What does the AI know?

[AlrightyThen]

Does the AI know at the beginning of the game which forms of victory are allowed for that game?

Does it know when the human and/or other computer players are close to one of those goals?

Does it know both of the above?

I'm getting attacked a little too coincidentally for no discernible reason as I come close to reaching a victory condition.

[Spaced Cowboy]

Originally posted by AlrightyThen
Does the AI know at the beginning of the game which forms of victory are allowed for that game?

It would have to because it is playing towards the same victory conditions as you are.

Originally posted by AlrightyThen
Does it know when the human and/or other computer players are close to one of those goals?

I suspect that it does. Whether it uses the embassy and spies to determine or it has a hook is unclear. However, there is strong evidence that points to intimate knowledge of your empire.

Originally posted by AlrightyThen
I'm getting attacked a little too coincidentally for no discernible reason as I come close to reaching a victory condition.

Try to trade places with the AI in that situation. Would you attack yourself?

R:PM

[delfino]

The computer knows everything.

Your weakest spots and your stong characteristics.

It will always target your least defended city, like magic one of their units will pop up near your workers, will rush build a wonder even in a miserable city and without any real way to get a leader, their regulars will sistematically defeat your elites specially if you are trying to get a strategic city. Furthermore their cities pop up near strategic resources, even if several squares away from their main empire.

That is why we love this game.

[=DrJambo=]

if it didn't get some help it would be too damned easy, much like civ2 where the AI would constantly throw its units against well defended cities!

[chiefpaco]

Does the AI know at the beginning of the game which forms of victory are allowed for that game?

Excellent question. This is on my top 10 suggestions for the next Civ game.

From what I've seen, the AI is vaguely aware of something called the Space Ship. That's about it. It will build the UN & hold a vote, but I don't know if it accepts it in a strategic move.

Right now, the AI seems to be built to make the human lose. I'd prefer it to go for a win instead. Perhaps, at some point in the game, preferably before the modern age, each AI can select a Victory condition to gun for. It would have to best suit their situation & ability. Like, a large warmongering empire might go for domination or conquest. A high culture religious civ would go for a culture.

As a result, currently, we have very few "races" in the game, where the human has to thwart an enemy winning, other than by SS or UN. We only have the human trying to sustain pressure from all civs to make us lose.

Furthermore, by getting rid of the "Double-culture" rule for a Cultural victory, we would get a much better race to the limit.

[benjdm]

Tougher question. Does the AI know what selections the human player made for map generation ? When I play islands, the computer does not seem to place enough emphasis on Lighthouse / Astronomy / Navigation etc.

[Nym]

Originally posted by delfino
Furthermore their cities pop up near strategic resources, even if several squares away from their main empire.

I am quite sure of this too, but I get a surprise within my present game. French are located south from me. Many turns ago, we both went at war against the Zulus, north-west from my empire. I captured most of the ennemy's cities and kept them, while the French razed those they captured, except one. I found this really strange because the city was in the middle of toundra, without water or any known ressource. I was so quite sure that aluminium or uranium will appear there (I was the most advanced and still in Industrial Era). And... nothing.
Very strange. The city is now completly surrended by my empire and about 20 tiles apart form the french empire. I do not understand.

[rah]

Fist time (and only time) I ever saw a nuke launched against me. Was it aimed At my super capital or a strategic city at the front, Noooooooooooooo, rather the city that was building my last SS part. The AI knows almost everything.

RAH

[Jamski]

Just a quickie, has anyone tried using a spy to catch a spy - so to speak - maybe the AI has infilterated an agent, and if you don't look... I check about every 10 turns on Regent+ and I find them quite often, so maybe this is how they know where your workers are, even if they can't see them, and do those stoooopid "worker round-up" attacks....

-Jam

[Coracle]

Originally posted by rah
Fist time (and only time) I ever saw a nuke launched against me. Was it aimed At my super capital or a strategic city at the front, Noooooooooooooo, rather the city that was building my last SS part. The AI knows almost everything.

RAH

The AI cheats outrageously. And that's another reason why I've turned the Manhattan Project into a comerce-producing Internet Wonder.

[Inverse Icarus]

i swear the AI is more agressive is domination is the onlyvictory type... and rightfully so.

[player1]

Originally posted by Coracle


The AI cheats outrageously. And that's another reason why I've turned the Manhattan Project into a comerce-producing Internet Wonder.

Don't forget you can do the same.

Spy cities, and find out whioch builds SS parts.
Then nuke it.

Also nuke Aluminum connections.

[LaRusso]

Only in my last (1.21) game did AI try to build a spaceship. Then, however, they all rushed and I won only because my research was quicker.
BIG note for XP: Firaxis, consider having ALL techs as prerequisite for spaceship. It is stupid, anyway, to have a ship to Alpha Centauri before stealth or SDI.

[Hurricane]

I´m pretty sure that the AI has no clue about victory conditions. The AI will build a courthouse in its capital, so why would it not build space ship parts even if it hadn´t a clue what they were good for?

The problem is, of course, that the AI scripting is very simple. I would say the AI reasoning from beginning to end is something like:

if { unsettled land = build settler
else build improvement (50%) or military unit (50%)
except if wonder available = build wonder
or if at war = build military unit }

research % = maximum without having a negative income

trade:
buy everything you can afford
sell everything anybody can afford

[Akka]

Try to trade places with the AI in that situation. Would you attack yourself?

R:PM

I am playing a game. I would like the AI rather act like if it's was its very existence that was in the balance (ie : no systematical "I'm a 5-cities large civ with gunpowder and I will attack this continent-size empire using hundreds of modern armors").

