Does anyone else agree that the AI is the highest priority?

[urlm]

I couldn't agree more. I don't really play Civilization for the graphics, but purely for the gameplay, and when it's single player, the AI is what makes the experience.

And let's face it, the AI in Civilization V is terrible.

The only way they buff up the AI's with difficulty level is giving them unfair advantages such as more production, gold, etc.

The difference between an easy and a hard AI is that an easy AI throws 2 units at your city while a hard AI throws 20.

It's equally boring to see an AI suicide 2 units at your city as seeing an AI pop up 20 units out of nowhere while you've barely managed to create a farm and see them obliterate your city thanks to their ridiculous advantages on gold and production.

What I'd love to see is an AI that would progressively play smarter. I reckon it's not that difficult for an AI to mathematically calculate what building order and what strategy is the most profitable, sort of like how a human would improve in the game.

I'd like to see the AI's treat Civilization more like chess. Rather than giving the AI extra pieces or unfair rule advantages, make them smarter.

The AI in Civilization V is really disappointing, and made me grow tired of the game.

 

[urlm]

And I don't think that's a problem either, AIs aren't "opponents", AIs are puzzles to solve.

It's hardly a puzzle when you don't have to out-strategy the AI, but find some way to exploit their stupidity to the point where their insane advantages doesn't outweigh it.

The AI in Civ 5 is just terrible. There are chess AI's that manage to calculate the perfect move every single time, that are unbeatable. I know Civ5 has a lot more mechanics than chess, but it shouldn't be that difficult to make an AI that progressively knows what building order is mathematically best, and that just throwing units at a city to die isn't the way to go.

As it is now that's exactly what they're doing. Just throwing units at your city. It's boring.

The AI's in Civ V turned the game into a fancy tower defence game, especially at higher difficulties.

 

[poom3619]

I'd like it if there was an option to make the AI play to win.

What did you meant option? I totally understand why you want to make AI play to win, and pretty much agree with it. But I don't think making 2 sets of AI that changed by a tickbox is a good idea.

If possible, I wish the AI would be more cut-throat if the player play above certain level, where AI basically tell you that you shouldn't expect to have best friend forever on Emperor, and Bismarck teach new player what "Realpolitik" meant...

 

[noto2]

What did you meant option? I totally understand why you want to make AI play to win, and pretty much agree with it. But I don't think making 2 sets of AI that changed by a tickbox is a good idea.

If possible, I wish the AI would be more cut-throat if the player play above certain level, where AI basically tell you that you shouldn't expect to have best friend forever on Emperor, and Bismarck teach new player what "Realpolitik" meant...

I said that in response to what people were saying earlier in the thread. While I want an AI that plays to win, not everyone does. Some people have expressed that they genuinely enjoy an AI that role-plays. I'm not discounting their opinions. I see the enjoyment of role-playing AIs as well, and I would like it in another game, but in Civ I'd prefer that they did not role-play, and instead played like humans in MP.

Not everyone wants that, though. Firaxis designs their AIs with that in mind. So to please both kinds of players, I wish it was something that the player could choose in game settings.

 

[dexters]

I said that in response to what people were saying earlier in the thread. While I want an AI that plays to win, not everyone does. Some people have expressed that they genuinely enjoy an AI that role-plays. I'm not discounting their opinions. I see the enjoyment of role-playing AIs as well, and I would like it in another game, but in Civ I'd prefer that they did not role-play, and instead played like humans in MP.

Not everyone wants that, though. Firaxis designs their AIs with that in mind. So to please both kinds of players, I wish it was something that the player could choose in game settings.

Role-playing/play to win means different things to different people. So it's very difficult asking Firaxis to write two sets of code.

I think generally when some people say role-playing, they want an AI that reacts to them. Friendships formed are expected to be permanent, maybe a few backstabs allowed, but deterioration in friendships or inflection points are not desired.

But role-playing can also be understood to be an AI the role-plays their role as the leader of a Civ, pushes their agenda/self-interest forward and presents a diplomatic challenge to the human player.

So I rather they add more levers to diplomacy, add more transparency to relations and let role-players weave a tale to that.

I get the sense people don't mind an AI that plays to win when it comes to weaving their role-playing immersion, but there needs to be clarity for them as to who can be friends and who can't be ahead of time because they invest resources/turns/effort into a friendship. No surprises. Which hopefully the agenda/hidden agenda system will address. So I'd venture to guess we can find out things like what ideology the AI prefers fairly early on too.

