General > Biology

Selfish Gene

<< < (2/4) > >>

Numsgil:

--- Quote from: deoxymoron ---All these higher level competitions could theoretically start occuring naturally?
--- End quote ---

Yes and no.  If you google search some variation of "haystack" "evolution" "altruism" and "game theory", you should fine some interesting stuff.  I have a link I could post Mondayish, if I remember.

Basically, think of it like this: the complexity of chemistry is not apparent from the simple() rules of quantum mechanics.  Likewise the complexity of genes is not apparent from the rules of chemistry.  Likewise the complexity of selection on larger orders is not apparent from the rules which govern the success/failure of genes.  At each level another level of complexity is created in the hierarchy above it from the interactions at that level.

So we could say that all evolution represents the conflict between genes, but we could also say that it's a natural consequence of the laws of quantum mechanics.  Both are true, but both miss a great deal of the story.  There's a reason physicists aren't licensed to practice medicine.

But these higher levels aren't guaranteed to exist.  The laws of chemistry are a bit immutable, but the laws that govern life is a little more plastic.  Under certain circumstances (this is where game theory comes in), the selection of groups outweighs the selection of individuals.  

So if you want to evolve bots, the question becomes how to form an environment that is able to support additional emergent levels of complexity.  It's a bit like being a gardener.  If I, as a programmer working on the program, just blithely assume that greater and greater complexity will result if I have self replicating entities, I'm either going to be extremely lucky or rather frustrated.


--- Quote ---For example, one thing that may be hampering 'evolution', is the way the gene's are effective in dawinbots. for example, in Dawkin's understanding of a gene, gene's overlap and a reason this could not really be implemented into dawinbots is because, the genes instruction consists of words, not four basic chemical acids. i realise it would be impossible/incredibly difficult in trying to mimic gene's exactly in essence as they behave in reality. lol
--- End quote ---

Actually, when you really get in to it, DNA is remarkably like a computer program.  It's written for a massively parallel computer, which is different from modern serial computers (and Darwinbots DNA), but at it's core it's a symbol manipulating machine.  It's just a symbol manipulation machine whose source code is absolute spaghetti code, with all variables global.  Imagine a 1 billion line program written by a script kiddie mixing and matching lines of code from all sorts of different programs.

If I had to decide exactly, I'd say that the main differences between DNA and computer code is that:

1.  Life is essentially a quantum computer (things are inherently probabilistic)
2.  Each base pair of DNA is also a processing element, or can spawn a processing element.  So the computational power grows as the complexity of the algorithm does.


--- Quote ---Btw, i sort of agree with what you say about dawkin's militant group, though this too could be seen in evolutionary terms? The tendancy for humans to form their views based on majority rules, laced with a splash of rational thinking (seemingly never enough) because this is what worked in the past. And when i say in the past, i mean, way back in the stone age. I'd prefer a bunch of dawin advocates running aroun than creationists telling me earth is 4000 years old, and if i think differently, i'm going to hell. it seems more a fight between ignorance and rationalism. while climate debate has bureaucratize into crap, politically, and most people don't really understand it (including me) but are willing to have such concrete positions. :/
--- End quote ---

Actually that's an interesting way to view it.  I was thinking like creationists et al are like an irritant and scientists are having a systemic allergic reaction.  But you could probably view it from an evolutionary stand point as well.

...

For your efforts at getting bots to evolve interesting behavior...  For zerobots, you'll need to play gardener with them.  Shepherd bots, restarting from time to time with promising strains, etc.  There have been some good results done like that.  I've been trying to do a 0 intervention zerobot sim for a few months and tens of millions of cycles but I've gotten nothing so far.

It's actually a good case for a god: without occasional (non-Armageddon) disruption, things stagnate.  Life on Earth was not more complex than bacteria for a billion years.  Along the same vein, this chart makes interesting food for thought.

deoxymoron:
Hm, i think we have roughly the same picture of evolution, i guess. Probably just seeing it differently. i'm sure you have done your research a bit more thoroughly than me though. lol, by the looks of it anyway.

Although, about the absolute reasons for life occurring in nature, I'm strongly leaning towards these main ideas in shaping my thoughts:


1) The universe is deterministic.

I think Plancks quanta just means we can't be certain, not that the universe is actually uncertain to a degree. Although ironically I'm the least certain of this idea as some things about quantum mechanics aren't completely clear to me, yet.


2) Life is just a product of the laws of entropy. (At least a good way to view it)

So most of the time, this law is used to describe simple happenings like hot/cold water becoming warm when mixed. In the case of 'life', it has favoured a completely opposite direction of complexity (insert understanding of evolution here). The sun's rays is the 'hot water' in a way. I think I have a pdf. paper that explains this if you're interested? (should be attached)


3) Evolution favours the gene, always.

Though i understand you're point that altruism and group selection do occur. but often, thinking in terms of groups leads to people thinking that evolution thinks for it's self. This is why i asked if you had read Dawkin's book.. (dw, there's no anti-creationist rants in there)

So for example, a man might save his 3 kids in full knowledge that he will die in the process. no he's not a 'mutant' defective person (speaking in evolutionary terms ). He does this because all his behaviour is only occurring to increase the population of his gene's, at least that particular gene that caused him to save his 3 kids. There are now a 150% (on average) increase of those gene's in the gene pool that caused him to die saving his kids, and thus, they were successful gene's.

This all being said, i'm sure dawinbots hold a lot of promise, even if it's somehow used in the future for something completely different, like developing bahaviours for the enemy in pc/ps3 games?


