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.
That was an invitation for you to elucidate.
In biology it is typically reproductive success. The same measurement can be made in EAs. What is your point?
If an organism reproduces it is fit and if it does not reproduce it is not fit, so the values “fitness” can take are 0 and 1? No.
So let’s ask then, how do we measure reproductive success? What exactly do we count?
Do we count the number of letters in the string that match the target phrase? Because that’s how the Weasel program assigns “fitness.”
No, as I made clear by talking about relative reproductive success.
Oh, we’re back on Weasel? I thought that had been done to death. That’s just a little pedagogical tool that demonstrates cumulative selection.
More generally, of course one counts the number of offspring if one wishes to measure biological fitness by that definition. Could you skip the silly questions and simply make your point, please?
My point is that you’re dissembling again.
I’d hoped to be able to make it without violating the rules, but since you insisted.
Reproductive success requires successful reproduction Patrick, otherwise it’s reproductive failure. So a measurement would have to be after the fact, not before any reproduction has or has not taken place.
If you assign a “fitness” of 0.2 in a GA to a genotype and this is used to give it a 20% chance at being included in the next generation, 80% of the time the genotype will not be included in the next generation, and the 0.2 “fitness” is not reproductive success at all in 80% of the cases. Reproductive failure is not reproductive success. You can’t measure nothing.
This explains the difference between an assigned fitness and a measured fitness.
It can be if you’re measuring the same thing. If you aren’t then it’s not the same measurement.
Patrick,
So what you are saying Patrick, is that any definition one wants to make about what is fitness in a population is appropriate in biology, is that it ?
Patrick knows what fitness is, he’s just confused about what it isn’t. That’s what happens when you don’t have a rigorous definition. Or definitions.
LMAO, the usual suspects still struggling with the most basic definitions
Go count the offspring that were never born dazz.
‘The Weasel program’? Which one? As you’ve certainly never actually written one.
You’re attempting to construct a narrative without evidence. Too bad, I thought perhaps NewMung was paying a visit.
If you are measuring fitness based on the common biological definition of reproductive success, yes.
It’s usually not a direct chance because calculation of the next generation population typically involves selecting two or more individuals in the current generation randomly then choosing which of those to reproduce either asexually or sexually. Individuals with higher relative reproductive success measures will tend to reproduce more.
No, it explains your equivocation on the term “fitness”. In an EA there is a measurement to determine how well an individual meets a particular criteria. This is called the “fitness function.” You are confusing, deliberately or otherwise, this term with the biological measure of fitness.
Please stop playing word games and address the actual concepts under discussion.
No, I’m saying it’s important to define one’s terms carefully and not equivocate or otherwise play word games. I know that’s very different than the rules for apologetics, but do try to keep up.
A wise man once said, you can’t play the blues unless the piano is blue.
You can’t simulate evolution unless the computer bits are actually fit.
If the computer doesn’t evolve it can’t be simulating evolution.
Everyone knows that.
Glen Davidson
Perhaps in real life biology. But in an EA it’s just highly unlikely.
Patrick,
Hence why I asked you for a definition, yet you were unable to provide one, other than saying it depends.
This is true if you are measuring reproductive success. As in counting the number of offspring actually present in the next generation.
It is not true if you are calculating “fitness” according to some criteria other than reproductive success, such as how close some string of characters is to some target phrase. That’s not fitness in the biological sense.
So far example when we see this comment in code:
// program stops when this fitness threshold is met or exceeded
Does it refer to when some genotype has left some threshold number of offspring as compared to its competitors? Say, 800 v 200 in a population of 1000?
Or does it refer to when you’ve found a perfect match of 28 characters to the target phrase?
#define FITNESS_THRESHOLD 28
Can we stop equivocating over the term fitness or at least admit it has a different meaning in different contexts?
I just wanted to repost this, because it is so entertaining.
That doesn’t make sense. Introducing probability to a matter of definition just makes you guilty of a category error.
The fuck is WRONG with you people?
This is simply hilarious.
Why not call it an objective function?
On any fair reading of what I have written, I was doing the opposite.
LoL.