 

[noto2]

Programming the AI to play its leader differently than another leader, when different civs have different abilities, is just common sense and fits either RP or PTW.

You brought up a good example, though, with friendships. If an AI is your "buddy" and stands by while you win a cultural or science victory even though they have the power to roflstomp you with their military, that is role-playing through and through. While some players might like that immersion, other players hate it. I am in the latter camp. I don't want the AI to just stand there and let me win.

Even in a MP game you will find players cooperating with each other a lot of the time - when it suits them. They will trade, they will sign research agreements, they will gang up on someone else, etc. You can be sure, though, than against a human player you will be backstabbed if it is in their interests to do so.

I want the AI to be like that.

A more ideal way of addressing this issue might be some kind of system in place that actually gives civs an incentive to stay peaceful with one another, and we do see that Firaxis has added pieces of such a system in place. If you have signed a research agreement that has not yet come to fruition, you lose the invested gold if you go to war with the other civ. You also lose things like trade routes. Perhaps they could further develop that theme.

Civ somewhat dabbles in this. In previous iterations you would get war weariness if you were democratic/representative, and you get economic bonuses in those governments. They could do more with this, though.

 

[Ryika]

It's hardly a puzzle when you don't have to out-strategy the AI, but find some way to exploit their stupidity to the point where their insane advantages doesn't outweigh it.

The AI in Civ 5 is just terrible. There are chess AI's that manage to calculate the perfect move every single time, that are unbeatable. I know Civ5 has a lot more mechanics than chess, but it shouldn't be that difficult to make an AI that progressively knows what building order is mathematically best, and that just throwing units at a city to die isn't the way to go.

As it is now that's exactly what they're doing. Just throwing units at your city. It's boring.

The AI's in Civ V turned the game into a fancy tower defence game, especially at higher difficulties.

Maybe you should have read more than just the last sentence of that post. I wasn't defending CiV's AI, I was merely saying that it's perfectly fine to help the AI with number bonuses to fill the remaining gap between player and AI, because the AI can never reach the level of the player.

If the initial gap is too big, then that's of course a problem.

There are chess AI's that manage to calculate the perfect move every single time, that are unbeatable. I know Civ5 has a lot more mechanics than chess, but

Chess is played on a static board, that's hardly a fair comparison. Try to create an AI for Chess960 and you'll see how randomly positioning a few pieces already makes things SO much more difficult.

It would however indeed be very easy to force the AI into the perfect build order once those build orders have been discovered, but that's hardly a good AI, it's just an AI that knows the good moves. If anything that's bad game design if you can do the same thing over and over again.

 

[unpossible251]

Chess is played on a static board, that's hardly a fair comparison. Try to create an AI for Chess960 and you'll see how randomly positioning a few pieces already makes things SO much more difficult.
.

Mmmmm yeah. Great AI is the aspiration. But it must be hard.

Years of research and thousands of programmers have finally managed to make competitive Chess AI. And chess has six types of moving pieces, on two types of squares.

Civ has one programming team, taking care of literally thousands of types of moving pieces.

Anyway, i'm a fan. Its not my job to offer excuses. Just, hope they can make something clever.

What I would like to see are advisors who can actually tell you something useful about rival civs. Rather than just "guarded" (ok, so the modifiers can help), an AI adviser who can give you a real qualitative judgement.

"ALEXANDER is generally an AGGRESSIVE civ. You've built wonders that he DOESNT covet, and made city state allies that he IS JEALOUS OF. He DOESNT care about your rapid expansion, but he HAS NOTED your warmongering. I think its LIKELY that he will declare war on you due to your CITY STATE RELATIONSHIPS"

That would actually impress me.

 

[Tabarnak]

the most important thing is features, mechanics and balance.

if the game doesn't have working features and mechanics, and isn't well balanced, then it doesn't matter if the AI is good or MP works.

I'm with MadDjinn here. With proper balance the AI will likely be more at par with humans. Too more broken things result into more AI raped.

 

[HorseshoeHermit]

The challenge of programming a glimmer of skill at the game is preceded by having any amount of skill at the game, something that no one with 2K does.

Playing to immersion and playing to win are completely at odds in the respect of AI design, which is sad. I personally don't know which one I like. If the AIs play to win then they're mostly all the same, and it also means that they will have to be a little bit unpredictable - so you have the Civ IV-V 'solution' tied up in this too. Is strategy supposed to be about a predictable playing field? Well, no, it shouldn't be when it comes to your opponents' actions.

