Likelihood we'll get a decent AI

[Leyrann]

If I have to keep reading 'ChatGPT AI' I might have to ask ChatGPT what would be the quickest way to jump off a bridge

Yeah, guess it's time for an obligatory reminder: ChatGPT can't even figure out what are and are not legal chess moves. It certainly cannot provide an AI for a game vastly more complex than chess.

 

[Krajzen]

I'm really curious about civ7 approach to diplomacy. Similarly to gov system struggles of flexibility vs consequences, civ games diplomacy struggles to balance human agency and AI agency.

In civ4 AI diplomacy was fairly transparent, numerical and rational, so it was easy to understand. But it was deemed to be too transparent and making AIs too predictable and easy to manipulate, so civ5 made its diplomacy much more opaque and volatile, and gave its AIs very complex personalities. So in this game AI leaders have a lot of will, agency and flavour... so much in fact, that it is really hard in this game for human player to influence their damn mind about anything, they are too wild.

Civ6 attempted to solve this dilemma by essentially giving its AI leaders autism (I hope its not offensive, I am on the spectrum myself ). AIs here are utterly obsessed with their binary fixations, but also very open and predictable about them, so they are wild but you can also tame them a bit. Success!

But then another set of problems has risen. Agendas don't make AIs rational political players but render them irrational, unstable and self-sabotaging. Agendas are also often frustrating to deal with and immersion breaking due to how clearly they are designed with gameplay in mind, not making in-universe sense.

Some time ago I have dug out this interview in which civ6 designers IIRC admitted they would like to see diplomacy being more about battle of wits between political players, and their dissatisfaction with civ6 diplo was visible, so we can assume some new angle here. Perhaps this time there will be effort to make it more about intrigue, cunning, self preservation, action and reaction? It would also facilitate espionage being integrated with diplomacy (and reworked yet again, which is obligatory at this point)

 

[GeneralZIft]

Civ6 attempted to solve this dilemma by essentially giving its AI leaders autism (I hope its not offensive, I am on the spectrum myself ). AIs here are utterly obsessed with their binary fixations, but also very open and predictable about them, so they are wild but you can also tame them a bit. Success!

Success?? Maybe I'm playing a different AI to everyone else...

 

[Krajzen]

Success?? Maybe I'm playing a different AI to everyone else...

(I mean personally I hate agenda system but I tried to give it credit at least on paper in theory)

 

[UWHabs]

(I mean personally I hate agenda system but I tried to give it credit at least on paper in theory)

I think the problem with the Agenda system is that while it added some predictable personality, it just always felt... random. As you said, not like civ 5 random where it felt random for random's sake and there was nothing you could do, but sometimes the civ 6 AI would just develop this weird random personality. Not even counting Nuclear Gandhi in there, but like a civ could get a random agenda that completely does not match their map environment just due to luck. "Doesn't like civs who compete with them in city-states" assigned to a civ that starts on an island without city-states.

It would probably be better if the agenda system maybe existed, but if AI characters could somehow develop their agendas naturally based on gameplay. So, like, while it's good that Harald maybe likes civs that build a strong navy, maybe you don't hard-code him to that agenda, but instead basically let that agenda develop naturally. Norway has a bonus with boats, and naturally in his play he will like to build boats. It feels like maybe once he has built up a big navy, the game engine would like roll a die, and then depending on the map, maybe he gets an agenda to respect other civs with a big navy, or maybe he gets an agenda to want to have the largest navy and want to attack anyone that challenges him on that. But maybe in his game and setup he doesn't end up in a place to actually build a big navy, and instead ends up with the agenda to have people who backstab others, or maybe ends up being pious and likes others who build lots of holy sites, etc...

Obviously it would be tough to figure out the "trigger" for each agenda, whether once an agenda is attached it's there for life, or they re-roll it every 50 turns, etc... Like the above with a navy, which agendas are like "I have lots of X, I want to be the only person with X" vs "I have lots of X, I like others who also like lots of X". But I think if they could come up with a setup like that, it might actually have a chance of being the best of both worlds. Plus, it might actually feel interesting if I send a lot of trade routes to my neighbour, and then shortly after that I get a message from them being like "you know what, I like you sending me trade routes. Keep doing that and we'll be great friends".

