from 2007. Long but worth the watch. Let’s discuss below?
176 thoughts on “Polyworld”
Leave a Reply
You must be logged in to post a comment.
The Skeptical Zone
"I beseech you, in the bowels of Christ, think it possible that you may be mistaken."
from 2007. Long but worth the watch. Let’s discuss below?
You must be logged in to post a comment.
It’s ok Richardthughes. As Patrick so kindly pointed out, “fitness function” has a specific meaning in the context of an EA. Next time don’t say there’s a fitness function if there is no fitness function. I’m sure it was just an honest mistake.
What’s his point anyway? He’s a TE. He therefore agrees with the textbooks 100%. He defers to the experts where necessary. Evolution happened, his claim is merely that something additional is present, a something that has no entailments from this side of the curtain. So be it.
His ‘disagreement’ is on a higher plane. So don’t sweat the details Mung. You might embarrass yourself by fixating on irrelevant, inconsequential details while the bigger picture is missed.
OMagain,
He doesn’t understand it, but is sure it’s wrong, and agrees with it.
Richardthughes,
Find time to watch the video you posted yet Richard?
I managed to link to relevant sections randomly. Are you that stupid?
I love the part where he admits there’s often cheating in evolutionary simulations.
What, if there’s a fitness function there must be cheating?
But he thinks natural selection takes place inside a computer, lol.
So what would you call it, Mung?
You always were lucky, Lefty.
Mung,
Luck beats skill 😉
Un-Natural Non-Selection.
It’s amazing how people can admit everything else in the simulation is non-natural.
Non selection? So we should observe no increase in capabilities , then?
What do you mean an increase in capabilities? What capabilities? The ability to increase?
phoodoo,
Certain behaviors and traits.
Yup. That’s the only increase that matters.
Oh look, it looks like it is trying to swim. LoL!
What certain behaviors and traits?
You mean the certain behaviors and traits that the computer programmers design into their program to search for-which is exactly why it has been explained to you why computer algorithms have nothing to do with blind evolution.
Every computer algorithm, even the ones Joe claims don’t search, but which he can’t name, does the opposite of blind evolution. It sets a goal, and then decides how to obtain it.
Tough for you to understand the not so subtle difference, but some people reading this might.
Phoodoo, Mung. How does “the ability to increase” happen? Are there multiple strategies? How does it happen in a dynamic environment?
Oh, wait.. design!
No not at all. Nor would you see “create swarming behavior” explicit in the code. But, I bet you haven’t actually watched the video. Because that is hard for you
Yes, EA’s are designed. Who ever thought otherwise?
Polyworld was designed to not repeat the cheating of it’s predecessors. It’s true.
Category error.
If their complaints had substance they’d be able to find parallels in the literature.
Unless, of course, nobody up till now has been as sophisticated as Mung and Phoodoo and not actually had the ability to point out these issues in such an erudite and informative manner.
Mung, it would not surprise me in the least if the Discovery Institute and the people at Biologic would be interested in a paper from you on EA’s and cheating. From looking at their output I’d say you were more then up to the required technical standard.
Yes Frankie, that’s correct.
The word “fitness” is used in different contexts in EAs, but it is always well-defined within that context. Unless you have actual examples of equivocation on this term, I don’t see your point.
First, you’re wrong about “every computer algorithm”. Polyworld and Tierra are examples of evolutionary algorithms without any goals other than survival.
Second, you seem to have a fundamental misunderstanding or lack of knowledge of EAs. Most work something like this:
1) An environment is created.
2) A random population of virtual organisms is created, each with a genome that affects its behavior in the EA’s environment.
3) The relative reproductive success of each individual is measured. This can be done explicitly through a fitness function or by allowing the organism to interact with the environment as in Polyworld and Tierra. This models how the phenotype of a biological organism works in the real world.
4) The next generation population is created from the current generation. Those organisms with higher relative reproductive success odds will tend to leave more offspring in the next generation than those with lower relative reproductive success odds. It’s typically stochastic, though, so there are no guarantees. This models natural selection in the real world.
5) Go back to step 3 with the new population or stop for whatever reason.
Note that, as Lizzie emphasized, this is actual evolution. It isn’t a simulation of evolution, it is an example of it.
Note also that it isn’t a search. As someone else here has pointed out, the person running the EA may be searching for something but the EA itself is just evolving.
Finally, note that the behaviors and traits that evolve are not designed in by the programmers of the EA. Quite surprising results happen.
(And a side note for Mung: It is possible to measure biological fitness at step 5, if one wishes.)
You should play around with some EAs before making patently incorrect claims about them.
