Evidence God Is

[nPeace]

@TagliatelliMonster
You said...
So, what does the programmer actually model in this algorithm?
1. the chromosome. The "genotype". And that in terms of structure, not in terms of encoding. The chromosomes of generation 1 could essentially be generated randomly. Usually though, in practice, they'll start with "the best" design that the human engineers came up with and optimize that one with the algorithm. But it will also work if you start with randomized strings - it will just take longer for the algorithm to complete as a lot more generations will be required.

So there is no possibility of anything being destroyed?
So you know that none of these generations will destroy anything, to the point you have to start with a new genotype?
So the more generations used do not have a significant impact on the success of accumulated beneficial mutations?

For one thing, the programmer assumes he gets to the point of "the best".... thus he programs for the best.
You said it.

 

[TagliatelliMonster]

Okay. How long does it take for evolution to take place in one of these algorithms. Can I have a look at your best one. Just supply a link to it. Thanks.

I'll spare you the technical models that we used back in the day when I worked on that project. I actually also wasn't on the team that did the module - we had experts for that who specialised in AI, GA, Search, machine learning etc.

I'm more of a business developer / analist and so was more involved with UI design, modeling the database and analysing business requirements. My code used their module by calling the API's they provided.

But here's a fun and more easy to comprehend showcase of such a GA implementation:

BoxCar2D

You can read more about the algorithm implementation this app uses here:
BoxCar2D: About

It runs in Flash, which today is blocked by most browsers. So you might have to explicitly allow access for the Flash plugin to activate. I also recommend running this on a desktop, or a laptop that is plugged in, because it will likely suck away battery power.

So, what does that program do?
It evolves cars for a given race track.

Generation 0 is a randomized mess of polygons with a random amount of wheels attached. The wheels spin at random speeds and are attached to the polygons with random amount of "force". So if the force is too light, the wheel will drop off when hitting a bump or whatever.

Each "individual" in the population will run the track once. The fitness test will measure distance traveled and speed, and give it a score. The top performers are then used to produce a new generation, with randomize mutations. The polygons that were mutated, will be given a new color if I remember correctly.

You can leave the website open for 30 to 60 minutes. When you return, you'll see highly optimized cars flying over your screen, most of which will be completing the track with good speeds.

What I find pretty interesting in this application, is the fact that you can suddenly change the track. There's this one track that has "rubble" on it.

I once left it open for like 3 hours and the cars clearly reached a local optimum. ie: they were fully optimized for the default track with very little room for further improvement left. Then I switched it to the rubble track. This has heavy influence on the selection parameters. Many of the cars could no longer complete the track as they got blocked by the rubble.

I left it open for another hour. When I returned, the cars had evolved into some kind of plower. It was really cool. A sharp angled polygon was added to the car's front and it took on the function of a plow. All the rubble just got grabbed by that plow and thrown in the air, clearing the path, with very little effect on the car's speed.


But the absolute coolest thing I've seen evolve on there, was literally a detachable part. It was a track with a lot of bumps. One of these bumps was really sharp, with some kind of ramp right before it. The best performer there, had an additional polygon attached right in front of the front wheels. When going on the ramp, that polygon touched the ground and broke off due to the impact force. The cool part: this caused the car to lift when driving over the piece that broke off. It has some kind of catapult effect. This "jump" enabled the car to overcome that sharp bump. Without that polygon, it couldn't overcome it.

It was fun to see

 

[TagliatelliMonster]

@TagliatelliMonster
You said...
So, what does the programmer actually model in this algorithm?
1. the chromosome. The "genotype". And that in terms of structure, not in terms of encoding. The chromosomes of generation 1 could essentially be generated randomly. Usually though, in practice, they'll start with "the best" design that the human engineers came up with and optimize that one with the algorithm. But it will also work if you start with randomized strings - it will just take longer for the algorithm to complete as a lot more generations will be required.

