How good is SuperPower's artificial intelligence?

            "The A.I. reacts to problems, or possible problems," SuperPower's creator, Jean-René Couture, once related to me, "A neighbor assembling a big army is a possible problem, and the A.I. will react accordingly, without making a difference if the neighbor is human or computer controlled. I think the big question is how will it react?  We can't really say ourselves.  I don't code pre-determined patterns of thinking.  The A.I. builds itself up from experimenting actions according to the current data. If it finds that buying iron from Nation A     starts a chain reaction that will eventually solve the military problem, it will do so.  That's the beauty (and complexity) of SuperPower."

            "Our intelligence model can do many things, but most of all, it can learn. It can think up new strategies and new ways to solve problems. It can think ahead and plan his actions by using simulations...  We'll give it a basic strategy it'll learn by playing against itself, but the rest will be relative to your gameplay... The A.I. will learn to play by playing with you."
 

The A.I. actually learns as it plays?

        Yeah.  Pretty cool, isn't it?
 

What is the A.I. code design based upon?

        The A.I. engine was developped  using techniques similar to neural nets, expert systems and Q learning, but also some psychology theories like the Maslow chain of needs.  It has been certified by professors in the artificial intelligence field as a new path of research.

But are the computer's actions doing to be realistic?

        Jean-René recently said in GolemLabs' Discussion Forum, "That is the most difficult part of the AI coding: not making it do intelligent things, but making it not do stupid things.  So far, it seems to hold.  We had to enter checks and double-checks all along the decision-making line, but it's doing the right things.  We try not to include too much of that code since we want the AI to make it's own decisions, and we want it to adapt to a changing situation.  So we couldn't just say: 'The US never invades Canada' because that would open the door to a canadian player to simply invade the US without retaliation."

        "But the code that was in [gave] really accurate results.  When we first launched it, and printed the "possible targets" of the US (country's it wouldn't mind attacking), we got Iraq, North Korea, China and Russia.  And keep in mind, we didn't tell it how to behave, it got those countries simply by analyzing the world.  So I think it'll hold."

        In the old Defcon Forum a poster once asked, "Do arms races ever occur?  For example, if you start building up a massive military force, what will
neighboring/opposing countries do?  Will they usually start to build up weapons as well?"

        Jean-René replied, "That's an interesting question... I can't answer it, because of the particularities of the EHE [evolving human emulator]. I don't know what it will do.  It won't do the same things every time, and every version won't do the same things.  Of course, if you build up your military, opposing countries will see this as a treat (possible problem), and try to correct it.  The way to correct it is to have a balance in forces, so either theirs must go up, or yours must go down.  To build up, they could construct, or create military alliances. They could ask for foreign powers to defend them, even build military bases on their territory.  To bring yours down, it could try sabotage, economic attacks, or start a war between you and somebody else. These are the options I think of rapidly, but the EHE could find out that by some course of events, raising it's taxes could bring your military power down. Don't know how, but it's just an example."

        "The point is that we won't give it predetermined paths of thinking. It will do something, that's for sure.  But what? That's the beauty of SuperPower. It will do
something it learned by playing against you, so the more efficient you become, the more efficient it will become also.  And it won't do stupid things many
times over.  It will learn, and do something else. So no one will be able to answer that question for you, unless you play the game a couple of times, and find out for yourselves.  But, believe me, it will try to stop you. Probably even succeed in giving you a couple of headaches along the way..."