 AM on December 26, 2007

Ick. The guy who made this video doesn't understand what they're trying to illustrate with the watch example.

Stunt, this technique is called "genetic algorithms" or "genetic programming". Nanopond is a simple, open-source program that implements this and is easy and fun to experiment with if you're a coder.
posted by XMLicious at 12:24 AM on December 26, 2007 [1 favorite]

As far as I can tell: swinging pendulum <>

Right but the video states that the program isn't given a desired outcome, and it seems to me that suggesting 'try to get a swinging pendulum' or 'hey, one hand on a gear is better than no hands on a gear!' is gaming the system in a way that invalidates his no-ID line. I think it's a weird example of evolution working, because he does have a desired outcome (i.e. a clock), whereas actual biological evolution doesn't-- it's just a game of whoever reproduces most successfully. I guess I'm wondering how he establishes that 'able to tell time' is a desired outcome without telling the program how and what steps to go through; if in fact he is telling the program that, the whole project is a bust IMO.
posted by shakespeherian at 12:35 AM on December 26, 2007

Aw shit, close tag.
posted by shakespeherian at 12:35 AM on December 26, 2007

Also, if anyone is interested in debunking Intelligent Design or would like to see a more articulate response to the watchmaker argument, this is the resource. (2-hour NOVA documentary)
posted by XMLicious at 12:42 AM on December 26, 2007 [2 favorites]

Well, a nice effort on behalf of the coder (I guess), but the video "design" is pretty awful. This would be a fairly obvious example to anyone who has studied genetic algorithms.

What's much more interesting is the fact that genetic clocks actually have evolved. There are all sorts of timing mechanism in cells.
posted by delmoi at 12:42 AM on December 26, 2007 [1 favorite]

The thing that's great about the whole evolution debate is the polite discussion it spawns.
posted by mullingitover at 2:08 AM on December 26, 2007

Thanks for the nanopond link, XMLicious.

I agree about the overall way the video was put together, delmoi. I liked the write up I found on that blog and found it a pretty good summary of a video that was a bit...less than smooth. Not even sure how useful the guy's code is, I just thought the overall work that went into the idea and the execution is neat.

Then again, I suppose ideas of Spore meets Sim Clock leading to me having a Rolex Terrier might have colored my vision. I thought "Huh, cool. EVOLVE! GO! MORE TEETH!", followed by "Hmmm, I wonder how well this guy's code is actually written" followed by "What? this is an ID counterargument? MORE TEETH! GO, BURNINATE THAT OTHER CLOCK!".

Clearly, there is something wrong with me.
posted by Stunt at 2:13 AM on December 26, 2007 [1 favorite]

I read the editorial until it got to where NoScript stopped it. Nothing up to that point made it seem worth continuing, so I didn't.
posted by Kirth Gerson at 3:33 AM on December 26, 2007

it seems to me that suggesting 'try to get a swinging pendulum' or 'hey, one hand on a gear is better than no hands on a gear!' is gaming the system in a way that invalidates his no-ID line

If you imagine a predator that eats the clock least able to tell the time, then it sort of works. But really any demonstration of evolution that ties survival to anything other than ability to survive is pretty worthless.
posted by cillit bang at 4:27 AM on December 26, 2007

Thanks for the link. I enjoyed it.
posted by RussHy at 4:41 AM on December 26, 2007

This would be a fairly obvious example to anyone who has studied genetic algorithms.

Which is not the audience.
posted by DU at 5:18 AM on December 26, 2007

But really any demonstration of evolution that ties survival to anything other than ability to survive is pretty worthless.

I don't see why. If you take 'fitness for purpose' to be the test of value in your conceptual model, you can simply assume that all those things that aren't fit for purpose would be discarded. Discarded = non-survival.

However, I'm a moron and know nothing about either biology or clocks, and I don't much like God-botherers, so I was easily convinced.
posted by PeterMcDermott at 5:48 AM on December 26, 2007

Nanopond is a simple, open-source program that implements this and is easy and fun to experiment with if you're a coder.

Nanopond is pretty cool and I'm having/going to have fun. But "implements this" is a little misleading. I thought I was being directed to a generalized genetic programming API/environment where I could define a biological toolkit that would interact similar to how the gears and ratchets of the video do.