 

[Gorbles]

Basically, what I'm trying to say is that Civ6 is way behind in terms of AI compared to other games and other genres, and it's meaningless to just blame it on the fact that it's a strategy game.

I honestly don't play many 4x games (anymore) aside from Civ, but I just don't see this. Not "way" behind, anyway. And the things other strategy games do well is in turn sometimes also smoke and mirrors.

How easy is it to find AI in Paradox games doing a silly? Pretty easy I think! I never got into the Endless series but I'm pretty sure it's possible there too.

I also think VI (now) trounces V (now) if you don't consider mods, and I think a large part of this discussion revolves around a bunch of fanatics (pun - not offense - intended) who no longer play the vanilla game (if they ever did). But maybe I'm wrong there. I'm certainly not knocking mods. But I've seen at least one poster who seems to think VI still has its launch AI, and I think in that case people really have memory-holed how poor V was on launch.

VI had low-hanging fruit left for years (not just in AI terms either), and that stings. But it's a lot stronger than it was, even if you can still fool it in a bunch of the usual ways. But it's like saying I can "cheese" rushing (or more accurately base harass) in the vast majority of RTS games, because that's generally the main strat the AI there is weak to (that you 100% need to leverage on higher difficulties before they're able to snowball).

Agreed 100% on "convincing and entertaining" though!

 

[pokiehl]

I think in that case people really have memory-holed how poor V was on launch.

The Civ 5 AI to this day cannot move an archer and shoot with it on the same turn.

People have ridiculously rosetinted glasses about all prior civs. I’m sure we’ll be hearing about how much better Civ 6 is than Civ 7 for quite some time as well.

 

[GeneralZIft]

I honestly don't play many 4x games (anymore) aside from Civ, but I just don't see this. Not "way" behind, anyway. And the things other strategy games do well is in turn sometimes also smoke and mirrors.

How easy is it to find AI in Paradox games doing a silly? Pretty easy I think! I never got into the Endless series but I'm pretty sure it's possible there too.

I also think VI (now) trounces V (now) if you don't consider mods, and I think a large part of this discussion revolves around a bunch of fanatics (pun - not offense - intended) who no longer play the vanilla game (if they ever did). But maybe I'm wrong there. I'm certainly not knocking mods. But I've seen at least one poster who seems to think VI still has its launch AI, and I think in that case people really have memory-holed how poor V was on launch.

VI had low-hanging fruit left for years (not just in AI terms either), and that stings. But it's a lot stronger than it was, even if you can still fool it in a bunch of the usual ways. But it's like saying I can "cheese" rushing (or more accurately base harass) in the vast majority of RTS games, because that's generally the main strat the AI there is weak to (that you 100% need to leverage on higher difficulties before they're able to snowball).

Agreed 100% on "convincing and entertaining" though!

The Civ 5 AI to this day cannot move an archer and shoot with it on the same turn.

People have ridiculously rosetinted glasses about all prior civs. I’m sure we’ll be hearing about how much better Civ 6 is than Civ 7 for quite some time as well.

I respect both opinions here, and I'm almost 100% convinced that the Civ6 AI plays BETTER than the Civ5 AI. But it is not convincing, interesting, entertaining, or unpredictable in the slightest.

Only recently I've heard people actually COMPLAIN about unpredictability with Civ5 AI.
I could understand the reasoning but it doesn't need to go the whole way to make the AI totally predictable.

At least I was surprised when a neighbour secretly built an army and decided to be expansionist.

There's nothing to surprise you in 6, I know some people like that, but surprises keep me coming back...

 

[pokiehl]

But it is not convincing, interesting, entertaining, or unpredictable in the slightest.

Only recently I've heard people actually COMPLAIN about unpredictability with Civ5 AI.
I could understand the reasoning but it doesn't need to go the whole way to make the AI totally predictable.

At least I was surprised when a neighbour secretly built an army and decided to be expansionist.

There's nothing to surprise you in 6, I know some people like that, but surprises keep me coming back...

Of course. I don't disagree with you at all that there are a lot of things Civ 5 AI did better. I think the "Flavor scores" did way more to define interesting, consistent AI personalities than Civ 6 agendas ever did, for instance.