So there is no possibility of anything being destroyed?
So you know that none of these generations will destroy anything, to the point you have to start with a new genotype?

I don't know what you mean with "destroyed".
If you mean going extinct, then no. You could do it by for example also programming in a chance that an asteroid will destroy the room in which the piping is being optimized, but what would be the point of that?

You could also implement a "minimum score" and simply "kill" all individuals that score less then that. So if an entire generation doesn't achieve the minimum score, there would be nobody left to produce a new generation. But again: what would be the point of that? The point of the GA is to look for solutions, after all.

In real life, not reaching the minimum score would indeed mean the end of the line: it won't survive to reproduce. But GA's aren't real life. They merely use the principles of biological evolution as an optimization module and / or search heuristic.

Extinction is the result of radical environmental change with the population's evolutionary pace not being able to "keep up" with these environmental changes. Such changes can be any number of things: new natural enemies evolving and/or migrating into the habitat and the population not having an evolutionary answer fast enough. Or indeed natural disasters or rapid climate change like the start or end of ice ages, etc.


If you mean "harmfull mutations" causing a result that is actually worse then the previous generation: sure, that can and does happen. Those will be discarded quickly though, as they won't find themselves in the top X performers and won't be selected for producing a new generation.

So the more generations used do not have a significant impact on the success of accumulated beneficial mutations?

Off course they do.... the more generations it runs, the more time opportunities there are for accumulation of improvements. This goes on until the population reaches a "local optimum". The "local optimum" is a genotype state from which there are no more easily accessible (or even possible) gradual evolutionary pathways towards further improvement. This is essentially what punctuated equilibrium is all about. Which means that if the selection pressures (= the total sum of environmental parameters) stay stable, then there comes a point where natural selection will favour the status quo. Pace of evolutionary change at this moment slows down, or even comes to a halt.

When the selection pressure change again, the local optimum moves and new evolutionary pathways towards further optimization open up. Selection now no longer favours the status quo by default.

In the Boxcar2d example I gave you, this would be the moment you could change the track, which would turn selection pressure on its head. The cars that were optimized for the track they evolved on, will no longer be fully optimized for the new track.

For one thing, the programmer assumes he gets to the point of "the best".... thus he programs for the best.
You said it.

Congratulations.

You managed to completely misunderstand everything I said, it seems.

Eventhough I really took the time to give you a comprehensive explanation of how all a GA programmer does is implement evolutionary logic and model an environment in which this logic is then applied, which results in optimized designs for that environment and where these designs are entirely the work of the blind algorithm, not the programmer.


I mean............ think about it......
If it was indeed the work of the programmer.... then why even go through the immens and costly trouble to build this GA algorithm and the entire simulation engine (which can be EXTREMELY complicated), only to end up with a design that the programmer was already programming against from the beginning?

It makes no sense at all.

The entire reason why people use GA's, is precisely because the GA's themselves are capable of coming up with better solutions then the human engineers (or programmers).

Please think it through next time.

 

[nPeace]

@TagliatelliMonster thanks for the link. I won't allow flash at all, so I won't see that one. I will get back to you when I have more time.

 

[TagliatelliMonster]

@TagliatelliMonster thanks for the link. I won't allow flash at all, so I won't see that one. I will get back to you when I have more time.

Apparantly there's an html5 version these days also

HTML5 Genetic Algorithm 2D Car Thingy - Chrome recommended

Don't know if it's from the same engineer, nore have I read about the algorithm on that one. I literally just found it.
It looks very alike though, so it seems as if it is just a new html5 version

 

[nPeace]

Apparantly there's an html5 version these days also

HTML5 Genetic Algorithm 2D Car Thingy - Chrome recommended

Don't know if it's from the same engineer, nore have I read about the algorithm on that one. I literally just found it.
It looks very alike though, so it seems as if it is just a new html5 version

Is this supposed to teach me what algorithms are? Then you just wasted a post. Sorry. Does nothing for me, and addresses nothing in my post, apparently.
As I said, I'll get back to you.