Which now that I've thought of it, would be awesome.
posted by DU at 5:48 AM on December 26, 2007

Which now that I've thought of it, would be awesome.

Throw some junk in a box, shake it around a bit, then open it up and discover the surprise.

Ohwaitaminute.

That was last year's must have Xmas gift.
posted by notyou at 7:08 AM on December 26, 2007

The "right" way to do the fitness function is to actually simulate the operation of the mechanism and check for periodicities in its configuration. As all the components only have one or two degrees of freedom, that wouldn't be that difficult. I hope that's what he did.
posted by topynate at 7:22 AM on December 26, 2007

Yeah, I'd like to have a closer look at the code.
posted by jeffamaphone at 7:46 AM on December 26, 2007

Clearly there was something more sophisticated going on in his program than nothing < pendulum, etc. as the 4 handed clocks outperformed 3 handed clocks in some simulations but not others.
posted by TypographicalError at 9:02 AM on December 26, 2007

Totally stupid. Won't convince even one person.
posted by lupus_yonderboy at 9:12 AM on December 26, 2007

I have yet to see an argument that will convince an ID/Creationist of anything. It's not a rational point of view to begin with.
posted by jeffamaphone at 9:59 AM on December 26, 2007

I have yet to see an argument that will convince an ID/Creationist of anything.

Or any religious person, really. If your worldview is based on the idea that pure belief is enough to understand the world, then evidence is irrelevant.

The point here is to race to vaccinate the ignorant before "faith" infects them too.
posted by DU at 10:06 AM on December 26, 2007

But really any demonstration of evolution that ties survival to anything other than ability to survive is pretty worthless.

Sure, but this is modeling the (fantastic) survival of clocks. They have no biologic functions whatsoever. The whole thing is a nice little metaphor. For the purposes of this simulation, more accurate timekeeping = survival quality. You claim that that's "gaming the system". Well, you have to define a baseline of survival qualities for non-biological organisms. Maybe pretend the clocks not to be clocks but cardiac cell timing mechanisms and it makes more sense. Then again, it would be interesting to do the model in which evolving structures where the gears don't bind is the biological imperative...
posted by nonreflectiveobject at 10:35 AM on December 26, 2007

Sure, but this is modeling the (fantastic) survival of clocks. They have no biologic functions whatsoever. The whole thing is a nice little metaphor. For the purposes of this simulation, more accurate timekeeping = survival quality. You claim that that's "gaming the system". Well, you have to define a baseline of survival qualities for non-biological organisms. Maybe pretend the clocks not to be clocks but cardiac cell timing mechanisms and it makes more sense. Then again, it would be interesting to do the model in which evolving structures where the gears don't bind is the biological imperative...

I think my earlier questions were poorly worded. I'm not complaining about the fact that clocks don't reproduce and therefore the simulation is rigged-- I'm wondering how 'more accurate timekeeping' was codified and expressed as a survival quality, since the terms used in the video make 'more accurate timekeeping' seem to depend on a human observer who knows what a clock is (a gear turning is good, but not as good as a gear with a hand, because the hand helps you keep track of how far the gear has turned). How is the 'more accurate timekeeping' expressed within the program without specifically outlining a desired outcome?
posted by shakespeherian at 10:48 AM on December 26, 2007

Looking at the video suggests that his fitness function had various round numbers of seconds (15, 3600, 10 etc.) as beneficial, although that contradicts his claim of 'no goals' somewhat. Probably he had the fitness of each hand increase dramatically very close to any one of several values, then added them together; this would make one hand at 15.003s and several at random values better than one at 15.1s, one at 3615s and one at 9.95s, say.

shakespeherian, that bugged me too. I just realised that the sprung gear's function in the proto-clock is simply to slow the pendulum down (the arm of the pendulum is a hand, so contributes to the fitness function). Then when a hand gets added to that gear, it rotates rather than oscillates (due to the ratchet), so it has an even longer period - probably giving the clock fitness by getting towards a target of one minute.

