tomh.bbs@shark.cs.fau.edu (Tom Holroyd) (02/18/91)
If we want the ants to create some higher level organization, they must cooperate with one another, as well as compete against constraints in the environment and rival ant species. Real ants form living bridges to cross large (to them) obstacles; why not try to evolve some ants that cross simulated barriers? Bees dance to inform other bees about food sources; how about some alife bees that communicate similar info to other bees of the same family? Note that several families of bees need to compete against each other to find the best communication strategy. I've seen too many "bugs" that evolve the ability to find "food" efficiently. Let's get out of the hunter-gatherer stage and try to evolve some bugs that FARM. Even one bug that raises a crop and defends territory would count, since then the bug is cooperating with the food species, leading to symbiosis. Don't worry so much about deciding what the ants should be doing at this stage; get them to cooperate effectively and new high-level behaviors based on simple sub-behaviors should emerge almost by themselves. tomh@bambi.ccs.fau.edu Tom Holroyd Florida Atlantic University Center for Complex Systems "An anthill is a queen's way of making another queen."
marek@iuvax.cs.indiana.edu (Marek W Lugowski) (02/18/91)
In article <16mJX2w163w@shark.cs.fau.edu> tomh.bbs@shark.cs.fau.edu (Tom Holroyd) writes: >If we want the ants to create some higher level organization, >they must cooperate with one another, as well as compete against >constraints in the environment and rival ant species. .... > >Don't worry so much about deciding what the ants should be doing >at this stage; get them to cooperate effectively and new high-level >behaviors based on simple sub-behaviors should emerge almost by >themselves. > >tomh@bambi.ccs.fau.edu >Tom Holroyd >Florida Atlantic University >Center for Complex Systems >"An anthill is a queen's way of making another queen." I think those two statements are incompatible. While I agree with the former, I can't quite see how ecosystemic constraints reduce to arbitrary sub-behaviors on which something meaningful is based. I would say: Worry very much about what the ants ought to be doing (i.e., what path in chaotic space they are to be computing) and by proper design at first and natural selection later, get them to cooperate effectively. I don't buy into this "should emerge almost by themselves." That's just alife propaganda that we should be careful not to get zapped with. In real life, the ecosystem and the pressures from without as well as from within form a complex slippage of constraints, none of which is arbitrary. To compute with ants ex nihilo is artificial intelligence pure and simple: you are just defining arbitrary symbols and shoving them around, with a bit of optimization thrown in for good measure. If so, anthill-climbing is not too far from hill-climbing (search) period. What we should be doing is trying to replicate a microcosm and computing with it. This will take nontrivial amount of design and tweaking, because we will be making up for the absence of natural constraints. I share Tom's call for better ants, but only via a better antworld. -- Marek cc: alife@iuvax.cs.indiana.edu (alife-request for additions)
tomh.bbs@shark.cs.fau.edu (Tom Holroyd) (02/18/91)
marek@iuvax.cs.indiana.edu (Marek W Lugowski) writes: > I don't buy into this "should emerge almost by themselves." That's just > alife propaganda that we should be careful not to get zapped with. In OK, sorry for the somewhat un-scientific statement. In the real world, interesting behavior DID emerge by itself, but perhaps only because the environment itself was already complex. As Marek says, better ants via better antworlds. Additionally, directed evolution can be a lot faster than waiting for it to happen "by itself". Cooperation between ants at least requires that the ants be able to communicate with each other. (If all the ants are turning to the left because their programs say turn left, I don't call that cooperation.) A uniform dynamical system with positive feedback can undergo a symmetry breaking, where small random fluctuations are amplified into some non-uniform structure. What I'm hoping is that an initially uniform scattering of ants which positively reinforce each other's behavior could spontaneously generate some structure (a higher level of organization than just one ant). Selection would then operate on the higher level structure. The point is that if I want to have the ants cross a river, I don't want to program each ant with "bridge building and river crossing" behaviors; rather, I would give them the ability to communicate and cooperate with each other, and let them self- organize a behavior which, if it resulted in the ants crossing the river, would be selected for. Tom Holroyd Florida Atlantic University Center for Complex Systems tomh@bambi.ccs.fau.edu > cc: alife@iuvax.cs.indiana.edu (alife-request for additions)
xanthian@zorch.SF-Bay.ORG (Kent Paul Dolan) (02/19/91)
A couple of things still seem to be lacking to have anthills that do interesting things: 1) Much bigger anthills; it is a bit much to expect behavior networks typical of colonies of thousands of individuals to be emulated by colonies of seven individuals. 2) Extra-genetic differentiation of individuals within a colony: at least bees, and I think ants also, raise up grossly different individuals from similarly genetically endowed eggs by particular feeding schedules/choices, and the colony is based on separation of tasks to suitable individuals. Thus, an individual ant's genetic complement should represent a variety of possible mature individuals. 3) Which brings up the need to emulate nurturing behavior. 4) And also cooperative food gathering; bee and ant colonies contain some individuals who function as the colony's liver or gut; create food that can only be digested to utility for other ants by ants who can't gather it, and too big for an individual ant to drag home. 5) And allow for this very low level specialization: locomotor, digestive, reproductive, etc. with tradeoffs and improved efficiency in the area of specialization. 6) Folks who haven't, should read Douglas Hofstadter's Goedel, Escher, Bach, An Eternal Golden Braid for motivational material on ant colony functioning, especially when trying to envision things an ant colony could do. Kent, the man from xanth. <xanthian@Zorch.SF-Bay.ORG> <xanthian@well.sf.ca.us>
william@oreo.berkeley.edu (W. E. Grosso) (03/06/91)
I've just started reading this newsgroup. This discussion about ants seems interesting. Could people send me, privately, some info on what ants are ? Thanks William E. Grosso (william@math.berkeley.edu)