[Hurricane]

Originally posted by Nym
I am quite sure of this too, but I get a surprise within my present game. French are located south from me. Many turns ago, we both went at war against the Zulus, north-west from my empire. I captured most of the ennemy's cities and kept them, while the French razed those they captured, except one. I found this really strange because the city was in the middle of toundra, without water or any known ressource. I was so quite sure that aluminium or uranium will appear there (I was the most advanced and still in Industrial Era). And... nothing.
Very strange. The city is now completly surrended by my empire and about 20 tiles apart form the french empire. I do not understand.

They probably made a decision each time they captured a city if a) they hade a good chance of keeping it and b) the chance of the city flipping. The small tundra town had probably no culture and was small in size, so the AI concluded the chance of keeping the city was high. Of course, the AI did not take into consideration the fact that this only city soon would be inside another civ´s borders.

[SpencerH]

Originally posted by player1

Don't forget you can do the same.

Spy cities, and find out whioch builds SS parts.
Then nuke it.

Also nuke Aluminum connections.

Given the cost of spying on a city and the number of citys the AI (or we) must check, and the limited funds the AI has is it reasonable to think that occurs? I dont think so. I think the AI is all knowing.

[DrFell]

Yep, the AI does have considerable 'cheated' knowledge of the game situation. But then again, a human can make educated guesses as to the strength of an enemy empire with much less information than the AI would take to do the same thing, so while the peeking is annoying, it is necessary.

[LaRusso]

Originally posted by SpencerH


Given the cost of spying on a city and the number of citys the AI (or we) must check, and the limited funds the AI has is it reasonable to think that occurs? I dont think so. I think the AI is all knowing.

well given that we are looking here for an appearance of AI competence, and not the competence itself, this is acceptable

[SpencerH]

Good points. I'm still 50/50 with it though. I definitely dont like the amount of info we (and the AI ) have about other civs in ancient or middle ages though.

[Spaced Cowboy]

Originally posted by Akka le Vil


I am playing a game. I would like the AI rather act like if it's was its very existence that was in the balance (ie : no systematical "I'm a 5-cities large civ with gunpowder and I will attack this continent-size empire using hundreds of modern armors").

I am not quite sure that I completely understand your point, but if you are winning or appear to be the biggest threat to win, why wouldn't the AI attack you, even if you are stronger. Maybe it thinks that it will be able to get friends to help ?

Personally, I think that the AI would be better suited to pick on the small dogs instead of taking on the big dog. If the AI swallows a few smaller players, then there should be a better chance to pick on the biggies.

Then there is the Sadam scenario. Explain that one!

R:PM

[Akka]

Originally posted by RPMisCOOL


I am not quite sure that I completely understand your point, but if you are winning or appear to be the biggest threat to win, why wouldn't the AI attack you, even if you are stronger.

Because it's anti-immersive.
I don't mean that I want that nobody is never wanting to fight a big dog, but I would like the AI acts like if it was NOT a game, hence acting more than like a state leader and less like a player that will just loose a game if it gets eradicated.

Maybe it thinks that it will be able to get friends to help ?

Personally, I think that the AI would be better suited to pick on the small dogs instead of taking on the big dog. If the AI swallows a few smaller players, then there should be a better chance to pick on the biggies.

Then there is the Sadam scenario. Explain that one!

R:PM

The computer players DO gang-bang on weakest civ. Would like, again, see less mechanical leaders that would not act only according to supposed efficiency, but considering also friendship, hatred and moral constraint.

[delfino]

Originally posted by Hurricane
I´m pretty sure that the AI has no clue about victory conditions. The AI will build a courthouse in its capital, so why would it not build space ship parts even if it hadn´t a clue what they were good for?

The problem is, of course, that the AI scripting is very simple. I would say the AI reasoning from beginning to end is something like:

if { unsettled land = build settler
else build improvement (50%) or military unit (50%)
except if wonder available = build wonder
or if at war = build military unit }

research % = maximum without having a negative income

trade:
buy everything you can afford
sell everything anybody can afford

Did you have any saying in the Civ3, computer programming??
Bullseye!!
Exactly what I think.

[Random Passerby]

The single concept that I find most lacking in the AI's understanding is balance of power. Here's a scenario that I've seen unfold many times:

Two great powers (let's say England and Russia, for sake of argument) are located on the same continent. Both are AI civs, and tenuous allies. At some point in the industrial age, that tenuous connection snaps and they go to war. They're evenly matched, and the war swings back and forth tiny amounts for many turns, never more than the same one or two cities changing hands. Then, let's say the English decide to pull in a few allies; they call in Russia's smaller neighbors, say, the Aztecs and Indians, and they happily join England, eager to be allied with such a powerful nation. Now Russia's forces have to be split; the Russian forces easily trounce the tiny Aztec and Indian armies, but now they have fewer forces to fight off the English, and England begins to break through. Now the "allies" here pull in a few more civs into the fray, and soon the entire continent and quite a bit of the rest of the world happily declares war on the Russians, seizing the opportunity to join the winning side and see a major power defeated.

The outcome of this war is certain: England will conquer the vast majority of Russia, swiftly destroying one great power and thrusting another into near-superpower status. The smaller civs bordering Russia are likely to gain little, if any land; some of them will even LOSE land first to Russia, only to have it reconquered by the English. Various random powers across the globe will grab a few tiny cities, though they will never get any use out of them due to their being half a globe away. Now England is in a very dominant position, not from her own efforts, but because the rest of the AIs were overly eager to all gang up on the losing side of the war.