General > Biology

Natural Selection as an abstraction for game theory

<< < (5/6) > >>

EricL:

--- Quote from: Numsgil ---An interesting idea in the mechanics of natural selection came with Dawkin's selfish gene "framework".  I use the term framework instead of theory for a very important reason.  My first premise is that saying natural selection works on individuals trying to pass their genes to the next generation, and saying that genes are acting to propogate themselves are equivelant statements.  They're simply different ways of viewing the same phenomenon, and can be used interchangably depending on which one makes the most sense for the given question at hand.
--- End quote ---
I disagree that they are equivalent statements.  Where the decision/motivation to perform an altrusitic act resides is the crux.  With genes as the principle actors, altruistic behaviour by the organism such as kin selection is explained by pre-programmed tendencies in the organism (including meachanisms that go with it such as the ability to guage relatedness) I.e. people perform altrusitic acts in cetain cases because their genes have found it a succesful strategy to do so.  Saying the individuals are the actors implies a choice made by the organism independent of the genetic programming.


--- Quote from: Numsgil ---Even black holes slowly evaporate into nothing through electron tunneling.
--- End quote ---
I beleive Hawking's theory has been disproved for all but quantum black holes.  Black holes above a certain mass threshold will never evaporate.


--- Quote from: Numsgil ---Cannibalism and the leagues - There is something of a pardox to the following fact: conspec recognition in leagues is important for success.  Bots that eat their children or other members of the species will not do as well against another species that doesn't.  However, in evolutionary simulations, this conspec recognition always, except in extremely contrived situations, is selected against and disappears.
--- End quote ---
As I have said before, there is little reason for asexually reproducing organisms not to be cannabilistic.  There is no gene flow between members of an asexually reproducing 'species' through sex and recombination, only through descent.  Thus, there is no gene-based evolutionary advantage to evolving species recognition or relatedness mechanisms (much harder than it sounds) and practicing kin altuism.  But what we are really seeing in DB evo sims IMHO is even simiplier than that.  In every evo sim I've seen where selection is operating (not the first days of a zerobot sim) bots in proximity are all descended from a recent common ancestor.  Thus, ALL bots in the sim are a close relatives through descent.  There is no way be a carnivore and NOT to be cannabilistic.  We don't have the diversity of niches (yet) to support speciation in proximity.  We really can't even talk about multi-species issues yet becuase we have yet to evolve multiple species.


--- Quote from: Numsgil ---When a bot species is fighting another bot species, the entire battle usually is measured in the low thousands of cycles.  "Cannibots" usually appear after some time of mutations has occurred, somewhere on the order of hundreds to low millions of cycles.
--- End quote ---

This is not an evolutionary phenomem.  This IMHO is simply a degrading of hand-coded programming when a human authored machine is subjected to mutations.  Multiple hand-coded organisms in a sim is not a natural or evolved state and it is incorrect to call them different species.  Humans wrote the code to avoid attacking other bots based upon some artifical identification mechanism.  Humans invented this notion of species.  It has nothing to do with evolved darwinian speciation or kin selection.  None of the code, behaviour or identification is evolved.   What we think of as species in DB is a flawed concept or at least not a darwiniam concept and species related conclusions based upon it is IMHO flawed.


--- Quote from: Numsgil ---Battling against other species occurs in a slightly different time frame.  In this time frame, the two species are battling for a single niche (victor of the round).  All bots of a single species are essentially the same agent, and it would not be wise for an agent in a zero sum game to fight against itself, which is why cannibalism doesn't work in the leagues.
--- End quote ---

Cannibalism doesn't work in leauges because humans programming their agent machines to destroy one another is simply not a winning strategy.  It has nothing to do with evolution, species or evolved cannabalism.


--- Quote from: Numsgil ---However, eventually a cannibot will develop.  This cannibot has a huge advantage in the game of surviving because it actively destroys other agents.  It no longer needs to get lucky and find a new veggy patch or have one spawn, because it's eliminating any other potential rivals for food.  And it's easy pickings because they don't fight back.  What I'm saying is that cannibots develop not because it's a new food source (although that's an important plus), but because it's a trait for eliminating rivals.
--- End quote ---
I would claim this is just the human authored machine deteriating further, falling down off a naturally unobtainable peak in the fitness landscape and that making conclusions based upon simulations which start with anything more complex than the simplest replicator (and certainly conclusions which rely upon some human defned concept of species) are flawed.  As above, IMHO the whole notion of what we call species in DB is artifical, arbitray and very unrelated to actual speciation in DB, if and when we see it evolve (to my knowledge, we have not seen it evolve when bots are in proximity).  We should not act surprised when a human coded gene breaks down and starts doing something different than it's author intended.  Your overall theory may be correct, but IMHO, using examples based upon human authored bots and artifical, human defintions of species and drawing big conclusions about cannabalism or something from that is not supportted.


