I find that all of this is interesting. I believe that I would show the way I see things, and probably the way some other poeple see how this would work.

First, everything will evolve. We are normally happy for others, or try to help others since sub-conciously we know that it will help us. (Evolution is survival of the fittest)

The structure may look like this:
struct creature
{
int commands[20][10];
int vars[20][1]
int x,y,z;
int d;
}

The x,y,z are obviously the location, d the direction.
Notice that the commands are in a 2D matrix. This gives the advantage of line numbers.

Therefore, we can have like in basica (gwbasic):
IF var1>60 THEN 3

And assignment operators, etc.

Also, the vars matrix is useful since it allows the computers program to remember information.

I''m emphasizing on giving it the ability to store vars. and jump around in its code since it would give the chance to evolve with even more abilities.

Probably this was already self-evident (I did my best to read all the posts).

~Mike2000

Hey! I use this same idea for the learning mechanism in my program. It''s darwinism in 0''s and 1''s!!! What the critter in my programs goal is to eat more food then anyone and reproduce more of itself then anyone.
There are simple properties of each critter that mutate over time, and get better. But the intelligence aspect is a little diffrent. Each ameoba does something when they are roaming around. They are given an infinitely long array that is changed on the fly. They "learn" from other ameobas. They have this array and they fill it with actions that are followed through. The better the array the less change that is made to it. It''s a simple mechanism but it makes for REALLY REALLY REALLY smart critters after a few generations. The array is filled by the critter seeing that another ameoba is doing something and then it plugs its action into its own array of action.

The idea behind this is DNA. DNA are made up of neucleotides, 3 amino base acids. There are 4 base acids. These base acids can be put in hundreds of diffrent kinds of orders to create billions of different creatures!

So by creating 4 basic actions, and putting them in any order I can create an open ended intelligence. Don''t beleive me? Here download the program for yourself!
www.geocities.com/wavewash

It really works. I am almost done with an update for it that I''ll be sure to put up when I am done.
~Wave

quote:

---

The_Minister
I think the way to do this is to create atoms, and design a level editor that let's you paste em together. This and the set of physics to go with it, yep all of em.
Now it could take years to build something incredibly simple, but once it has been done, we are one step closer to simulating this universe.

---

Hey, i agree with minister. We can simulate everything in the universe, in the near future, where the computers will have a very huge speed and memory, we'll able to simulate our universe, and more, if we can build a computer fast enough to simulate millions of years in a day, we can "spy" the intelligent creatures that will "born" in the simulation and "copy" their inventions and technologies

Edited by - AnarquistA on November 5, 2000 12:42:54 PM

What I don''t get is, going back to the Matrix idea, do you simultaneously run several matrix monsters, or do you run one, then run his offspring, then run his? If a create is *less* successful than his parent, do you then go back to the parent matrix, or do you continue on the existing one?

Also, I think each number should have a totally different meaning depending on the context, so that a given matrix, e.g. (123737378) *always* has a meaning, so if 2 means ''if'' then the 3 that follows it is a question, if the 2 means ''move'' then the 2 is used as a direction.

Incorrect about the universe simulator!

You''d need some storage device to store the basic states of every particle in the universe... how would you do that using less than a particle?

So a universe simulator would be at the very least as massive as the entire universe! Sorry, buddy boys.


suppose you wished to simulate a small portion of the universe instead. Two things exist to keep it from truly representing a physical "reality"- heisenberg uncertainty principle, and the fallacy of a closed system.

but i''m sure that''s not as detailed as where you''re thinking of going. You just want to simulate CHEMISTRY, not physics.

Note: if you want to create a simulation of chemical life, why not use regular chemistry? look at this thing called the Miller-Urey experiment. So far it has not been able to produce life from any chemical soup. What makes you think a computer will? Because it will be faster? No. Chemists have their own ways of speeding up chemical reactions. Another issue is "how will you detect the presence of ''life'' in your simulation?"

qt:

Incorrect about the universe simulator!

You''d need some storage device to store the basic states of every particle in the universe... how would you do that using less than a particle?