Looking at the Civ V AIs as players, they never looked psychotic to me... actually they often seemed very cunning, trying to screw up my diplomatic efforts with their rivals. In my games they had a lack of fight, which made some wins disappointing, not the "lol backstab" situation.

I hear that in Deity Civ IV, the trouble is that the AIs making too much sense makes your diplomatic efforts just a set of 'tricks' or 'conditions' to clear, just a bunch of expenses you have that are mostly gifts, which get you insurance through the most devastating and vulnerable parts of your world domination scheme (which is what every game of CIV on a high difficulty feels like... a scheme and quick for world domination as you make sure to use the entire playbook to ensure relations with the foe) edit: World domination is a nice tune for CIV, a clear goal and everything in the game reinforces that play experience, but it does contribute to CIV's samey-same replayability issue (everything is cuirassier rush, etc.). Certainly by contrast, CiV's balancing create that unsettling feeling of playing for a game's end conditions rather than just evolving toward supremacy in some respect or other, and it's made more artificial with the circumstance that you have to specialize from turn 4. So I don't mean to say the world domination gameplay of CIV is bad, but the other players' interaction with it is. Only one victory keeps all goals clear and the game's "respiratory system" very elegant, but then there's a limit to how rich the rivals' behaviour can be even without PTW... A CIV ai that turtled is like a rival that's not even playing, for instance.

My preference then would seem to be one thing which I've read is others' nightmare or even no-buy scenario, which is AIs that have immersion, but without machinery behaviour. Maybe if just all of them were given a competitive spirit, just in the scope of the game's main success factors (like population size, or settlements). This is partly motivated by an acceptance that these computer players will never have a frigging clue how to play this title to win.

 

[dexters]

I think people are talking about two different AIs here.

1) Tactical AI is without a doubt in need of work
2) Diplo AI, there is a raging debate about a) how they should be orientated b) the UI & game systems that go in it

Firaxis has already hinted there will be a new mechanic and presumably new UI layer for diplomacy with the Agenda/hidden Agenda system.

On the latter, it's less clear what is in need of fixing as the complaints generally boils down to transparency (can be improved) ; immersion (subjective) and AI goal seeking (player dependent).

 

[Hail]

I'm with MadDjinn here. With proper balance the AI will likely be more at par with humans. Too more broken things result into more AI raped.

the AI needs to understand how to play the game.

given current limitations, the AI will not be on par on humans. nor does it need to be.
what the AI sorely needs is to completely avoid stupidity. for most, that wil be enough.

the AI should be a high priority, but it never will be, because it's expensive and unstable. visuals & audio give a magnitude higher bang for buck than AI. civ5 proves this once again.

 

[noto2]

@Horseshoe: yeah one of the things I hated about Civ4 was that it felt too "gamey" in diplomacy. I'd often do whatever it took to get my neighbours to "friendly" and I'd memorize which ones backstabbed at which relations level. I quickly modded the game to get rid of that aspect.

 

[Spartan_X]

I think the problem with AI in CIV V is that apparently it is pre-programmed to do certain things and build certain stuff, and whenever it cannot do what it has to do, either because the terrain doesn't allow it, or for whatever other reason, the AI appears to revert to stupid actions. I think there is one simple way to fix this in CIV VI, and that is, program the AI to behave differently according to its surroundings and player actions, give it a large set of instructions that could be "if *that*, then do *that* or *that* or *that*" etc... many different actions, chosen semi-randomly. In this way, it would appear to act randomly ( while it will not be random at all ) and much more logically than before. I guess this will be a strain for slower CPU's, as it will require more calculations from that part of the computer - but then again, Civ is a strategy dammit! If a strategy game doesn't put a strain on your CPU, then there is a problem with your game

 

[TheMeInTeam]

the most important thing is features, mechanics and balance.

if the game doesn't have working features and mechanics, and isn't well balanced, then it doesn't matter if the AI is good or MP works.

The most important thing is that the game functions at all, then what you say, then AI.

But working features and mechanics are mechanics are those that work in MP too. If features/mechanics are relying on AI stupidity just to be viable at all, they're not well-tuned.

The AI comes after MP-viable features and mechanics, because it's only at that point you can have an AI that acts plausibly within the constraints of the game rules without throwing or behaving in a vexing manner.

 

[stiiknafuulia]

The most important thing is that the game functions at all, then what you say, then AI.

But working features and mechanics are mechanics are those that work in MP too. If features/mechanics are relying on AI stupidity just to be viable at all, they're not well-tuned.

