Floys - Territorial artificial life flocking java creatures

[Shadow]

Hey all,

 

I've recently started to delve into genetic algorithms, and things associated with them, including alife. In my searches of the internet, I stumbled upon Floys, a simulation of creatures that share some characteristics with living creatures, among others "the social tendency to stick together" and "being territorial animals that defend their territory against intruders". There are two other, more advanced versions, other than the basic one I link to, called iFloys (Individualistic Floys, having individual personality) and eFloys (Evolving Floys, evolve by genetic algorithms) . You can find links to those two on the site.

 

I find this a very interesting idea, and though I would share. It also gave me an idea of my own that I will attempt to implement, and subsequently post on this forum. Has anyone ever tried programming anything similar?

 

Cheers,

 

Gabe

[itay390]

Yes, i made http://www.youtube.com/watch?v=jTfSlg7SRKk

Evolute was a personal, and a college project of mine, i don't believe it yet reached it's maximum potential

i believe what i made can be at least 10x better.

the thing i made is general, not restricted into animal simulation.

but an artificial world with animals was one of the demonstrations i coded.

other "demonstrations" i coded were guessing a function of the shape that the user draw on the screen,

and picture compression (to a function).

i think genetic algorithms can be promising, but at the same time they require enourmous computing resources.

i don't know really what the maximum potential of my library (Evolute) , so i can't tell you exactly what to expect. but don't expect for something too complex to evolve. because that's what i expected , and i was disappointed.. but maybe future devlopment will suprise me.

well .. i can discuss that subject for hours. i don't know what else you want to know.

Edited October 12, 2009 by itay390

[Shadow]

Interesting. That's sort of what I was thinking off, but not entirely. I'll post the program when (if) it's finished. Thanks for your input.

 

Anyone else tried anything similar?

Edited October 12, 2009 by Shadow

[Mr Skeptic]

There probably are others out there. But if you do make one, poke me and I will have a look at it. I think genetic algorithms have some serious potential, but we might need more computing power.

[itay390]

The only restraint we have is computing power.

its just a bad fact , that in order to evolve something you have to evaulate every individual/code/behaviour in the population to know how good/bad it is. this means , for an example, simulating the whole world for every individual (on in other way : simulating a world with the number of animals so that there will be at least 1 animal for every different behaviour)

and after that, you will have to simulate the world for quite some time in order to see how the creatures act and what animal is better than the other, and get a result for every animal (fitness/ how good/bad the animal is) for the next generation.

another burden is the simulation overhead. you see, when you are executing a behaviour that was randomly produced, the more easy thing would be to interpret the basic behaviour commands in a "while" loop (or recursion loop, if wer'e talking trees here). but that's gonna cost you a lot, because if you make genetic algorithm that runs on machine code, you won't have that overhead , and in result, your code can run at least 20 times faster than interpretation.

programming in machine code is a hard thing. unless your'e some crazy guy from the "demoscene", which knows to hack you a 3D presentation in one hour, who knows only assembly.

 

the human mind, yes - the simple daily human mind you and I carry inside our head each day, is more powerful in the processing power (calculations per second), than the best super computer we have today. at least 10 times more powerful than the supercomputer.

Edited October 13, 2009 by itay390

[iNow]

the human mind, yes - the simple daily human mind you and I carry inside our head each day, is more powerful in the processing power (calculations per second), than the best super computer we have today. at least 10 times more powerful than the supercomputer.

 

Actually, that's not true at all, itay.

[bascule]

Actually, that's not true at all, itay.

 

It's very hard to gauge, but I wouldn't go saying that statement is wrong, per se.

[Mr Skeptic]

the human mind, yes - the simple daily human mind you and I carry inside our head each day, is more powerful in the processing power (calculations per second), than the best super computer we have today. at least 10 times more powerful than the supercomputer.

 

For now, perhaps. I think it depends on how you measure its processing power; neurons don't correspond nicely to computers. For example, the amount of computing power to build a neural net like a human's would be immense. The Blue Brain project, for example, uses a supercomputer to simulate a single neocortical column of a rat. On the other hand, computing power is often measured in floating point operations per second. On that metric, a calculator would kick our sorry butts.

[itay390]

...

indeed..

 

yes we can't do as much "floating operations" that a computer can do. the faster (normal) people can do it in a minute (to me it takes a few). people like autistic savants can do it in less than a second. in our brain when we do this kind of action its very high thought process, it probably involves much circuitry in the brain, instead of doing it very primitively like a computer does. this is because we don't have a dedicated circuitry for this type of operation, evolution never needed us to have such a skill. instead it developed dedicated circuitry for logic/reason, eye movements, coordination, face recognition, social intelligence - yes we have that information from birth, and other things..

 

i wonder if it will ever be possible to program the neurons to have a predetermined circuit. it will then probably be possible to store any kind of normal computer program in a circuit of neurons. that would probably be very very akward because of how neural network works. as far as i know neurons are pretty accurate. almost like a clockwork, they can generate a pattern of repeating currents for hours without one shift from it.

maybe if you take all the neurons in the brain and divide them into groups where you could construct from each group a floating number calculator, then you will be able to compare to a calculator (except that a computer is much more than a calculator)