Spend too much time at TSZ. Do you have Patrick on ignore? Go argue with him. I actually agree with you. We shouldn’t call it fitness if it’s not fitness.
That’s not at all how the conversation has gone. Here are our last two interactions:
The common measurement of biological fitness is reproductive success (the number of offspring in the next generation). What point are you trying to make?
So far we’re in agreement.
It’s still a measurement, but it is not the same measurement as biological fitness.
Weasel is not a good example of an EA. It simply demonstrates the power of cumulative selection compared to random selection.
That’s what I’ve been saying. Check my last few comments to you.
The word “fitness” is used in both measurements. In context each is clear. Again, what is the point you’re trying to make?
So there is no rigorous definition of fitness. But I guess no one ever claimed otherwise, so it’s all good.
All life is artificial life. There it is, in the first three minutes. LoL!
Thanks Richardthughes!
Not without considering the niche. Remember the niche!
Statements like the following probably turned people off:
Evolution is an algorithm. Regardless of substrate you get evolution.
When you design your evolutionary simulations …
I really liked the Karl Sims demo, especially the anthropomorphisms and the blatant admission that the “creatures” were “evolved” in order to do this or that.
Keep those vids coming!
The fact that someone makes use of a vague or different definition of fitness from the one used in biology, does not establish that there is no rigorous definition of fitness. All that would establish is that someone has failed to abide by it, or has abandoned it in favor of another for whatever reason.
The point is that it’s not “just an implementation detail.”
My point is that you’re equivocating.
https://github.com/polyworld/polyworld
Someone want to point me to where the “fitness function” can be found?
And there was this Patrick.
You said “it depends on the definition of fitness you are using”.
And you said the “two definitions” could yield different answers. So how many definitions are there for fitness?
Two?
Rumraket,
So its not an evolutionary algorithm if they use a definition of fitness that is not reproductive success, is that correct Rumraket?
Our name is legion.
Every EA can have a different “fitness function.” They can all define “fitness” to mean whatever they need it to mean. That’s what we mean by “objective.”
What I found intriguing was the claim that in Polyworld the fitness function was actually based upon the biological definition of fitness.
But no evidence has been offered in support of this claim.
Evolution is an algorithm. Didn’t you watch the video? If it’s evolution, it’s an algorithm. No “objective empirical evidence” or “rigorous definition” required.
If you have a population
and selection
and mistakes
You’ve got evolution. Regardless of substrate.
Have a word with yourself, Mung. Disingenuous. Disgrace.
That you are wrong whatever you say! Therefore ID.
Mung,
Not only is there no evidence, but instead they flat out admit that they couldn’t allow the one that reproduces the most to reproduce the most, because then you might end up with the least fit reproducing the most!
And Rumraket’s response to that is, that’s impossible! hahaha.
Chuckle. You win!
OMagain,
Maybe you should watch the video.
Obviously Richard and Rumraket haven’t.
Hey! Marginal progress!
So please explain why you don’t consider it an implementation detail and what you think that means for EAs generally.
On the contrary, the issue is that you are the one equivocating. You seem to have some problem with the word “fitness” being used in two different, although well-defined, contexts. Why exactly is that? What is your actual argument?
On the contrary, the issue is that you are the one equivocating.
Are we having fun yet?
Nope. And that doesn’t qualify as me equivocating over the term either. Do you have an actual example of me using the term in an equivocal manner?
That when you claim the same measurement is done in EAs you are equivocating, because it’s not the same measurement, it’s two different measurements going under the same name. That’s an equivocation.
One for each specific concept. What is important is those concepts, not the specific words we’ve agreed to use to label them. As long as we’re careful not to equivocate, there’s no problem.
Which concept are you interested in discussing?
Do you have examples of anyone making that claim?
Yet now we all know that when “fitness” is measured in a EA it is something completely different that is being measured.
Richardthughes is right. I should have watched the video rather than relying on what he said about the video.
Richardthughes:
Video: There is no fitness function.
You’re such a shitty quote miner. *golf clap*. The slide clearly says
“o No Fitness function
o Fitness is determined by natural selection alone”
Your brother keiths would be proud.