It's 2005, why does the AI still need to cheat?

[Urban Ranger]

Originally posted by padillah
I'm saying that it's computationaly infeasable to create a fully functioning "AI". The research hasn't figured out the perfect algorithms and the technology can't support the algorithims they have developed.

As far as I can tell, the "AI" is just an expert system. At least the core is an expert system, probably got other stuff thrown in for good measure.

[Urban Ranger]

Originally posted by padillah
And they weren't even going for AI. They just had to automate a vehicle.

That's a completely different beast.

[Azuarc]

Originally posted by nbarclay
The catch to that approach is that it does not allow the AI to coordinate its units effectively. That's not a problem if the units are so far apart that their activities are essentially independent of each other. But when higher levels of coordination are needed, the AI can't afford to wait until after it's moved its first unit before it starts thinking about what to do with the second. It has to consider all of the things it might want to do with its second and third (and so on) units - not just in that turn, but also in future turns - as part of the process of deciding what to do with the first unit it moves. Once the AI does that, you're back into the realm of very, very steep exponential complexity.

That's fair enough, but if you have the AI decide certain goals it has...take this city, pillage that resource, colonize this land mass...and weight its actions - at least ones that aren't hopelessly uninvolved like the scout in Siberia trying to get into a war in South Africa - in favor of those goals, with queued prefered actions per turn, it just might be possible to have the AI perform something cohesive with relatively independent decisions.

I hope you didn't take offense. I was joking.

None taken. I knew I was setting myself up for the fall. After all, that is what pride cometh before.

[Drachasor]

Originally posted by Azuarc
That's fair enough, but if you have the AI decide certain goals it has...take this city, pillage that resource, colonize this land mass...and weight its actions - at least ones that aren't hopelessly uninvolved like the scout in Siberia trying to get into a war in South Africa - in favor of those goals, with queued prefered actions per turn, it just might be possible to have the AI perform something cohesive with relatively independent decisions

I think there is a fair amount of room for improvement with the AI. You can make it more willing to take troops out of its core cities to take back and defend its territory. Right now it is far too conservative even when it has a force that could overwhelm you.

The AI needs a little more give and take. I'm sure the community will improve on the AI bit by bit.

-Drachasor

[Willem]

Originally posted by Sleestax

It is exactly because it is hard that I am proposing that Firaxis open the AI code to the open source community.

As mentioned already the AI code is going to be released. That's been Firaxis plan since before the game was released. It will probably be available in January.

[Strudo]

Originally posted by Solver
True AIs haven't been invented yet. So all that programmers are stuck with are making algorithms to make their AI play the game optimally. For a game as complex as Civ, where you BTW couldn't go through all the different move combinations if you tried to, this is a very non-trivial task.

I have often though that the term AI should be changed to SI, or "Simulated Inteligence". It better describes whats going on internally, and doesn't give any false hope that the program is actually smart.

Just what I thought...

[Willem]

Originally posted by Urban Ranger

The moves for the three warriors are independent unless they are in close proximity of each other.

That's not true. All units, at least military ones, are connected simply by being part of the military. If a war is going on, the AI has to analyze where it's taking place, has to calculate a likely target, then adjust the moves of those two units accordingly. If there is peace under way, it has to determine the most likely civ that will declare war next, then send troops nearby in order to counter any future attacks, or preferably choose a spot where it can defend attacks from several civs.

It makes a lot more sense to think of the entire military as one unit, and all the individual units as simply components of this.

[Alashar]

Look at it this way... What size computer did they use to defeat the Grand Master of Chess? A supercomputer. i.e. a REALLY REALLY Big computer that is super specialized and probably cost over $50 million dollars.

The best AI you'll get in most games is as someone posted above, simulated. It will use tricks and cheats in order to provide humans with enough challenge to keep them occupied. If you really want a challenge, go multiplayer and play some other humans.

[padillah]

Originally posted by Azuarc
There aren't 729 choices for moving 3 units one space. There are 3 to decide who to move first, 9 to move it, 2 to decide the second, 9 to move it, and then 9 to move the last, for a total of 32 total evaluations.

Yep, 32 evaluations... of 729 different choices.

And this is not that big a deal. I was writting a test script for a billing system and we came up with 10,000 different scripts in the first phase alone. Just because a branch of logic isn't followed doesn't mean it's gone. Or doesn't have to be coded for.

You are refering to distinct choices made, I am talking about the ability to make those choices. It's like saying because you had McD's for lunch there was only one choice made. Yes, but how many options did you evaluate?

Tom P.

[padillah]

Originally posted by Urban Ranger

Originally posted by padillah
And they weren't even going for AI. They just had to automate a vehicle.

That's a completely different beast.

True. It's a much smaller task with much more confined criteria. It doesn't have to plan several moves in advance it just has to get around this obstacle in front of it.

By the same token, it could be considered to be more difficult since it's interacting with the real world which is messy and inexact.

So, take your pick.

Tom P.