At the moment i haven't even attempted creating my own bot, i've used I flammas and others with the algae. i've just started using bots from this site. I'm thinking if i can create different environments (with the shapes) in different areas on the map, maybe different species will develop in each area giving them a 'safe zone' then allowing the 'in between' zone to be fought over. constant fighting between two would then lead to favour on complex bot, as simple behaviour could be exploited (not sure, maybe, if they can remember another bots likely bahaviour. probably) The more complex behaviour, met only with even more complex behaviour? well hopefully!

I found a good way to create an environment is to use lots of tall/thin and wide/short shapes crossing each other. Perhaps they will just learn the environment, even so, that would be a good outcome.


BTW thanx Houshalter for the tips!



--- Quote ---It's actually a good case for a god: without occasional (non-Armageddon) disruption, things stagnate. Life on Earth was not more complex than bacteria for a billion years. Along the same vein, this chart makes interesting food for thought.
--- End quote ---

It would be funny if you did come to this conclusion after all you're work on dawinbots lol

ikke:
As stated earlier on this forum I am more a Dawkins adept. On boh counts actually. On evolutionary biology I think Dawkins has a convincing case against group evolution as an independent driver. Basically his argument is that of evolutionary stable strategies. If there is a group where doing the benefit of the group outweighs that of the gene is is open to invasion of a parasitic gene usurping the group, with the group not being able to respond.
The god delusion sums up his stance on religion pretty eloquently. I have seen the dawkins vs lennox debate and he didn't do as good. He is not a debater and he got succerpunched by the set up of the debate:
the mediator would give a quote and ask dawkins to expand and then let Lennox hammer away against it. this left dawkins having to split his time between explaining the quote and lennox attack, where lennox only had to demolish. With dawkins familiarity on game theory he should have looked at the rules. He could have predicted the outcome.

Numsgil:

--- Quote from: deoxymoron ---1) The universe is deterministic.

I think Plancks quanta just means we can't be certain, not that the universe is actually uncertain to a degree. Although ironically I'm the least certain of this idea as some things about quantum mechanics aren't completely clear to me, yet.
--- End quote ---

Just the opposite in fact.  Or that's the commonly accepted scientific consensus.  On a quantum level things are uncertain (probabilistic).  And the consensus is that this is not a measurement problem, but a fundamental nature of the quantum world.  (See: Copenhagen interpretation).  There is even a hypothesis that there can not be any hidden variables.  Here's the wiki article.

There is an argument to be made for "superdeterminism", but it is not considered probable or even testable, so it can't be theory.


--- Quote ---So for example, a man might save his 3 kids in full knowledge that he will die in the process. no he's not a 'mutant' defective person (speaking in evolutionary terms ). He does this because all his behaviour is only occurring to increase the population of his gene's, at least that particular gene that caused him to save his 3 kids. There are now a 150% (on average) increase of those gene's in the gene pool that caused him to die saving his kids, and thus, they were successful gene's.
--- End quote ---

Exactly.  Just reasoning it out that should be the outcome.  But there is empirical evidence for local populations of neolithic humans and animals limiting their reproduction to benefit everyone.  This is the selfish gene vs. altruistic group problem.  Just reasoning on the gene level altruism should always die out to selfishness.  Since the selfish individuals will out-compete the altruistic individuals.  And it's actually one of the first things evolution will do, is break altruism.  We can see that in Darwinbots with conspec breaking, as well as in real life, with certain communal slime molds which quickly "devolve" into selfishness.

But we see altruism all the time in the natural world.  We can dismiss it as kin selection and therefore really a form of selfishness by the gene.  But that doesn't explain why altruism develops in the natural world but breaks down under laboratory conditions.  Or why only some animals are altruistic and others aren't.  It in now way helps us form a scientific theory of altruism, with real and testable predictions.  It's just a form of handwavium that says "if it's an observable trait, it benefits the genes".  That would be like saying "if it's an observable result, it's because God said so".  Maybe true, but it doesn't make testable predictions.

Which is where selection on a group level comes in.  This is this aggressive mouse/altruistic mouse in the haystack model for group selection.  In a single haystack, if there are any aggressive mice, they will always win out and the entire haystack will be aggressive mice.  But there are other haystacks.  Maybe ones with just passive mice.  Passive mice will do better and there will be more of them.  Every now and then the haystacks get removed and all the mice scurry around in a mouse orgy then repopulate new haystacks.  Using game theory and a bit of math, it's possible to formulate exact conditions under which aggressive mice will become 100% of the population, and under which altruistic mice will become 100% of the population.

These conditions are dependent on aspects of how individuals interact once the haystacks are removed (whether mating is purely random, that sort of thing).  So while the behavior (and thus these conditions) are controlled by genes, the end result is not immediately clear from that.  It is emergent complexity.  The unit of selection essentially is groups in this model.  Aggressive mice can win out in every single local haystack fight and still lose globally.

See multilevel selection theory.

Houshalter:
First of all, evolution is alot more complicated then a few genes fighting each other. DNA codes for RNA which codes for protiens, which help assemble more complex structures, themselves building more complex structures, and so on untill you get to the individual, and theres no reason it should stop there because, as numsgil pointed out, evolution often acts on the level on an entire population, not just individuals (look at colony insects like bees and ants.) The concept of genes fighting against each other is rediculous because in order to have a gene that could actually do that, it would have to be rediculously complex to the point were its improbable it could have evolved from chance in the first place. There are examples like the T gene in mice, which makes 90 % of the sperm carry its gene, but its not like theres genes which litterally change the way we behave like in your dad saving his kids example. After 16 generations, you are related to 65,536 differirent organisms that lived 16 generations before you. A small population of organisms will very quickly become interrelated, and so the big differences matter more in seperate populations. In order for a single gene to survive the incredible odds of being passed down after 16 generations, it will have to have a significant effect on the organism.

Navigation

[0] Message Index

[#] Next page

[*] Previous page