Artificial Life

[Oblivion]

I've been hooked into finding programs of Artificial Life a while back. And it seems the addiction is coming back.. Anybody know of a free great artificial life program, preferably 3d?

[www2]

I've been hooked into finding programs of Artificial Life a while back. And it seems the addiction is coming back.. Anybody know of a free great artificial life program, preferably 3d?

i think that this also a idea is for vegastrike includet AI...

[Stakhanov]

Well , there are lots of neural network a-life programs and games , but true code evolution is rare.

CoreLife is one such a-life environment , I spent spent hours watching countless species of little code worms competing for the total domination of my screen. They are ruthless , resistant and clueless just like our own bacterial microcosmos.

I don't know if or how it runs on Linux (MS-DOS program) and it seems to not work on damn windows XP. I ran it with windows 98 , but afaik it only uses DOS resources - very little memory to grow on.

Life32 is not exactly about artificial life , it only grows cellular automata. Large scale patterns have very interesting properties , like filling space with stripes or other various objects , duplicating themselves , or exhibiting various mathematical properties ; but even tiny patterns can grow into a blossoming chaos.

GeneSaver is a funny screensaver displaying colored creatures gifted with neural networks to evolve and adapt. They eat each other or plants to grow and reproduce. Life-like and pretty.

[klauss]

Life32 is not exactly about artificial life , it only grows cellular automata. Large scale patterns have very interesting properties , like filling space with stripes or other various objects , duplicating themselves , or exhibiting various mathematical properties ; but even tiny patterns can grow into a blossoming chaos.

In bold: What a misleading statement. I guess you may know that there's a cellular automata which is, in essence, a universal turing machine (I mean... if you know about deterministic chaos). Then... there's no real difference about Life32 and any other a-life project - if it runs on computers, it's a turing process, and if it is, it can be modelled into a cellular automata.

[Stakhanov]

Conway's Game of Life is very different because it is deterministic. There is no real randomness - the same starting conditions always give the exact same result at generation 2047. Also , chaotic patterns all die off eventually - they can only keep growing in "explosive" environments (where each pattern is likely to fill the screen anyway)

Real randomness can only be extracted from our material world , and our computers can interface with it to provide random numbers (instead of pseudo-random algorithms) which are processed by a-life code. CoreLife cells alone are not alive (they fill the screen without evolving) , but random mutations create evolving populations. For instance , if you change the rules to allow cells of huge sizes , they may not grow alone , but generations affected by random code changes will. It becomes an artificial biotope once cells affect each other reccursively - when they fling code bits at each other.
Likewise , neural networks show true evolution when they have random changes upon duplication.

There is a debate about the nature of virtual life , some think that mere images cannot be alive. I think they can - virtual environments may not be good emulations of our own , but virtual forms that process food to grow , reproduce and evolve should be considered alive.

[klauss]

Good point, about randomness.

But mine was much more technical: if it runs on a computer, it can be modelled by cellular automata (in particular, this universal turing machine-ish one). If so... then both programs are (in an esoteric theoretical value) equivalent.

[Zeog]

Creatures is a quite good artificial life program. It simulates an entire eco-spheres with plants, animals, weather and some higher "Creatures". The more advanced inhabitants, the Norns, have a gene code which expresses body features. They kind of learn your and their language. You can breed them, insert genes from different species (Ettins and Grendels), cross-breed via the internet, examine and alter their body functions, teach them using tools or just watch them evolve. It's all quite lovely and time consuming...

[charlieg]

Try the Noble Ape Simulation. 3D, AI-centric, free (open source) - sounds like what you are looking for. Only issue is that it may be Mac OS and Windows only.

[Stakhanov]

I don't get it... what exactly are those creatures capable of ? I see them move around and interact , but I have a hard time figuring out the nature of that interaction - just competition and mating ? The manual says that they have a language , how complex is it ? Obviously such questions are not answered by the neural network display (which has no explanations either)

[charlieg]

Got a better one...

NERO: Neuro Evolving Robotic Operatives

I think that's probably what you are looking for Oblivion.

[Oblivion]

...oof

I was a biologist first before I aspired to be a programmer, so CA's don't hold that much fascination for me as actual organism sims. Though, Game Of Life did start the whole shebang about AL, I don't see much of the resemblance unless you reduce life to just its really basic principle - reproduce. Though it's good tho to generate startling patterns and fractals.

You don't know how much I've been fascinated by living things ever since I was a kid. A puddle of greenish slimy water would hold hours of fun for me, as well as anthills (incluing abortive attempts to create viable ant farms), aquariums, fish ponds, my own little mushroom garden, fossil collecting (all cenozoic stuff, boring, but I find surprises now and then, and my favorite fossils are crustaceans, crab claw and carapace molds, detached mostly ). And lots and lots of other stuff... often dragging my barkada (that's our word for cliques, or a group of freinds) along in nature treks that are rather silly.

I remember once when we reduced our school pond to nothing but lifeless muck. Like me to tell the story? Our highschool (Central Mindanao University Laboratory High School), had this little pond just below the flagpole filled with waterlilies (the pink and white kinds). Well, it was pretty so we started to get really mad ideas about improving it.

So first, we introduced some frog eggs (both from an imported Bufo and an endemic tree frog).

No change in the ecosystem.

So we added a few water hyacinths and another aquatic plant we call Water Cabbage (not sure, but I think it belongs to the duckweed family). And lo! They multiplied like mad. And pretty soon, the whole pond surface was nothing more but a floating carpet of water hyacinths with the occassional dragonflypreying on the striders, damselflies and other insects.