He could have explained himself a bit better. It's really cool.
posted by topynate at 11:22 AM on December 26, 2007

I think a case can be made that the biological function of humans was to invent and perfect clocks, because we really, really are quite excellent at making clocks, and not that we have achieved this, we have served our function and should just go away, perhaps to emerge every so often, when clocks need winding or whatever.
posted by Astro Zombie at 11:26 AM on December 26, 2007 [2 favorites]

Maybe I'm reading something in to this, but I thought the fitness function was related of the range of hand oscillation periods in each clock. Remember that the pendulum was made from a hand, so by that simple metric the lone pendulum is fitter than no oscillating hands (and explains why single pendulum clocks appear in the first place).

What amazes me is he never got clocks with multiple pendulums dominating the gene pool.
posted by noble_rot at 12:26 PM on December 26, 2007

Clearly atomic clocks in orbit prove that god, sorry, God is fucking with us.

Explain that Evilutionists!
posted by Sparx at 12:51 PM on December 26, 2007

Maybe I'm reading something in to this, but I thought the fitness function was related of the range of hand oscillation periods in each clock. Remember that the pendulum was made from a hand, so by that simple metric the lone pendulum is fitter than no oscillating hands (and explains why single pendulum clocks appear in the first place).

That makes some sense, but I don't understand then why in one simulation there was a brief period of four-handed clocks that then went back to three-handed clocks.
posted by shakespeherian at 1:12 PM on December 26, 2007

Personally, I thought it was cool enough from a computing standpoint to share with others in my field.

As an aside, I haven't bought a single computer game since Spore was announced and I saw the video. They better get this game out on the shelves, because I plan to become completely addicted.
posted by Kickstart70 at 1:51 PM on December 26, 2007

...but I don't understand then why in one simulation there was a brief period of four-handed clocks that then went back to three-handed clocks.

Reality is that sometimes beneficial mutations don't survive, if other factors in survivability are more important (or just dumb luck, like all the telepathic humans were so engrossed in others' thoughts that the saber-toothed tigers ate them).
posted by Kickstart70 at 1:54 PM on December 26, 2007

Wow. This reminds of of a similar (more simplistic) project I coded, back when I was teaching myself Javascript. It's so old, it uses F-R-A-M-E-S!

For the sake of history and nostalgia, I preset Alphalution. It evolves random strings of letters into words.
posted by grumblebee at 1:55 PM on December 26, 2007

although an interesting - this clock example misses one of the main points on evolution. Natural selection does not aim to select for a purpose other than survival. Natural selection is a blind watchmaker because it does not see ahead, does not plan consequences, has no purpose in view.

To look at an example; natural selction does not set out to select for a wing so that a bird can fly- but rather a wing shape is developed through small step by step changes - each resulting in an advantage to the species eventually this wing shape allows the species to fly and so we have a wing the combination of genes that developes wings - it is not aimed for by evolution. The wing is then widely selected for because of the huge advantage (being able to fly away from the nasty things trying to eat you).

Evolution does not see into the future and it does not design for a purpose. It does not get a bunch of parts and decide it will keep rearranging them untill it reaches a clock. but rather the parts are rearranged and the design repeated if it helps the species survive.
posted by Lesium at 2:36 PM on December 26, 2007

If anyone is interested in reading a great and simple explanation of evolution and arguments countering I.D Then read Richard Dawkins 'The blind watchmaker". Even those that believe in I.D should read it - as it is always good to look at the other point of view and have the theory of evolution explained properly. Before you refute it.
posted by Lesium at 2:39 PM on December 26, 2007

This is my best guess at what his fitness function is:

First, for each hand, work out its periodicity.
Secondly, compute a pre-adjusted fitness f(x), for each period x. This function is probably the sum of several sharply peaked Gaussians, for instance f(x) = e^-[5(x-15)^2] +e^-[5(x-3600)^2] + e^-[5(x-1)^2], or some such. The constants inside the brackets are the target periods; the factors of five out front are to make the function more sharply peaked.
There is a problem with just adding them together here, because you can have two hands with similar periodicity (or the exact same), and get twice the fitness, which is not what you want. So, thirdly, for each pair of periods x and y, you compute a factor 1-e^-3(x-y)^2, say, which is 0 for x=y but rapidly increases to 1 further away, and multiply f(x) by it, giving an adjusted value, and the same for f(y). Keep doing this for all pairs. What you get is fitness of near zero for both hands, if they are close together. Thus a fit clock will have one hand per target period.
Finally, add all the adjusted f(x) together to obtain the fitness of the whole clock.