--- Quote from: Numsgil ---Big berthas - Big berthas are very large, very sterile bots that constantly shoot.  They used to develop in evo sims before I increased the problems associated with waste.  Basically, a bot would break its gene to reproduce, and would instead constantly acquire nrg and body.  The body would make it stronger and basically unkillable by the smaller, reproducing bots.  5 or 6 Big berthas would eventually kill all strains of reproducing bots, causing a mass extinction event when the big berthas eventually died.  What's interesting is that these 5 or 6 big berthas would all be independant mutations since the big berthas never reproduced.
--- End quote ---

Big Berthas are in my opinion, simply an aberation that occurs in sims that do not have the costs/environment/diversity to select against them.  In a rich enough, large enough, diverse enough sim, some bot would eventually evolve to take advantage of the nrg source they represent.  

You theories as related to gaming, etc. may be sound, but I think you are on shaky ground drawing conclusions from DB which are based upon hand-coded bots and human-defined notions of species.

Numsgil:

--- Quote from: EricL ---I disagree that they are equivalent statements.  Where the decision/motivation to perform an altrusitic act resides is the crux.  With genes as the principle actors, altruistic behaviour by the organism such as kin selection is explained by pre-programmed tendencies in the organism (including meachanisms that go with it such as the ability to guage relatedness) I.e. people perform altrusitic acts in cetain cases because their genes have found it a succesful strategy to do so.  Saying the individuals are the actors implies a choice made by the organism independent of the genetic programming.
--- End quote ---

If you are examining things with individuals being the agents, then you need to use the concept of intrinsic fitness.  By redefining the concept of what an "individual" agent is for natural selection to operate on, you're changing the rules.  This is key: they are equivalent statements (isomorphic), but the mapping changes the basic rules of selection.  What I'm proposing is that you can do this all the way up, using intrinsic fitness at each level to make the concept of an "agent" more inclusive.  But doing this changes the rules of the game for the agents to play.  Determining these rules is outside the scope of my framework.  What it does say is that faster paced games have more weight for selection.

It's a sort of recursive subdivision.  Hopefully this makes sense to you?


--- Quote ---I beleive Hawking's theory has been disproved for all but quantum black holes.  Black holes above a certain mass threshold will never evaporate.
--- End quote ---

Okay, I'm going to call you on this!  I haven't heard anything of the kind.  Where did you read this?  What was Hawkins reply?  Is it a new, still contested point that disproves the idea of black hole evaporatioin, or is it an overlooked flaw in Hawkins proof?


--- Quote ---As I have said before, there is little reason for asexually reproducing organisms not to be cannabilistic.  There is no gene flow between members of an asexually reproducing 'species' through sex and recombination, only through descent.  Thus, there is no gene-based evolutionary advantage to evolving species recognition or relatedness mechanisms (much harder than it sounds) and practicing kin altuism.
--- End quote ---

I am aware of nothing that prevents asexual creatures from practicing kin selection.  Actually, since close relatives are more likely to be 100% genetically identical, kin selection would seem to be even stronger.  Why doesn't kin selection develop?  Under what conditions does it "devolve", and under what conditions does it persist?  I also find it extremely naive to simply say "there is no gene-based evolutionary advantage".  It's begging the question.  It's about as useful as saying "because God wants it to".  It's almost a truism, and it doesn't help us understand why.


--- Quote ---This is not an evolutionary phenomem.  This IMHO is simply a degrading of hand-coded programming when a human authored machine is subjected to mutations... It has nothing to do with evolved darwinian speciation or kin selection. None of the code, behaviour or identification is evolved. What we think of as species in DB is a flawed concept or at least not a darwiniam concept and species related conclusions based upon it is IMHO flawed.

 Multiple hand-coded organisms in a sim is not a natural or evolved state and it is incorrect to call them different species.  Humans wrote the code to avoid attacking other bots based upon some artifical identification mechanism.  Humans invented this notion of species.

...

I would claim this is just the human authored machine deteriating further, falling down off a naturally unobtainable peak in the fitness landscape and that making conclusions based upon simulations which start with anything more complex than the simplest replicator (and certainly conclusions which rely upon some human defned concept of species) are flawed.
--- End quote ---

Here is where I will vehemetly (sp?) draw the line.  The hand coded bots are acted upon by mutations.  They must be subject to the law of natural selection same as every other mutating and reproducing entity in the whole of the universe.  Shot bots don't lose the ability to shoot.  That's a hand coded trait that's evolutionarily stable.  Reproduction and movement are iffier, and are related to the size of the simulation, and how much it is well mixed.  Conspec recognition is the first thing to go in an evo sim.  Clearly there is strong selecting pressure against it.

Remember, mutations act on the genotype, but natural selection acts on the phenotype.  Your argument assumes that because a genome is highly artificial, and prone to any mutations breaking it, that must mean that the phenotype is likewise easily broken.  If this were true, we'd see all the behaviors of bots (vision, movement, reproduction, shooting, and conspec recognition) fail at frequencies related to their genotypical complexity.  We don't.  Our bots are subject to natural selection.  Always.