As soon as we realized our mistake, we decided to add some snail species (the introduce Golden Kohol, and some native snails, including ramhorns) to the pond to counter the explosive expansion of the water hyacinth.

Uhuh. Very BAD idea. The Kohols (which are quite destructive in the ricefields here), ate the water hyacinths, the remaining struggling water lilies and everything besides. The whole pond was reduced to a floating pile of brownish gunk.
And it dried up in a matter of months from the accumulated dead plant matter. Until now, I think it's still lifeless..

lol. Now that's a lesson in miniature about disrupting ecosystems..

Anyway. THAT was really off-topic. lol

@www2:
Ya. I think. I've been reading the python documentation and it's really quite easy, not much different from C++, but I still have to examine the VS py files. I'll see if I can concoct some AI in the future, but not anytime soon. . The CA programs though could be used to create terrain textures, I think.

@Stakhanov:
I've tried all of those. CoreLife doesn't run (im on XP). Life32 is CA. Genesaver looks interesting, but I prefer Darwin Pond
(http://www.ventrella.com/Darwin/darwin.html) and Gene Pool (http://www.ventrella.com/GenePool/gene_pool.html).

@klauss:

hehe. The discussion is way over my head. I don't even have the vaguest idea what a turing machine is.

@Zeog:
Yep, Before I stumbled upon VS, I was actually making some "agents" for Creatures. I succesfully made a toy and a vendor, both seems to be fun to my Norns. One thing about it tho, is that it's commercial. Out of my reach. The version I ahve is the Docking Station plus various awesome mods to increase habitation space.

Stakhanov, same problem with me. I was fascinated with them in the beginning, ad I bred a lot of Norns, but then I began to doubt if the claims to a neural net is true. I think only the genetic claims are true in Creatures. Tho I did breed some creatures who havevery obvious habits, likes and dislikes, etc. they have problems when faced with a new environment. The language they have is nothing more than syntax and learning. They do learn how to string words tho, and sometimes, it's funny, but you can't teach them new terms. They take words solely from a preset vocabulary database, disappointing.

@Charlieg:

I've tried Noble Ape. Not that attractive.

But thanksalot for NERO! Tho it's more AI-building than AL, it does promise to be interesting. I think I've come across something like this once, RoboForge I think.

ANd no, I'm still looking for a stand-alone ecosystem sim. I've come across a few, but they were all network based. Still got me drooling tho.

I've also tried Framsticks, but it's commercial. Right now, the AL program I'm playing with is Breve (http://test.spiderland.org/breve), it's promising and I'm really looking forward to releases. Coding is a must, however, and I'm still learning STEVE, their coding language. But I've made a reproducing (kinda ) agent already. It's fun but difficult and there's no neural net, they only use sine/cosine functions for evolutions. And tho it can simulate an environment, it's extremely difficult to make.

Well, still crossing my fingers. But if you guys ever come across anything AL, anything at all. Just post here. Thanks everyone.

-very long post

[klauss]

hehe. The discussion is way over my head. I don't even have the vaguest idea what a turing machine is.

Simply put, a computer.
A universal one is one that can run any possible program, without change to the machine itself.

[Stakhanov]

NERO is a great find indeed... the little guys are clever but , as the starting guide says , their intelligence is alien to us. It is not easy to fully understand the evolution of instinct ruled by arbitrary fitness parameters. It takes some patience too - as if you were teaching little kids that 1+1 is a low positive number and waiting for them to guess it.

[Oblivion]

@klauss:
I better start reading more technical books if I ever want to create artificial life. I've been going on too long with just fiction..

@Stakhanov
Indeed! I wish there were more ways to customize them tho. I managed to change their uniform but not much else. Hey, fancy a duel between our kids someday? We could post our .rtf files here and duke it out on our machines (me being without the net, ).

hehe. I've pitted some of the pretrained teams against each other, and I have to admit, it's not that spectacular, but still it's wonderful!

I can understand the hesistance of the group (from U of Texas, I got a friend studying there ) to open-source the game. It's going to be the future of RTS, methinks.

[klauss]

We have such competitions at our school every once in a while.
They're small robots, playing soccer.
Pretty fun.

Each team programs their robots, and the idea is to get the better playing algorithm. Sadly, we don't have the resources to build biped robots, and most of the time they're virtual matches played on one of the big projectors, or they're simple rolling robots, so the match goes a bit dull for the nonprogrammer.

[Stakhanov]

Something I noticed is that most robots do not make good all-around fighters and maze explorers - for instance , the sniper types are buggy and often show aberrant behavior in battle when the arena is not empty , while maze explorers are sloppy fighters , sometimes not even turning around to hit an adjacent enemy. They usually have big problems with collision.

[Oblivion]

@klauss

They're small robots, playing soccer.

kewl!!

Aren't there tournaments just like that in the US, where they do destruction derbies with em?

I remember seeing documentary about robots using artificial life algorithms once, i forgot exactly what and where. it was cool, the robots learned to walk around obstacles, find the sunlight, and etc. all without true AI programming.

@stakhanov:
yep.they focus more on one task, I think they forget older lessons quickly and i can't get them to move realistically at all. Tho I've seen a preview of a newer version not yet released, it looks more promising.. But it's boring eventually because they can't be good at everything.

Something to do with their brains maybe. Learning something that goes against previous conditioning will make them forget the earlier lessons.

e.g. snipers are trained to move away, hit, and evade. If they were schooled into another strategy, e.g. move in, hit,evade, it'd be at odds with the earlier lesson, hence, they must forget and learn again.