General > Off Topic
Specialization again
Numsgil:
1. Humans are composed of roughly 400 cell types. Each has specifically changed itself from its stem cell ancestors to become a specialized cell type. In DB, we can model this but the end MB will usually be less fit than a bunch of single cells alone.
This is really why I'd want any specialization system at all (aside from enzymes, which is a seperate but highly related thing).
2. Which specific laws of nature allow creatures to both photosynthesize and/or digest other creatures? Knowing this, at what level do we have to model it in DB? The more abstractly we model it the easier it is to handle but the less flexible it is.
The best system would be to model individual atoms, and the compounds they make. But that's obviously too complex. In the end I'm thinking of some kind of system where predefined and user defined compounds intereact in predefined ways.
Then from these rules the program figures out enzymes that can use those reactions.
3. You can't deny that in real life a 2 step deep food chain is rare. Usually you have plants (plankton usually) -> herbivores -> carnivores. Even this simple relationship won't develop naturally in the present situation, since plants are as good to bots as animals.
You seem to agree that food types (and presumably enzymes to digest them) are okay, right?
Without some form of specialization, bots have no incentive to only eat one kind of food. The best bots will have enzymes to eat everything. Which means all bots are omnivores, which means none can find any distinct niches and our sims are inevitably pretty dull.
Adding specialization of mechanics is only useful for multibots. Yet I don't want to give multibots something that isn't available to single bots, since that'd be pretty artificial.
PurpleYouko:
--- Quote ---PY points:
1. Hmm, did you see the this page? It explains how "protein/carb/fat system" may lead to appearance of different species. No additional physics necessary.
2. The mechanism for specialization is in genes. They allow different behaviour, thus - specialization.
3. In the thread linked above Nums explained how the feeding would work in such system. It is done for the bot, so there is no need to code specifics, just general decisions based on data ("crunchy" -> "spit out").
--- End quote ---
Point 1
I read every thread. Doesn't make it any more possible for that to happen though.
DBs live in a world defined by the VB code. In that code it not possible for any robot to tell the difference between carbs, fat or proteins. in each case all you see is energy and body. You shoot it and all that comes back is energy.
It makes no difference whether a robot shoots a veggie or another robot. He gets the same energy back. Not only is there no incentive to specialize but there is no ability to do so without a change in feeding efficiencies for food types.
Almost all creatures specialize in some way as to what they eat. Dogs eat meat and get sick if they eat grass. Cows eat grass and would probably have a hard time digesting a dog.
point 2
only up to a point. The genes can only do certain things within the Visual basic code. DB isn't an infinitely variable system like real life is. There is absolutely no way in the present system for any robot to become a specialized feeder. Even a robot that evolves to only fire shots at other robots with no eyes (specific veggie) can still get energy back from shots that miss the veggie and hit another bot by accident. In a specialized feeding system we could stop these shots doing any damage at all unless the feeding robot had the right enzymes to feed from the other robot type.
point 3
that is exactly my point. I want this to be sub genetic level and behind the scenes. I don't want a robot to have to fire a specific shot at a specific type of food or use a genetic instruction to metabolize sunlight. It should be automatic in the presence of the correct resources. Only bots that feed on other bots should need to shoot at all.
:D PY :D
shvarz:
OK, guys, good job in convincing me that some form of specialization is necessary. Your biology-related arguments are weak, but PY has a very good point about introducing something in VB code that makes it possible for cells to get better in a chosen task (how that would be done we can still discuss). And Nums has a good point about the level of detail we want to model. Certain arbitrary desicions must be made and accepted (after all, I accept the idea of eating by shooting particles at food :) ).
It was not as much fun as I hoped it to be, but.. well, that's life.
BTW, since so many discussions do center on biology (at least my discussions do), what about making a separate "folder" for this? People could talk about biological issues there and how these are modeled (or can be modeled) in DBs.
Botsareus:
--- Quote ---(how that would be done we can still discuss).
--- End quote ---
I think what PY and NUM are working on right now does not need any more modification until we actualy test it.
Botsareus:
--- Quote ---I think it'd be cool to be able to run some sims with a primitve bot that slowly evolves into plants and animals.
--- End quote ---
Why dont you like this idea shvartz?
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version