BTW thanks for trying.

 

[TagliatelliMonster]

Is this supposed to teach me what algorithms are? Then you just wasted a post. Sorry. Does nothing for me, and addresses nothing in my post, apparently.
As I said, I'll get back to you.

You asked for an example, I gave you one.

If it doesn't interest you, don't ask for examples.

 

[nPeace]

You asked for an example, I gave you one.

If it doesn't interest you, don't ask for examples.

Forgive me for failing to communicate properly.
Could it be this is the same problem you had when you read the Bible - a communication problem?
I remember saying exactly this...
How long does it take for evolution to take place in one of these algorithms. Can I have a look at your best one
To me, that doesn't read, "Give me an example." Perhaps skeptics and Atheist read thing different to believers.

 

[TagliatelliMonster]

Forgive me for failing to communicate properly.
Could it be this is the same problem you had when you read the Bible - a communication problem?
I remember saying exactly this...
How long does it take for evolution to take place in one of these algorithms. Can I have a look at your best one

If you fail to communicate properly, then the fault doesn't really lie with me, now does it?

Evolution takes place in every generation, as every generation is subject to a fitness test after wich the best performers are used to produce a new generation while the others are discarded.

To me, that doesn't read, "Give me an example."

You said "can I have a look at one?" where "one" refers to a GA.
If that doesn't mean that want an example of such an algorithm to look into it, then I don't know what you are saying.

Perhaps skeptics and Atheist read thing different to believers.

Is that why there are thousands of denominations of christians who all read the same book, yet all walk away with beliefs different enough to classify them as a different denomination?

 

[nPeace]

If you fail to communicate properly, then the fault doesn't really lie with me, now does it?

Evolution takes place in every generation, as every generation is subject to a fitness test after wich the best performers are used to produce a new generation while the others are discarded.



You said "can I have a look at one?" where "one" refers to a GA.
If that doesn't mean that want an example of such an algorithm to look into it, then I don't know what you are saying.



Is that why there are thousands of denominations of christians who all read the same book, yet all walk away with beliefs different enough to classify them as a different denomination?

Okay. So you read my two statements and totally misunderstood, and want me to take the blame.
I understand the kind of person I am speaking with. So I have said all that I need to say on this.

So did you give me your best "example"... which it is clear from my post, I did not ask for ?

 

[Valjean]

Okay. How long does it take for evolution to take place in one of these algorithms. Can I have a look at your best one. Just supply a link to it. Thanks.

Are you asking for examples of evolution or of the natural selection mechanism/algorithm?

 

[nPeace]

Are you asking for examples of evolution or of the natural selection mechanism/algorithm?

I'm asking for the natural selection w/ mutations algorithm. Thanks.

 

[nPeace]

Right. I saw no "cars" that grew wings an flew along the track. Nor did I see any wheels become electric motors allowing the "car" to be evolve to a hover board... and there were no jet packs.
Just the same old "box cars" repeating a process over and over again.
Sort of like what we expect if the programmer did not include code for those.
If he did though, what would that tell us....

I don't know what you mean with "destroyed".
If you mean going extinct, then no. You could do it by for example also programming in a chance that an asteroid will destroy the room in which the piping is being optimized, but what would be the point of that?

You could also implement a "minimum score" and simply "kill" all individuals that score less then that. So if an entire generation doesn't achieve the minimum score, there would be nobody left to produce a new generation. But again: what would be the point of that? The point of the GA is to look for solutions, after all.

In real life, not reaching the minimum score would indeed mean the end of the line: it won't survive to reproduce. But GA's aren't real life. They merely use the principles of biological evolution as an optimization module and / or search heuristic.

Extinction is the result of radical environmental change with the population's evolutionary pace not being able to "keep up" with these environmental changes. Such changes can be any number of things: new natural enemies evolving and/or migrating into the habitat and the population not having an evolutionary answer fast enough. Or indeed natural disasters or rapid climate change like the start or end of ice ages, etc.

