http://www.newscientist.com/article.ns?id=dn7674

European scientists have been making virtual societies using artificial intelligence. They are trying to mimic human society to study and get answers to questions about real society. BTW they are using the Source Engine.

Meh.. I really don't understand how you can learn anything about people from a simulation coded by a bunch of programmers. The AI clearly isn't even close to the behavior of real humans. If you want to learn about human behavior, I guess I'm saying wouldn't it be better just to study.. humans?

I know, but the interesting thing is that they picked a game engine from popular game designers for their work.

That's almost a non-issue when you consider what they're trying to achieve.

The fact that they're using source demonstrates that graphics aren't really the focus at all. (Heck, they're using Breen there).

It actually says they are using source ONLY for the graphics, and they are making their own program to design the behavior patters and whatnot.

Hence what I said.

Because if you can reconstruct human society from scratch, you know exactly what it is composed of.

They start off with some basic rules of interaction and stuff, leave it running for a while and see what they get. Most of the time they'll get something weird and crazy (but fascinating and interesting nonetheless), but then they'll tweak it slightly and try to create something that looks more like human society.

I also think this will be significant for Richard Dawkin's theory of Memetics:



Basically, in that ideas or thoughts or information generally may be spread by one individual to other individuals, who will in turn spread it onwards, and the ideas/thoughts/information that is most useful at the time will spread the most.
This virtual world will help to show exactly what constitutes as 'useful' in terms of ideas/thoughts/information.

Fascinating stuff.

Where is the source code for this thing?

But what happens when they become aware they're not real?

It launches its missiles against the targets in Russia. Because Skynet knows the Russian counter-attack will eliminate its enemies over here.

Well, we haven't.

In that case, what happens when the AI creates its own AI? :p

Wouldn't it be awesome to see them come up with their own little religions? The simulation would have to be extremely powerful though.

It'd be more amusing if they turned out to be atheists

If a large part of the population turns out to be atheist, something is wrong with the simulation.

What do they probably have to discuss first? "Hi!" "Hello!"

But how can they prove that they weren't created by intelligent design (by the almighty Valve)? Along they way, they might figure out they were created through the eyes of somebody. There isn't exactly science, evolution, historical data, etc. to support "atheist" views in their little world. So, being atheist would be wrong, I guess.

They could argue who is the original "God" and start a holy war. That would be awesome.

IMO, its completely impossible for humans to program another human. Our brains are way too complex and we don't know nearly enough about them. It's impossible. At best we get shotty robotic AI.

Artificial intelligence as "smart" as we are is entirely possible. Computers can indeed learn...if we'll be able to achieve such AI anytime soon...well, probably not.

Not entirely impossible. Programmers have already successfully created software to simulate how the human brain learns. This topic, is, of course, artificial neural networks. While I'm not sure about the age of the subject, it did not just appear yesterday. It's actually fairly easy to find articles and tutorials on neural nets. You might want to google the "Hopfield Network" for a neural net story.

While I'm not saying neural nets by themselves will allow scientists to easily recreate an artificial human society, I wouldn't be surprised if it played a vital role.

In sociology the individual level doesn't matter as much. We don't have to look at individual action and psychology, just at group level dynamics. It's still complex as ****, but it should still be interesting, and variables could be added over time.

Heres a good idea to what a Neural Network can do (besides character reconition)

http://www.mindcreators.com/



Another:

http://www.seattlerobotics.org/encoder/200205/abionin.htm