What do you expect from the AI in Civ 4

[eddie_verdde]

I expect the AI of CIV4 to:

1 - not declare war if my civ is on the other side of the globe (let's say, in the Age of Discoveries scenario, what's the point in Iroquois declaring war on Portuguese if the Portuguese don't have any unit or colony in North America?)

2 - not send stacks and stacks of units across the world to attack me, then make peace and withdraw all those units again (what's the point in sending dozens of units to a place that is dozens of turns away, if there is no interest in resources, cities or land? is the AI only a bully?)

3 - not to try to "invade" my continent, making landfalls of 2 units at a time

4 - not build cities wherever it can, even if the spot has no productivity and is very near a enemy

5 - accept an offer of 4000 Gold and some luxuries/resources for a stinky city in the middle of nowhere.

6 - not give away all its fortune for a stinky city in the middle of nowhere.

7 - not make global military alliances against me, just because...(god, do I sometimes feel alone)

8 - not cheat (come on!!where did that spearman in Syracuse came from??)

9 - respect alliances and peace treaties more than 10% of the times


among other things that I can't remember right now

 

[KrikkitTwo]

I expect the AI of CIV4 to:

1 - not declare war if my civ is on the other side of the globe (let's say, in the Age of Discoveries scenario, what's the point in Iroquois declaring war on Portuguese if the Portuguese don't have any unit or colony in North America?)

2 - not send stacks and stacks of units across the world to attack me, then make peace and withdraw all those units again (what's the point in sending dozens of units to a place that is dozens of turns away, if there is no interest in resources, cities or land? is the AI only a bully?)

3 - not to try to "invade" my continent, making landfalls of 2 units at a time

4 - not build cities wherever it can, even if the spot has no productivity and is very near a enemy

5 - accept an offer of 4000 Gold and some luxuries/resources for a stinky city in the middle of nowhere.

6 - not give away all its fortune for a stinky city in the middle of nowhere.

7 - not make global military alliances against me, just because...(god, do I sometimes feel alone)

8 - not cheat (come on!!where did that spearman in Syracuse came from??)

9 - respect alliances and peace treaties more than 10% of the times


among other things that I can't remember right now

1..Why not, that makes it a low cost thing to do, and it might gain diplomatic points

2.. Well that's planning ahead, perhaps the Situation changed

4... possible access to resources, grabbing terrain

9..do YOU respect alliances and peace treaties, better yet is it a game winning strategy to respect alliances and peace treaties (the AI probably undervalues it, but when playing against a Human...I can just imagine what the AIs would complain about in terms of the Humans they have to play, and how they are so totally unrealistic)


Overall some good points though, AI needs to take more into account in its planning

 

[apatheist]

Isn't it really tough to program a computer for Go?

The best Go program in the world has yet to beat the best Go player in the world, AFAIK, but it can still beat most players. We're not talking about the best, though; we just need it to be good enough to beat most players without cheating. Besides, Go is a much older game, so the accumulated knowledge of human players is much greater than it would be for Civ, so the bar is lower. Also, Go is visually simple, so a player can focus on the essentials more easily, while there are a lot of superficial variations in the Civ world that make it harder for the human player to "see the Matrix." Finally, Go is a little too simple; you can describe the game in a single paragraph. I believe that Civilization should have a simple set of rules, but not that simple. I do think you could have a pretty good Civ game described on 1-2 sheets of A4 paper in a readable font size, and I think that game would most likely be better than a Civ game that takes 10 sheets of paper to describe. I'm not talking about the manual; I'm talking about describing the game's core.

 

[warpstorm]

I do believe the design doc is a bit bigger than 2 pages. Hell, the design doc for the one I'm working on is well over a hundred and it isn't as ambitious as Civ.

 

[Dida]

I expect the AI to be able to make some global strategic decision. The AI in civ3 is stupid, and can never see the big picture.

 

[apatheist]

I do believe the design doc is a bit bigger than 2 pages. Hell, the design doc for the one I'm working on is well over a hundred and it isn't as ambitious as Civ.

I used imprecise terminology. What I meant was, you should be able to describe the mechanics of the game in 2 pages. It doesn't spell out every single implication, but it does convey sufficient details that the implications are logical and obvious. It explains what corruption is, how trade works, how railroads work, what food is, etc. It doesn't describe every unit or every civ trait or graphics or sound or improvements or wonders or whether jungle adds a 10% defense bonus vs. 15%. The former is code; the latter is data. The former is fundamental design principles; the latter is refinement and tweaking.

 

[Superkrest]

I'm not really a fan of the AI in Civ3 and I don't see how it was an improvement over any of the previous Civs. The AI is way too mercurial to be considered good. It does not value trade in the least. There was one instance where I had been trading with one AI since the dawn of time and then it declares war on me for no reason. I've always said this but there needs to be more reasoning behind AI war declarations.

the great wall of antartica...the massive(one unit) expired unit invasion...the building a city one square from yours. the everybody pick on the BIG guy. lol...but civ 3 was still stupid..but i think a massive improvement ..so hopefully they'll make an even bigger leap in civ4

 

[Urederra]

Also, Go is visually simple, so a player can focus on the essentials more easily, while there are a lot of superficial variations in the Civ world that make it harder for the human player to "see the Matrix."

In Civ III and therefore, I suppose also in civ IV the luck factor is important, whereas in Go or chess there is no luck factor. You don't roll any dice in Go or Chess, but you do in Civ III and, I suppose, in civ iV.

 

[Urederra]

I'm not really a fan of the AI in Civ3 and I don't see how it was an improvement over any of the previous Civs. The AI is way too mercurial to be considered good. It does not value trade in the least. There was one instance where I had been trading with one AI since the dawn of time and then it declares war on me for no reason. I've always said this but there needs to be more reasoning behind AI war declarations.

