F1 Alliance

[googlyeyesultra]

I was thinking. . . What if several of us bot programmers got together and create a whole slew of "allied" bots (with the same conspec) and some sort of teamwork strategies. Presumably, a few programmers making allied bots could quickly take over the internet mode. We could have veggies that would give our bots a small portion of their energy in exchange for protection, and whole armies of symbiotic combatants.

Anyways, does anyone have any interest in trying something like this? It might be interesting.

[Numsgil]

It might be interesting, but there's a huge incentive to cheat and feed on bots that think you're a conspec.

[googlyeyesultra]

Although that's perfectly possible now. If I really wanted to, I could store the right values in the right spots to defend against all existing memloc based conspecs, make a bot with the exact right number of eyes, and a bot that stores the correct value in .out1. However, that's really against the spirit of the game (people actually bothered to write a memloc/memval copying gene instead of just telling the bot to store the same value that Ymir does, for instance). And, if someone does, we can always change our conspec and blow 'em out of the water.

We could have both hunters, and perhaps veggies that would lend energy in exchange for protection.

Anyways, I just think this might be a nice idea.

[Peter]

Well, it sounds nice.

Problem is that if they mutate conpec. rec. could break and it would turn into chaos.

We could have both hunters, and perhaps veggies that would lend energy in exchange for protection.

And for this, if I understand you want to make the different bots that are made by different bot programmers live with eachother and behave well and protect eachother, and have cooperation with eachother and living a socity on eachothers strong points.

Sounds nice, but making this with different bot-programmers sounds hard.Fore example, I am making simple bots consisting out of genes like the original cond-start-stop system. I can barely read 1G-bots.    So I don't even know for sure if I can read yours. And if the bots have to work together it could be nice to be able to see what is written.

I sounds easier to make a system like this with just one programmer, who knows what heself is doing. Could be nice to see a real working together between bots and vegs for example.

[googlyeyesultra]

You've got the basic idea of it, Peter. Multiple bots by multiple programmers designed to either cooperate or at least consider each other conspecs. And if someone creates some inventive bot code, we can share it amongst all of the bots in the system, improving the overall fitness of the alliance.

I think as long as we use the same conspec system (perhaps a cycling variation of memloc/memval for ease of use + security?) it really doesn't matter how well we understand each others bots. That aside, should you decide to join this (possible) alliance, I'll be more than willing to explain any snippet of code of mine that doesn't make sense. I don't do SG (sqr div mult sub wha?) stuff, but I am using the new conditional paradym (I actually find it a bit easier to use, but that's just me.).

For the sake of conspec, though, we'd need to disable mutations for all of our bots (you could still do something along the lines of epigenetic evo, like lionfish).

My reasoning is that different programmers have different ideas and different fortes. Presumably if we combine all of these into some huge conglamorate army of bots, at least one bot from the alliance should do well in any given sim.

Anyways, this is still really just an idea. Anyone want to consider making an internet mode alliance? If I get some serious interest, I'll work on making a new conspec system based off of memloc/memval.

[Peter]

I think as long as we use the same conspec system (perhaps a cycling variation of memloc/memval for ease of use + security?) it really doesn't matter how well we understand each others bots. That aside, should you decide to join this (possible) alliance, I'll be more than willing to explain any snippet of code of mine that doesn't make sense. I don't do SG (sqr div mult sub wha?) stuff, but I am using the new conditional paradym (I actually find it a bit easier to use, but that's just me.).

That new conditional is easy to use. It works really fine, yes. Even I do understand it.  

Anyways, this is still really just an idea. Anyone want to consider making an internet mode alliance? If I get some serious interest, I'll work on making a new conspec system based off of memloc/memval.

Well, I think I am not really interested, sorry. Basicly the idea is fine but the fact that the bots may not mutate strikes me a little.
Maybe I am even going to trow a bot in that doesn't even know what a conspec is. And just attacks everybody or almost nobody.  

[fulizer]

good idea im in!
but one question.
what conspifec gene will we use?
im programming a sweeperbot that will send out some other bots to find food then report bact to him radar style!

[Commander Keen]

This could be done in a league, but since it's one bot type vs one bot type, you'd want a bot that contained the genes for all variations and randomly picked a variation, possibly deleting the genes for the other uneeded variation to save energy.

[fulizer]

how about a cancourus bot that will mass reproduce. trying to make sure that one bot will get each type(or does that not work on leuges)

[rsucoop]

This sounds like a very great idea. I think if you used some code like:

42 rnd 110 store

This gene would create 42 possible bots for one DNA strand. The Conds would hanve to use the 110 value check a lot though, adding the extra cost to the gene.

Before we decide the conspec recognition gene, we need to establish a maximum remaining gene number, so that it could be matched within 10% +/- of eachother and be ok. It would also be very smart to come up with a Multi anti-virus that would be active in say 10% of the bots. Organization of bot movement should be kept to tmemloc and tmemval.

The best way to update a lot of ties is this

cond
*.numties *150 !=
*.multi 0 >
start
*150 .tienum store
*.tienum .readtie store
1 *150 add 150 store
stop

'place feed instructions and all tmem related genes here

cond
*150 *.numties =
*.multi 0 >
start
1 150 store
stop

This allows for the tienum to update every cycle, so that a new bot's mem is read for changes and reacted upon. A lot of memory will have to be used for reference of each bot.

[bacillus]

Is it possible to change more than one memory location at a time?
Here's a basic model for a hunter/veggie/producer symbiosis:

def isVeg = 10
def hunter = 20
def supplier = 30

'I/o channel 1 contains whether this bot is a veggie, hunter or produces resources

cond
start
 *.in1 .tie store
stop

cond
 *.out1 .hunter =
start
 -6 .shoot store
 50 .sharenrg store
stop

cond
 *.out1 .veggie =
start
 20 .sharenrg store
stop

cond
 *.out1 .supplier =
start
 50 .mkslime store
 30 .shareslime store
 10 .mkshell store
 50 .shareshell store
 10 .mkpoison store
 50 .sharepoison store
stop

end


It's a rough idea, but it may be a start.

[abyaly]

What is the advantage of different bots which have advantages in different situations over a single bot that can emulate each of those behaviors in each of those situations?

[Numsgil]

Aesthetics at the moment.  There's nothing in the engine to encourage this, except for maybe DNA length costs, if they're set up.

[bacillus]

At the moment I'm thinking modularity. In this way, we can first write all our bots, and adapt them to each other, then when all the bots are complete, we can throw them all together or whatever else you had in mind.

[Moonfisher]

You would also be able to make complex alge that support allied bots and fight or flee from enemies, this way you don't need to channel a huge virus first.
Also it would be an advantage for bots that mutate, shorter simpler bots would be less fragile.