Not extinction.

If you mean "harmfull mutations" causing a result that is actually worse then the previous generation: sure, that can and does happen. Those will be discarded quickly though, as they won't find themselves in the top X performers and won't be selected for producing a new generation.

Bingo.
That's quite a significantly low "top X performers", considering we have the harmful + the neutral overweeningly significant.
Discarded makes it sound so simple, doesn't it? How do you discard a mutation?
As far as I know, these are changes - whether harmful, neutral, or beneficial. They don't go anywhere, even if they are not entirely crippling.

Off course they do.... the more generations it runs, the more time opportunities there are for accumulation of improvements. This goes on until the population reaches a "local optimum". The "local optimum" is a genotype state from which there are no more easily accessible (or even possible) gradual evolutionary pathways towards further improvement. This is essentially what punctuated equilibrium is all about. Which means that if the selection pressures (= the total sum of environmental parameters) stay stable, then there comes a point where natural selection will favour the status quo. Pace of evolutionary change at this moment slows down, or even comes to a halt.

When the selection pressure change again, the local optimum moves and new evolutionary pathways towards further optimization open up. Selection now no longer favours the status quo by default.

In the Boxcar2d example I gave you, this would be the moment you could change the track, which would turn selection pressure on its head. The cars that were optimized for the track they evolved on, will no longer be fully optimized for the new track.

Let's beg for many "ifs".

Congratulations.

You managed to completely misunderstand everything I said, it seems.

Eventhough I really took the time to give you a comprehensive explanation of how all a GA programmer does is implement evolutionary logic and model an environment in which this logic is then applied, which results in optimized designs for that environment and where these designs are entirely the work of the blind algorithm, not the programmer.


I mean............ think about it......
If it was indeed the work of the programmer.... then why even go through the immens and costly trouble to build this GA algorithm and the entire simulation engine (which can be EXTREMELY complicated), only to end up with a design that the programmer was already programming against from the beginning?

It makes no sense at all.

The entire reason why people use GA's, is precisely because the GA's themselves are capable of coming up with better solutions then the human engineers (or programmers).

Please think it through next time.

Wait.

they'll start with "the best" design that the human engineers came up with

I must not be reading too well.
The human engineer comes up with a design which he / she considers the best to start with, but has mot programmed his little program to start with an assumption that his best will be reached. ... not to mention, all the thousands of times it is ran.
Oh,,, my... head hurts.

 

[Valjean]

Right. I saw no "cars" that grew wings an flew along the track. Nor did I see any wheels become electric motors allowing the "car" to be evolve to a hover board... and there were no jet packs.
Just the same old "box cars" repeating a process over and over again.
Sort of like what we expect if the programmer did not include code for those.

Cars don't reproduce with variation. You really don't understand the mechanisms of evolution.
The mechanism of natural selection does not apply to manufactured items.

Discarded makes it sound so simple, doesn't it? How do you discard a mutation?

How did you not learn this in school? The mechanism for "discarding" a harmful mutation or reproductive variant is simplicity itself.
If you don't understand even the basics of natural selection, how can you have a valid opinion on this subject?

As far as I know, these are changes - whether harmful, neutral, or beneficial. They don't go anywhere, even if they are not entirely crippling.

What are you talking about? "Don't go anywhere?" If they confer improved survival within a given environment; if the variants are more reproductively successful, the trait will increase in the general population.
I mentioned the peppered moths in a previous thread. Do you not understand the mechanism they illustrate?

The human engineer comes up with a design which he / she considers the best to start with, but has mot programmed his little program to start with an assumption that his best will be reached. ... not to mention, all the thousands of times it is ran.
Oh,,, my... head hurts.