Maybe because it is programed to win. Won't you do the same? C'mon, suppose you have a trade agreement with the egiptians, but, in order to win, you have to conquer them (there might be a lot or reasons, because the egiptians are growing too fast, they are winning against the persians and they are conquering more than a half of the continent... etc.) Then, what would you do? Not to attack because you had a trade agreement?


Oh, no, I cannot attack the egiptians because they are my friends and although they are going ahead faster and faster, I cannot attack them because they are my friends. I would attack, sincerely.

Best regards,

Civ III AI

 

[jkp1187]

Maybe because it is programed to win. Won't you do the same? C'mon, suppose you have a trade agreement with the egiptians, but, in order to win, you have to conquer them (there might be a lot or reasons, because the egiptians are growing too fast, they are winning against the persians and they are conquering more than a half of the continent... etc.) Then, what would you do? Not to attack because you had a trade agreement?

Agreed. The AI can meta-game all it wants, since the good Civ player will do that, too. It's the AI's tactical control of units that needs to be addressed.

As an observation -- I didn't notice much improvement between the AI in Civ I and II (mostly b/c Civ II was just a gloss on the game engine of the first game, not a true ground-up re-write.) The AI in Civ III, in my opinion, was a VAST improvement over what went before.

 

[Avayaman]

firaxis has a tall job of identifying what makes a good civ player and build the AI from there, instead of identifying what makes a bad civ player (building cities on the artic, etc) and forcing the AI not to do those. AI then becomes a semblance of an OK player, because it doesn't make silly mistakes, but this means that in order to have amore difficult game, AI definitely has to cheat.

 

[apatheist]

Agreed. The problem isn't that the AI breaks agreements, it's that it does so foolishly.

Urederra, your point about luck differentiating Civ from Go is a good one. An implied corollary is that Go and chess boards are the same simple board every time, while there's a near infinity of possible Civ worlds. Both those elements mean that the AI has a better chance of being able to challenge the human player.

 

[eaglefox]

the points reflected by most people here match my own. the AI needs to act in a way that will benefit its country in the long run. i don't want to see present day east timor attacking america for anything. i want to see the ai calculating at every step whether that step is viable or not, whether there's any gain from it, and what the risks are. i want to see a better trading system. i want to see quantitative resources. not just arbitrary graphics on the map that supply the entire nation no matter what the demand is. if these changes are not available in civ 4, then it holds very little interest for me, because it'll be like playing civ 3 with civ 4 graphics. i don't wanna spend money on that.

 

[Himalia]

I certainly want to see an improvement with alliances. More detail at the very least both sides must state the minimum that they will put forward in this alliance or protection pact. Im happy when the AI just sends a loan spearman or sometimes a couple of militia just for them to leave as soon as possible. More depth with MPP. How often have you had the AI get a MPP with you for it to declare on someone else and for them to attack the AI back. I dont like that i feel that a MPP could be ignored if they are the protagonist just as one additional option.

 

[Arizona_Steve]

I'm waiting to dig into the AI SDK that's supposedly coming out alongside the game. I think that's where the AI improvements will be seen...

 

[MasterChief]

I was always disapointed to reduce another civ to one city only to have them demand stuff from me a few turns later then declare war again when i refused to give it to them. Do the Carthinagins really believe they can win that war with one spearmen? No one can win a war against someone whoes army is 400x its size, i dont care what random number generator you use.

 

[Crazy Jerome]

I expect the Civ4 AI to be a solid improvement on the Civ3 AI, but not enough to satisfy most Civ Fanatics. (And I'm not even considering those with such a totally misguided view of what is possible that they would never be satisfied.)

About a year after release, I also expect there will be a modded AI or three that is considerably tougher than the Civ4 initial release AI. I expect it to be tough enough to drive everyone down a difficulty level. From there, I doubt it will get much tougher, though it will probably get a lot more flexible (e.g. for any given task, picking from 2 or 3 solid strategies developed by veteran players). The best players will still whip it, but they'll have more fun doing it.

 

[Aussie_Lurker]

My big hope for the AI, aside from many of those listed, is for it to be a bit more BLIND . Namely, I want it to know as much-or little, to be more precise,-about the map and future resources as I do, and I want it to not even know which civ is controlled by the human, and which are controlled by the AI (if you think it can't be done, try SuperPower 1 and 2 on for size!)

Yours,
Aussie_Lurker.

 

[joethreeblah]

I expect the Civ4 AI to be a solid improvement on the Civ3 AI, but not enough to satisfy most Civ Fanatics. (And I'm not even considering those with such a totally misguided view of what is possible that they would never be satisfied.)

About a year after release, I also expect there will be a modded AI or three that is considerably tougher than the Civ4 initial release AI. I expect it to be tough enough to drive everyone down a difficulty level. From there, I doubt it will get much tougher, though it will probably get a lot more flexible (e.g. for any given task, picking from 2 or 3 solid strategies developed by veteran players). The best players will still whip it, but they'll have more fun doing it.

would it be possible for them to run a zillion games of AI's vs AI's and do like a survivial of the fittest AI program evolver

 

[Leuf]

Combat:

Where the AI is always going to have an advantage is to be able to look at all the odds of all possible moves it could make. The human player is generally not going to do this as thoroughly.

The human player is always going to have an advantage in strategy. It's very easy for the player to look at the map and see what is important. It's extremely difficult to get the AI to understand this.

I expect marginal improvement in this area. The AI will be very tough in combat the first couple times you play it, but you'll quickly learn to manipulate it again because the AI will be built on rules rather than understanding.

Everything Else:

I expect vast improvement.


-Leuf