[Combining AI, AL and the science of complexity] - A New Kind of Science: The NKS Forum

A New Kind of Science: The NKS Forum

Pages:1


Combining AI, AL and the science of complexity

(Click here to view the original thread with full colors/images)


Posted by: Tommy Petersen

Hi there and thank you for this forum.

I want to share with you an idea I have to integrate the study of artificial intelligence, artificial life and the science of complexity. It is based on using a cellular automaton so I think it is applied NKS.

I have attached a paper describing the idea. In short I embed an autonomous agent in a cellular automaton and later in the paper I consider a population of agents controlled by a population of strings. These strings are used in the study of artificial life.

I am very interested in comments or questions on this idea.

By the way, I read that Stephen Wolfram gave a talk about the future of artificial life at the A-life conference in september 2004. Can I read about what he talked about somewhere?

Thank you,
Tommy Petersen


Posted by: MikeHelland

I have skimmed through your paper (I really should be working) and I am intrigued enough to read it more carefully, mayhaps this weekend.

One quick comment, you say:

"3 Evolution of time in the CA
In this kind of CA, time does not exist on its own. Time in the CA is induced
by the actions of the agents situated in the CA, such that a time step occurs
in the CA if and only if an agent makes an output. Whenever an agent
makes an output, the whole state of the CA is updated using the CA-rule
extended with handling of cells containing agents as described previoulsy."

And I agree overwhelmingly.

I believe that if you begin to flesh out your idea, and really start to discover its consequences, you will have to find a model where what you have to say about time is equally true about space and matter.

You might find this very helpful:

http://www.techmocracy.net/science/time.htm

There are other several other threads here about time where the position you've written about is also discussed.


Posted by: kenr

An MP3 of Wolfram's talk at ALife 9 is provided by Complexity Digest here:

http://complexity.vub.ac.be/~comdig/ALife9/Wolfram.mp3

_Ken


Posted by: Tommy Petersen

I have made some java-applets demonstrating the CA-embodied agents, see http://www.ai-agents.com/Research/Demonstrations/.

There are some changes to the original sketch posted. These changes are described at the above url.

One of the demonstrations shows four agents transporting a static structure of size three. Are there any static structures of size greater than one moving around in a normal 2D-CA?

- Tommy


Posted by: Todd Rowland

The short answer to whether there are structures which move around in any 2D CA is yes. It depends on which 2D CA you mean, and what you mean by structures. For instance, look at the typical behavior of some of the 64 totalistic rules (5 neighbor)(p.246), as well as the variations with more neighbors.

As far as I know it has never been nailed down, exactly what the 5 neighbor totalistics do.

Stephen Wolfram uses the term local structures to refer to structures which are well-defined in their nature and extent, like those found in class 4 rules like the 1D rule 110(p.677).

It would probably be helpful for you to separate out the components of your system, and identify them with systems already studied. For instance, the type of CA you seem to be discussing is calledoutertotalistic. The agents are acting likegeneralized mobile automata. (It is not clear from your description what happens when they collide). Later, your agents' actions are history dependent. There are many ways to do this, for instance global control,finite state machines, or even coupling the agents behavior to a completely different CA or Turing machine.

Make some decisions about the components, keep it simple, and try to implement it, probably in one dimension before trying two.

The second point is about the underlying argument. It is not clear what this has to do with AI, AL, or complexity. I am sure you have some interesting ideas, and it would probably make things more clear if you had a set of simple examples. Look at Stephen Wolfram's book, and you will see that he bases his conclusions on a set of experiments, as simple and as unbiased as is possible. Keep in mind that you can get complexity from even the simplest of setups, and it is better to search rather than to bias how things should work. In the end, what is the minimal setup needed?




Forum Sponsored by Wolfram Research

© 2004-2008 Wolfram Research, Inc. | Powered by vBulletin 2.3.0 © 2000-2002 Jelsoft Enterprises, Ltd. | Disclaimer
vB Easy Archive Final - Created by Xenon and modified/released by SkuZZy from the Job Openings