Patrick,
Hm, Patrick, you have really given me a challenge with this one. How do I respond to your incredible lack of understanding in this post without insulting you personally. Dam, you have made this one tough!
You believe that Polyworld is an example of a computer program whose only constraint is to see which organism can reproduce the most. Holy Cow Patrick, you, you…no, I can’t.
Ok, this is what you believe Patrick. You got me. I simply can’t respond under the rules of this forum. Whew!!
Phoodoo, read what people say rather than tell them what they think. It’s a shitty habit you have and I’m sure contributes to your unlikeability.
You could simply point out how what Polyworld and Tierra are doing qualifies as a search. That would require clearly defining your terms and analyzing the code. Clueless mocking is certainly less effort.
Patrick,
+1.
Biological fitness requires, erm, a biological.
phoodoo +1
If the definition of “biological fitness” is “the number of offspring present in the next generation” then it is easily measured in EAs.
This is why definitions matter, Mung. Resist the equivocation!
The definition of biological fitness can be applied to non-biologicals. And you accuse me of equivocation. LoL.
Patrick,
In biology, biological fitness requires nothing more than existing.
Try again. But try harder.
OMagain,
Sorry, I am straight.
No offense, I don’t judge.
Subject matter is simply beyond Mung and Phoodoo.
Instead of mindlessly mocking, perhaps you could describe why it could not. I realize that would slow down your seagull posting, but that wouldn’t be a bad thing.
No, it requires reproducing.
Patrick,
A ha, I see. So in order to measure fitness of a particular genotype, you don’t measure those that exist in a population. You only count those that have already had offspring. So one’s fitness status changes overnight then?
What if their offspring are sickly and only last a day, then does that constitute biological fitness, or is the individual that had offspring that subsequently dies, put back in the unfit category?
The definition of biological fitness that Mung and I are discussing is the number of offspring in the next generation. If you feel there is a better measurement, you can define that. The important thing is to not equivocate.
Patrick,
So which is it Patrick, are you counting the offspring, or are you counting the ones who have had offspring? Its either one or the other.
So you either are counting the ones that exist (like I said) or you are counting how many offspring, ones that already exist have? If its the latter, does it matter if the offspring of the ones you are calling fit only survive a day? Does it matter if the offspring of the ones you call fit never reproduce?
The definition of “biological fitness” under discussion in this subthread is “the fitness of an individual in an ancestral population is the number of offspring from that individual in the subsequent generation.”
Other definitions are possible. What is important is that the definition is agreed by all parties and used consistently.
So, what’s your point?
Patrick,
My point is other definitions are NOT possible under the theory of evolution. The ONLY requirement for being fit in a population is you exist. You count those that exist.
You don’t check the sperm count of those that exist. You don’t don’t give fertility tests. You don’t wait to see if one is attractive to the opposite sex. The gene exist therefore it must be because it was fit enough to exist. Are worker ants fit? Are worker bees? Are lone wolves?
Fit means existing in biology. That is it.
Because that is the dumb theory your side has created. Like Sal said, dumb and fast, might beat smart and slow. You can’t know until it no longer exists.
Patrick,
What is all of the offspring of the fittest individual are sterile? Are they still the fittest?
That assertion is contradicted by the fact that I provided a commonly used definition that does not consist solely of “it exists.”
By the definition we’ve been using in this subthread, the fittest individual in generation N is the one with the most offspring in generation N + 1. You can certainly create a definition that includes offspring in generations greater than N + 1. It might even be useful. Just be careful to avoid equivocation.
Patrick,
Its not contradicted if you are wrong Patrick. If you make up a definition that makes no sense, no one has to accept your definition.
Again I ask, if what you call the fittest (the individual with the most offspring) only reproduce sterile offspring, are they still the fittest?
Patrick,
What do you mean by “we’ve” been using? Who has been using that definition besides you?
From the first link returned by Google for “biological fitness”
Biological fitness, also called Darwinian fitness, means the ability to survive to reproductive age, find a mate, and produce offspring.
Basically, the more offspring an organism produces during its lifetime, the greater its biological fitness.
(Emphasis mine.)
As I said, it’s a very common definition.
Patrick,
First off that definition says “the ability to” it doesn’t say anything about the emphasis that you decided to tack on. So at what point is an individual fit, when they have the ability to reproduce or when they actually reproduce? What if they have the ability to reproduce but a tree falls on them and kills them? That means they were never fit?
What if another individual has many offspring, all of whom have severe birth defects and are unable to reproduce? They are more fit, definitionally, than the one that had the ability to reproduce but had the tree fall on them?
And what are we to make of all the worker bees and ants, and mules of the world. They are all unfit?