Natural selection comes up with new models with every generation -- all random. The variation is random. The selection is not.
Natural selection selects the best fits. Don't you understand this mechanism? It has nothing to do with human engineering. It's totally different.
If you don't understand the mechanism you can't have a valid opinion on it. It obviously works, you're not a bacterium.
Oh,,, my... head hurts.

 

[nPeace]

Cars don't reproduce with variation. You really don't understand the mechanisms of evolution.
The mechanism of natural selection does not apply to manufactured items.

Oh. How am I supposed to know cars don't reproduce with variation? Thanks.
So right. Since I didn't know the above, how can I understand the mechanisms of evolution.
The mechanism of natural selection does not apply to manufactured items, yet I am given algorithms of manufactured items to demonstrate the mechanisms of evolution.

How did you not learn this in school? The mechanism for "discarding" a harmful mutation or reproductive variant is simplicity itself.
If you don't understand even the basics of natural selection, how can you have a valid opinion on this subject?

No. I did not learn this in school. Never heard of it until now.
I'll go learn about the basics of national selection, and see what I am missing.

What are you talking about? "Don't go anywhere?" If they confer improved survival within a given environment; if the variants are more reproductively successful, the trait will increase in the general population.
I mentioned the peppered moths in a previous thread. Do you not understand the mechanism they illustrate?

And if they don't, where do they go?

Natural selection comes up with new models with every generation -- all random. The variation is random. The selection is not.
Natural selection selects the best fits. Don't you understand this mechanism? It has nothing to do with human engineering. It's totally different.
If you don't understand the mechanism you can't have a valid opinion on it. It obviously works, you're not a bacterium.
Oh,,, my... head hurts.

I could understand. Communication can be hard sometimes. I think listening is an art - whether spoken word, or written word, which we are not born with. It is learned.
I don't believe I said anything as you are suggesting here. I think 'best fit' and 'best design' are two different things. Do you? Fit and design don't appear to even match.

 

[Subduction Zone]

Cars don't reproduce with variation. You really don't understand the mechanisms of evolution.
The mechanism of natural selection does not apply to manufactured items.
How did you not learn this in school? The mechanism for "discarding" a harmful mutation or reproductive variant is simplicity itself.
If you don't understand even the basics of natural selection, how can you have a valid opinion on this subject?
What are you talking about? "Don't go anywhere?" If they confer improved survival within a given environment; if the variants are more reproductively successful, the trait will increase in the general population.
I mentioned the peppered moths in a previous thread. Do you not understand the mechanism they illustrate?
Natural selection comes up with new models with every generation -- all random. The variation is random. The selection is not.
Natural selection selects the best fits. Don't you understand this mechanism? It has nothing to do with human engineering. It's totally different.
If you don't understand the mechanism you can't have a valid opinion on it. It obviously works, you're not a bacterium.
Oh,,, my... head hurts.

I am lucky. He could not handle my constant refutation of his claims so I am "temporarily on ignore".

 

[Subduction Zone]

Natural selection is almost a tautology. Something so obviously true that it is hard to even come up with a way to test it. Natural selection only says that the individuals best suited to an environment are the ones most likely to pass on their genes. Let's take some extreme examples. We have three antelopes on the Serengeti. One has three legs and cannot run that fast, one has five and tends to trip itself, one has four. On a plain full of lions, which one is most likely to pass its genes on?

 

[TagliatelliMonster]

Okay. So you read my two statements and totally misunderstood, and want me to take the blame.

Instead of engaging in this peeing contest, perhaps you could just clarify what you meant instead.

I think my answer and interpretation of your question was rather reasonable given how you worded it.

Instead of simply repeating that I didn't understand it correctly, perhaps just clarify instead. That would be more constructive, wouldn't you agree?

I understand the kind of person I am speaking with. So I have said all that I need to say on this.

You said nothing at all. You threw out some accusations and claims and are unwilling to support them and clarify.

So did you give me your best "example"... which it is clear from my post, I did not ask for ?