--- Quote ---Cannibalism doesn't work in leauges because humans programming their agent machines to destroy one another is simply not a winning strategy.  It has nothing to do with evolution, species or evolved cannabalism.
--- End quote ---

That's my point!  It's not a winning strategy for a species.  Which can be seen as a single agent, because all the bots are somewhat genetically similar compared to another species.  It's immaterial how we define the species, because at the start of the league, all bots that share a "species" label are 100% identical, and 100% unidentical to the other species.  At the end of a round, the game is "won", there is only a single agent.  This causes this levels game to be given a weight of zero, and the next weighted game is the competition between individuals.

Strategies that are successful in one game can be harmful or unstable in another.  What matters is the weight of the respective games.  Faster paced games (league battles) are more strongly weighted.  Again, we can test this hypothesis by figuring out a way to extend most leagues battles into the millions of cycles (larger arena maybe, we'd have to test it) and allowing mutations into the round.  My hypothesis says that species who develop cannibots will lose, or actively fight the cannibot.  If it loses, it won't be represented in the future of the sim, because it will not exist.  We will see the longterm incidence of cannibalism be very low in winning species.


--- Quote ---IMHO, using examples based upon human authored bots and artifical, human defintions of species and drawing big conclusions about cannabalism or something from that is not supportted.
--- End quote ---

Using artificial bots is important because it allows us the specificity and variable limiting that is important in good science.


--- Quote ---Big Berthas are in my opinion, simply an aberation that occurs in sims that do not have the costs/environment/diversity to select against them.  In a rich enough, large enough, diverse enough sim, some bot would eventually evolve to take advantage of the nrg source they represent.
--- End quote ---

This is very poor science.  You cannot place all the data that you don't like in the neat little category of "outlier", especially if it's data that's very common.  And not just in Darwinbots, there are all sorts of circumstances where evolution has proceeded down a course that ultimately leads to the population's extinction.

My framework allows us to examine the reasons why.  And it's not all that novel.  All it's saying is that evolution is very short sighted, and will react to local pressures more strongly and more immediately than it will react to global pressures.  Evolution will lead to its own destruction at certain times.  It doesn't need a mutation meltdown, it doesn't need mueller's ratchet.

Really, this shouldn't be a hard pill to swallow.


--- Quote ---You theories as related to gaming, etc. may be sound, but I think you are on shaky ground drawing conclusions from DB which are based upon hand-coded bots and human-defined notions of species.
--- End quote ---

The idea works eaqually well with real observed phenomenon.  For instance Slime mold altruism.  Developing slime molds that break their altruism gene is very easy in the laboratory, even common.  My framework allows us to see selection in nature as a competition between rival slime mold colonies.  In this competition, cheaters cause their host slime mold colonies to be less successful, and they'll be eliminated by other successful colonies.  In a lab, this higher, faster paced selection is not available, and the next slowest game, competition between individuals, is the dominant game.

Note that my framework does not allow us to see how.  It's outside the domain of my framework to say how slime molds react to cheaters, or how cheaters influence the success of their colony beyond saying that if it will quickly cause their colony to no longer exist in the genotype of the future generations, it will not exist in future generations.

------------------------------------------------------

Any real criticism of my framework should really be of the counter example variety.  Everything my framework is based on should be either self evident or at the very least a valid current scientific theory (though perhaps hotly debated.  I won't spend the time to defend intrinsic fitness).

Phrased another way, my framework basically says is that evolution is short sighted, and short term benefit overrides long term detriment.

EricL:

--- Quote ---Okay, I'm going to call you on this!  I haven't heard anything of the kind.  Where did you read this?  What was Hawkins reply?  Is it a new, still contested point that disproves the idea of black hole evaporatioin, or is it an overlooked flaw in Hawkins proof?
--- End quote ---

Will reply at length to main topic later.  On this aside, I must apologize.  I mispoke.  The theory has not been disproved.  Rather, the calculations show that the mass loss due to Hawking radiation from any black hole above a certain relativly small size will be exceeded by the mass intake due to cosmic background radiation.   If I remember correctly, this size is much much smaller than the minimum size possible for a black hole formed via stellar collapse and thus only theoritical, so-called quantum black holes will evaporate (and even these, if such exist, will take many times longer than the age of the universe to do so).  This is pretty old news and came to light shortly after Hawkings famous 1974 paper was published.

Numsgil:
Interesting.

My understanding is that if the present acceleration of the expansion of space continues unabbated, eventually it would reach a point that even the strongest gravitational forces would be overcome by the sheer expansion of space, eventually destroying all matter, including blackholes.

Interesting stuff, even if it's sort of off topic

Numsgil:
A short postscript.  If you take the entire phylogenic tree and bind it together as a single agent, you would basically have the same things as Gaia theory.  I think this fact is very helpful in illuminating where I'm comming from.  It's very much in the same boat as Gaia theory.

Navigation

[0] Message Index

[#] Next page

[*] Previous page