Testing a hypothesis

[Numsgil]

Came to me after watching the robots Jez linked to.  I just want to write it down before I forget.

Basically, you'd have a controller veg that gets fed basically infinite nrg.  It would create insanely large amounts of both slime and shell, to prevent it from being a viable food source.  Poison too.  It would be programmed to send info shots to bots that it sees that have certain criteria, forcing them to reproduce.  Maybe this threshold value changes over time as the number of bots in the sim increases/decreases.

Another dummy veg species would exist as little more than food pellets.

Another controller veg species would act as the dynamic predator that Eric has mentioned.  It would get stronger or weaker depending on how many animal bots it sees.

The animal bots would have rudimentary behavior to feed from dummy vegs and move around a bit if there isn't any food.  It would seek out the controller veg when it feels that it's time to reproduce.

The hypothesis I have is that cooperation becomes easy and natural when you're reproduction is handled by another, independent agent.  I think this sort of environment should foster cooperation as opposed to cannibalism.

The hard part is deciding the criteria that the master controller veg uses to decide wether or not to reproduce a bot.  I'm thinking possible candidates are either age, nrg and/or body, or kills.  Another option is maybe having the bots give nrg to the master controller as a sort of "bribe".

I'm betting that the bots won't learn to independantly reproduce very quickly.  Though I might have to add some new features to ensure that only the master controller allows reproduction.  I'll probably need to add a few features here and there anyway to get things like my predator up and running.

[Endy]

The bots should need to have a certain minimum body/nrg level as the minimum requirement. You'd also need to make sure the two main vegs, predator and reproducer are seperated though. Could be problems if the repro shot goes wild.

Not sure if repop would be an issue, with the different veg species, although rapidly reproducing vegs could possibly solve this problem.

[Numsgil]

Looking at it after a day or two, I think the criteria for reproduction should be a minimum stored nrg and age.   You want to encourage collection of nrg, without encouraging overt "cheating".

[Endy]

Ran a basic test and found that it does appear to at least limit canni's to an extent; they still crop up but since reproduction is less easy they don't last. Might want to add a timer to your controller, the shots come out so rapidly, that a bot will typically be forced to reproduce multiple times until the area surrounding it is filled with offspring(4-6).

[Numsgil]

Hmm, yes.  Or perhaps just have the controller suddenly turn all the way around after it shoots, so it can't see what it just shot, and shoot it again on accident.

[Endy]

Running a related test with bots that control each other's reproduction via info shots. Interesting effects, the population deffinitly appears more stable than normal. Some bots still figured out a way to game the system by not spurring other's reproduction, others have taken to limited canni feeding. Still sems relativly controlled compared to normal, primarily since the bots are reliant upon others to reproduce.

[Numsgil]

I'm considering adding a feature to the DNA that lets you "mask" certain sysvars from being changed by the DNA.  That would mean you could run a long term sim like this where the only way to reproduce is to use another bot, without the potential for reversion to autoreproduction.