Any proper GA implementation is as good as the next one.
What I gave you is one proper implentation of a GA and simplistic enough to understand it for laypeople.

I could also point you to expontially more complicated examples, but in those cases you'ld have to take my word on it.... Because I don't have access to those programs, considering companies have paid millions of dollars to have them developed. Like Boeing who used GA's to optimize fuel distribution systems in one of their airplanes. I never saw that program and the code is completely proprietary. I merely read a case study of it in a journal.

(ps: you might also ask yourself why companies are prepared to pay millions of dollars to have a GA developed to optimize their products.... Could it perhaps be because the ROI is worth it because GA's actually work?)


Having said all that, it is completely unclear to me what exactly you are asking an example of.
It seems to not just be an example of a proper implementation of GA, because I gave you an example and you rejected it.


So unless you **mod edit** actually clarify your statements instead of engaging in a peeing contest, I can not help you further.

 

[TagliatelliMonster]

I'm asking for the natural selection w/ mutations algorithm. Thanks.

That makes no sense.
Mutations is an intrinsic part of evolution as much as the selection process is.

Without mutation, there is no evolution.
Without selection, there is no evolution.


Evolution occurs when you have systems that reproduce with variation (=mutation), are in a struggle for survival and compete with peers over natural resources.

You need mutation and you need competition.

 

[TagliatelliMonster]

Right. I saw no "cars" that grew wings an flew along the track. Nor did I see any wheels become electric motors allowing the "car" to be evolve to a hover board... and there were no jet packs.
Just the same old "box cars" repeating a process over and over again.
Sort of like what we expect if the programmer did not include code for those.
If he did though, what would that tell us....

These cars don't evolve wings and jet packs for the same reason that living beings won't evolve plastic legs: the nature of genotype doesn't allow for it.

It has nothing to do with what the "programmer" is programming for and everything with what is possible (and impossible) in context of the nature of the genotype.

Bingo.
That's quite a significantly low "top X performers", considering we have the harmful + the neutral overweeningly significant.

Not really.
For example, suppose the top performer scores a 9.
A neutral mutation would keep it at 9...
A beneficial mutatoin might push it to 9.2
A harmfull one might lower it to 6.

The 9's and the 9.2's would be selected. The 6 wouldn't.
What's the problem?

Discarded makes it sound so simple, doesn't it? How do you discard a mutation?

You just don't select the individual with the mutation to produce a new generation. It sounds simple, because it is simple.

In real life, an individual with a harmfull mutation likely won't survive till breeding age. And even if it does, it likely wouldn't succeed in finding a mate.

The mutation thus will be discarded.
What's the problem?

As far as I know, these are changes - whether harmful, neutral, or beneficial. They don't go anywhere, even if they are not entirely crippling.

Beneficial and neutral ones have a much higher chance of being passed on to offspring. So those are going places. Harmfull ones have little to no chance of being passed on to offspring, because the individuals that carry them won't be reproducing. So those mutations are discarded. They don't go anywhere.

Again: what's the problem?

This isn't exactly rocket science.......

Let's beg for many "ifs".

No clue what you mean by this in response to the quote you are responding to.
What "ifs" are being begged and what do you mean by that?

I must not be reading too well.

Yes, that's been clear for quite a while now.....

The human engineer comes up with a design which he / she considers the best to start with, but has mot programmed his little program to start with an assumption that his best will be reached. ... not to mention, all the thousands of times it is ran.
Oh,,, my... head hurts.

The practical application of GA's in the industry concerns optimization of existing systems.
They are used to solve design problems. This implies that you have a design with a problem that you can't solve, or of which you aren't sure that it can be improved. A GA will then be implemented to run on top of those designs.


When Boeing used a GA to optimize their fuel distribution system in one of their plane models, they didn't have the GA evolve an entire airplane from scratch.

The input of generation 0 was the airplane design they had and which they were trying to improve / optimize.