So a universe simulator would be at the very least as massive as the entire universe! Sorry, buddy boys.

endqt

if you only want to simulate the pure action of this universe, there''s no reason why quantum effects can''t decode whatever pieces of information you want to faster than the universe itself can evolve those pieces of information. See, for example, quantum computers. This may in fact be (one of) the secret(s) of conciousness (no scientific evidence of this provided b/c I haven''t seen any ;P)

If you want to simulate the action of ALL quantum realities, well, you''re just completely perverted


mikey

Yeah, i agree with you, it''s impossible to store all the data of the universe, because each atom will require much more than one simple byte, and how we will store various bytes in a physical space smaller than an atom? We''ll need a computer bigger than universe to do it

This may not be completely on topic, but:
1)There is an interesting fact that most people forget when thinking about this subject: DNA is written in binary!!! There are 4 genomes(They''re called something like that): GATC. An ''A'' can only pair with a ''T'' and a ''G'' can only pair with a ''C''. Therefore there are only two possible combinations, making it a base 2 number system. ie. binary! I suppose that this is mostly irrelevant to the discussion, but with computers get bigger and faster eventually they will be able to handle complete human DNA. This means that we could program in "ourselves" somehow. I don''t know. I guess I just find that interesting.

2)Also, witb the internet(ie. basically every computer in the world connected) does this not present interesting possibilities for AL? Suddenly, instead of running an evolution program on one iscolated computer you could run hundreds, thousands, millions of copies all over. It seems to me that this change the scope of things entirely. I''m not very knowledgable on AI though, so I''m really just guessing. Hey, maybe I''ll inspire someone.

3)What is it with AI and long posts? Anyway, talking about AL, wouldn''t it be interesting if you could expand the evolution possibilities so much that they could possibly modify they''re own code? This would complicate things considerably, but if it worked you might get some incredibly interesting results. Hey, the next time you have a math problem due make AL that''ll survive based on coming closer to a solution. Then just leave it on all night and in the morning... POOF!!! your computer has crashed because one of them wrote illegal code. Anyways, I''m really rambling now...

Keep on theorizing, so we don''t have to prove it!!!
-Mark
chaos1111@hotmail.com
ICQ:22527985

Gregor_samsa:

no, it is not written in binary. It''s base 4. A-T and T-A are not equivalent, neither are C-G and G-C. In fact, if you study the protein transcription process (studied in the 1970''s), you''ll see that a codon is 3 base pairs, and thus there exist 4^3=64 possibilities for transcribing amino acids (some are duplicated). They''ve got the code for transcribing each amino acid pretty down pat.

The fact that we do have evolving programs already is good enough to know that we CAN- the only problems we have to face are the "environmental" conditions and the practical extent of self-modification.

I don''t know how far discussion can go in terms of furthering our knowledge past the current bounds. Most scientific advancements come accidentally rather than deliberately. Interesting question though, and we all could be helped by being up to speed on the current issues...

I havnt read the entire thread so if this is covered, forgive me

I just want to point out that if you are using a matrix that is filled with comands, the creature will possibly learn new behaviours, but it will not learn new actions. For instance, 1 i your matrix means go forward. So the creature is permanently only able to go forwards in one way.

How do you let it evolve a behaviour like move forward and left at once.... I think if the interpretation of the matrix is static, the creature will never be able to properly evolve, if you do not have a Jump to instruction implemented in the matrix, it will never learn to do stuff in loops, for instance...

I think a different system is needed to try model evolution, but your idea is good for developing behaviours inside a fixed frame...GA style

There was an interesting program a while ago about little rectangle shapes that flip over and compete with each other, does anyone know the research I mean?

also I saw a nice program for evolving 3d geometric shapes, it had a ''matrix'' of instructions for drawing 3d objects, like base shape, how it is tiled, scaled, coloured and arrayed in 3d.. and it would produce a selection, and you pick your favourites, and it would combine elements of your favourites into a new panel of choices.. in this way the users input is the fitness algorithm, and it just mutates the designs somewhat until you are happy with what you have....