Bots and Simulations > Evolution and Internet Sharing Sims

How to avoid cannibotism?

(1/3) > >>

Ispettore:
Is there any way for you to avoid cannibotism in long-term evo sims? I mean, there should be a way (settings, different amount of predators, I don't know) to avoid the fact that cannibots are advantaged, because conspecs can't answer their fire. Is there for you to make evolution improve collaboration (or at least not killing) between conspecs?

EricL:
We should seperate out two types of cannabalism:

1) A mutation impacts species recognition such that I attack anything and everything including my own descendants.
2) A mutation impacts species recognition such that I attack anything not possessing the same species recognition DNA I have.

In therory, type 1 should not provide an advantage in the long run.  An individual gains near term advantage for a generation or two due to the abundance of easy prey, but its does not favor longer term reproductive success for the cannabalistic DNA since it attacks it's offspring or others related to it (via direct descent).  Type 2 should provide a long term advantage for the new DNA in asexually reproducing species and one might expect a new non-cannabalistic equalibrium to establish itself once all the others are wiped out.

But this whole argument is based on the presuption that the attacks by cannabots are not repulsed, that somehow the non-canablaistic members still think of the cannibal as one of their species despite it's rouge behaviour.   Real species recognition is not black or white.  It's complex, based on many clues including behaviour and members of the same species will often turn on and kill a rouge in their midst if it displays anti-social behaivour even when the rouge shares the same DNA.  Genes protect themselves.  Species form as a side effect of gene self preservation.

Unfortunatly to date our bots, evolved or not, are too simplisitc in their species recogntion to base it on behaviour.  In this light, cannabalism can be seen as an advantage selection will and should favor.  I.e. DNA so dumb as to not defend itself when attacked (even by one nearly identical to itself) is selected against.  Type 2 cannabalism at least defends itself via a good offsense against everything.   So, attacking everything can be seen as a first primitive step towards evolving true species recognition since the first step has to be for the DNA to defend itself and artifical (I.e. hand authorred) species recognition does not do that against cannabots.   Thus it is little wonder that cannibots are selected for in our simplisitc sims.  Nature doesn;t do thinng the way we do.  This is why I claim that evolving from starter bots with hand coded DNA is a dead end.   Hand coded bots are unreachable peaks in the fitness landscape.  Easy to fall from.  Necessary to fall from so that a real trail up the side of mount improbable can be climbed by evolution.

Peter:

--- Quote ---1) A mutation impacts species recognition such that I attack anything and everything including my own descendants.
2) A mutation impacts species recognition such that I attack anything not possessing the same species recognition DNA I have.
--- End quote ---
I would like to say that one certainly is canibalism. Two is like I see it a declaration of independence of one bot from the rest, aka I am the first of a whole new specie. As it wouldn't attack it's own young bot only the rest.

As by some handbuilded bots there are some defences against such cannibalism.
Such as shooting a virus at the cani bot to chance his id.
A small timeout after the birth. In the timeout there is no shooting.
There are different possibilities, different combinations.

In my bot I have set up different ID-systems, just with a small mutation in one id the bot will still be seen as desame specie. At the other end it is pretty strict on putting venom, poison or just shooting on everyone that attacks it, in couterattack genes there has no conspec-rec been set. Means if you punch it it will puch back without even looking at you.

I have to say, my way doesn't work perfect, in IM I have received bots back of multiply that just attacked everything. I've got to say that the dna has extremely changed. Nothing was desame.

You can look around on the forum and find some ideas for countering cannis, if you find a good working one (or anybody) I am pleased to know.


Could be interesting to see a bot build up a own id-system in any sim. Could be something interesting, instead of begining with normal bots, begin with simple canni-bots and see if there could evolve a id system. I doubt if that will happen. But it could be nice.

Numsgil:
The best way to counteract cannibots is to make your bots defend themselves.  However, this can quickly lead to anarchy if you use shots, since shots might stray from the intended target.

Another method is to design cannibalism into the bots from the start.  I have a bot I like to run that tries to kill everything, including its own young.  Young learn to escape from parents and bulk up on veggies until they're strong enough to survive.

Elite:
I had a huge post all typed out, but the forum ate it. Damn.

Yeah, first way is to design smarter conspec-recognition systems, recipriocity, offspring-checking, defence, poison canni-'marking' etc. into the bots.

The problem is that at an individual-selevtion, level, cannibiotism really is better than playing nice. Cannis can feed with impunity on their meek, unconditionally-cooperating kin, and this represents a huge individual advantage for a bot.

From an evolutionary standpoint, there are two solutions to this:

(1) Raising the priority of gene-level selection. Cannis, the selfish individualists they are, are poor at passing on their genes due to their disdvantageous habit of eating their offspring. By (for example) using a massive ramping age cost that prevents bots from living indefinitely, there is a selection pressure to pass your genes forward and safeguard your offspring rather than eating them. This mechanism has been proposed as a way that cooperation could have evolved in real life (advocated by Richard Dawkins), and is especially relevant when it comes to eusocial reproductive systems and kin-selection etc.

(2) Group-level selection. This mechanism has also been proposed to explain altruism and cooperation in the real world (Gould, I think (?) was a fan of this view). If you've got a situation where individuals must cooperate against an external threat, cooperation is favoured or the whole group goes bust. Pack-bots, territory-forming bots, antbots etc. where there are a plurality of groups in competition for limited resources favours the evolution of cooperation (and complex cooperation strategies). There has got to be a situation where bots' survival depends on others in its group. A load of individuals with a set of conspec-recognition codes wouldn't cut it.

Navigation

[0] Message Index

[#] Next page