The AI comes after MP-viable features and mechanics, because it's only at that point you can have an AI that acts plausibly within the constraints of the game rules without throwing or behaving in a vexing manner.

Can't they do a 'learning' AI that watches top MP players struggle each other and 'memorizes' their tactics? Or is that too costly or complex to program for a computer game?

 

[Acken]

I think the problem with AI in CIV V is that apparently it is pre-programmed to do certain things and build certain stuff, and whenever it cannot do what it has to do, either because the terrain doesn't allow it, or for whatever other reason, the AI appears to revert to stupid actions. I think there is one simple way to fix this in CIV VI, and that is, program the AI to behave differently according to its surroundings and player actions, give it a large set of instructions that could be "if *that*, then do *that* or *that* or *that*" etc... many different actions, chosen semi-randomly. In this way, it would appear to act randomly ( while it will not be random at all ) and much more logically than before. I guess this will be a strain for slower CPU's, as it will require more calculations from that part of the computer - but then again, Civ is a strategy dammit! If a strategy game doesn't put a strain on your CPU, then there is a problem with your game

The ai is already programmed this way. Its full of randomness and situation reading. Its problem are due to some decisions taken by developers, general difficulty of ai programming especially especially with 1upt, and balance. Balance because the randomness implemented means it will take crappy options despite the fact it reacted logically to the flavor of that option. For example taking piety with a religious civ.

 

[TheMeInTeam]

Can't they do a 'learning' AI that watches top MP players struggle each other and 'memorizes' their tactics? Or is that too costly or complex to program for a computer game?

I'm not the right person to ask for programming specifics, at least not as I am now, but my understanding is that AI "learning" that would be onerous. Developers could have AI mirror strong play tactics via patches, but usually this opens new holes in it and is enough of a resource sink to not be worthwhile.

Basically, it should be taught how to move and shoot, but not necessarily that it should always go liberty 3 cities to rush you or something.

You brought up a good example, though, with friendships. If an AI is your "buddy" and stands by while you win a cultural or science victory even though they have the power to roflstomp you with their military, that is role-playing through and through. While some players might like that immersion, other players hate it. I am in the latter camp. I don't want the AI to just stand there and let me win.

Asking for "role play" in this scenario is asking for a different game. If you have Civ's game rules, at least so far, AI behaving as you describe is game throwing. It is possible to design the game's rules such that you have incentive for lasting, even permanent friendship (Civ IV had a shallow implementation of this, turned off by default, with permanent alliances).

When people start talking role play, they're talking about playing the game in a fashion where one of its victory conditions isn't the aim. If you want to do that, turn down the difficulty or accept the challenge. It's not reasonable to ask that the AI game throws in general to satisfy a player's desire to self-restrain on high difficulties, nor is it reasonable to ask that the AI to pretend as if extra rules exist when they aren't there.

That's why I said the mechanics/features need to exist first. If the developers want diplomacy as part of the game, then they should set their mechanics to incentivize/necessitate diplomacy to win in MP. Use whatever tools is necessary; overwhelming strength from collusion, joint victories, economic victory that you can't reach without cooperation but risk your ally out-managing you, whatever. If that's the route you want to go, to make diplomacy engaging, then rather than making the AI throw...make the mechanics work in the first place.

If the AI has to throw for the mechanic to be usable and it isn't viable in MP, the mechanic doesn't work.

Balance because the randomness implemented means it will take crappy options despite the fact it reacted logically to the flavor of that option. For example taking piety with a religious civ.

Piety could have been made MP-viable, but it wasn't.

 

[Acken]

When it comes to balance there are indeed a lot of "could have been"

My point was only to illustrate that the AI reacts logically to how it was designed. A religious civ will pick piety more often, as it should a priori. But because of balance issues it is a posteriori a poor choice.

 

[poom3619]

Can't they do a 'learning' AI that watches top MP players struggle each other and 'memorizes' their tactics? Or is that too costly or complex to program for a computer game?

There are too many random factors in Civ game for AI to simply memorize those tactics,. There's lot of difference in ideal tactic for Pangaea, Oasis and Archipelago map, and the same AI that work well against Civ5 ranged unit before artillery could do poorly against longbowmen (who have farther range)

and if they did. It could end up taking long time to process on player's machine.

In other hand, If there are a game that AI usually do well because of they use only one or two strategy that work better than all other alternative in first 100 turns. Then I would think the game. have another problems that didn't directly involve AI...