If instead of five we use the number 20 in the second step, we give extra emphasis to precision, and thus favour clocks with fewer, more precise hands over clocks with more, less precise ones, which hit all the targets. We can also make the parameter lower, which will create sloppier clocks that have one hand for each target.
posted by topynate at 2:44 PM on December 26, 2007

Ah, a wrinkle. I knew there'd be one. If you don't want hundreds of hands proliferating on every gear, you need to add a small penalty for each hand. So subtract c * number of hands, where c is on the order of 1/100 to 1/10, at the end.
posted by topynate at 2:49 PM on December 26, 2007

Natural selection does not aim to select for a purpose other than survival.

I understand what you're saying, Lesium, but "to tell the time" and "to survive" can both be viewed as purposes. If you're designing a toy universe, you have to come up with some sort of driving force behind selection, so why not "to tell the time"?

I guess that an ID person might mistake this for a specifically utilitarian force, in the sense that, in general, an intelligent entity wants to be able to tell time. But if the programmer had used "to survive" as the "goal," he could be equally accused of stacking the deck (by choosing that goal).

In the end, it's all about survival. In his universe, ability-to-tell-time is the same as survival. Watches that don't tell time well don't have a good chance of surviving.
posted by grumblebee at 2:51 PM on December 26, 2007

In other words, entities aren't selected simply to survive. They're selected to survive in a specific environment, which bestows specific materials and constraints. For instance, on Earth, entities need to be able to (at least) tolerate oxygen if they're going to survive.

So in universe X, maybe creatures need to be able to tell the time in order to survive.
posted by grumblebee at 2:54 PM on December 26, 2007 [1 favorite]

You guys are leaving sexual selection out-- the females (or males) could just have a random preference for "accurate timekeeping" however you want to define that [a sense of rhythm, perhaps?]

Come to think of it, these were probably bisexual/unisexual-- but the point remains that there isn't just survival in terms of whether it "stays alive," there is also survival in terms of whether others will mate with it if you are talking sexual reproduction, which seemed to be the case here.

Either way, such demonstration of intelligence gets *this* female hot...
posted by Maias at 3:28 PM on December 26, 2007

To further clarify my original questions: I'm not suggesting that the preference for 'accurate timekeeping' is an argument for ID; I'm just wondering how the preference can be articulated mathematically without imposing a desired end. topynate answered it pretty well (I'm assuming, because I'm terrible at math).
posted by shakespeherian at 3:37 PM on December 26, 2007

Time Cube all around!
posted by humannaire at 5:00 PM on December 26, 2007 [1 favorite]

Natural selection does not aim to select for a purpose other than survival...
In other words, entities aren't selected simply to survive...

I have a lot of trouble with the term natural selection because of the way it's used in sentences like this: selection indicates purpose of will... choice. But it's not a decision that's made, it's an event that happens. When talking with a creationist, using a phrase like nature chooses is only going to confuse the issue, because they'll think the argument is framed in such a way that there's some kind of intervention going on, the question is just what is intervening. But that's not the argument at all.
posted by Civil_Disobedient at 4:00 AM on December 27, 2007

This is awesome.

I'm guessing what happened with the three handed clocks winning out over the four handed, is that they were able to produce gear ratios that better matched the desired periods. The extra complexity in managing another hand accurately couldn't evolve fast enough and got out-competed on accuracy by three-handed clocks.
posted by Arturus at 7:49 AM on December 27, 2007

This would make a nice parable, à la Abbott's Flatland, wherein our intrepid hero, A. Watch, visits the respective lands of Sundials, Pendulums, and Two-, Three- and Four-Handed Clocks, briefly expounds on the sacred mystery surrounding the word IIII (as it appears on most clocks), and teaches us all a lesson on evolution and natural selection before being smashed to bits as a heretic.
posted by steef at 3:03 PM on December 27, 2007 [4 favorites]