I'm just saying that the "new bad, old good" mentality that people have for Civ games can be shortsighted. Reading some posts here, you'd think Civ 4 or Civ 5 graphics were hand-drawn by a reanimated zombie Van Gogh and that every aspect of gameplay and AI was perfectly, flawlessly designed. The reality is more nuanced: all of these games have their good and bad parts.

 

[UWHabs]

I respect both opinions here, and I'm almost 100% convinced that the Civ6 AI plays BETTER than the Civ5 AI. But it is not convincing, interesting, entertaining, or unpredictable in the slightest.

Only recently I've heard people actually COMPLAIN about unpredictability with Civ5 AI.
I could understand the reasoning but it doesn't need to go the whole way to make the AI totally predictable.

At least I was surprised when a neighbour secretly built an army and decided to be expansionist.

There's nothing to surprise you in 6, I know some people like that, but surprises keep me coming back...

Also it is important to consider what I would call the macro AI vs the micro AI. Macro AI being stuff like building up an army, declaring war on a neighbour, competing towards a victory, and stuff like getting the proper balance of expansion and turtling.

Micro AI I treat more as "do they move their unescorted settler next to a barbarian", can they take a city with a reasonable amount of units, do they needlessly shuffle units around when they would be 100% better to park and shoot, etc...

VI in micro AI still has some bugs - I definitely see cities just not take shots at me when I have troops around. Not always, but occasionally. I don't think they ever fixed them to be able to actually use cultists. They were certainly a lot worse at all that micro stuff a few years back, when they would constantly wander their settlers in the wilderness, shuffle units around constantly, and were completely inept. I've definitely seen them take troops in and actually put up a fight. And there's definitely some times when I have my troops around their lands and they can pick off my units here or there in what I would call a coordinated effort.

As for the macro piece, that's harder to judge. Partly because anytime they fail at the micro, that can also cause a cascade of effects. Like maybe they have the perfect idea, but if their settler gets eaten surely that changes what they can do. Or if they have a great idea to invade their weaker neighbour, but then fail at actually capturing the cities, the idea was right but the execution failed.

 

[Vincent_lacruz]

Also it is important to consider what I would call the macro AI vs the micro AI. Macro AI being stuff like building up an army, declaring war on a neighbour, competing towards a victory, and stuff like getting the proper balance of expansion and turtling.

Micro AI I treat more as "do they move their unescorted settler next to a barbarian", can they take a city with a reasonable amount of units, do they needlessly shuffle units around when they would be 100% better to park and shoot, etc...

VI in micro AI still has some bugs - I definitely see cities just not take shots at me when I have troops around. Not always, but occasionally. I don't think they ever fixed them to be able to actually use cultists. They were certainly a lot worse at all that micro stuff a few years back, when they would constantly wander their settlers in the wilderness, shuffle units around constantly, and were completely inept. I've definitely seen them take troops in and actually put up a fight. And there's definitely some times when I have my troops around their lands and they can pick off my units here or there in what I would call a coordinated effort.

As for the macro piece, that's harder to judge. Partly because anytime they fail at the micro, that can also cause a cascade of effects. Like maybe they have the perfect idea, but if their settler gets eaten surely that changes what they can do. Or if they have a great idea to invade their weaker neighbour, but then fail at actually capturing the cities, the idea was right but the execution failed.

Yes, what we need is ont ONE AI, but several specialized AI that can work in tandem. For example a city AI that aims to maximize the city development, with as input a general target (science, culture, or just production), with optimal placements (no longer campus with one science adjency when it could produce 4 one tile away...). And another one for optimal unit movement. And a "global AI" that set the priority for the sub-AI, depending on the victory type it find the best candidate...

 

[BackseatTyrant]

Of course. I don't disagree with you at all that there are a lot of things Civ 5 AI did better. I think the "Flavor scores" did way more to define interesting, consistent AI personalities than Civ 6 agendas ever did, for instance.

I'm just saying that the "new bad, old good" mentality that people have for Civ games can be shortsighted. Reading some posts here, you'd think Civ 4 or Civ 5 graphics were hand-drawn by a reanimated zombie Van Gogh and that every aspect of gameplay and AI was perfectly, flawlessly designed. The reality is more nuanced: all of these games have their good and bad parts.

I honestly feel it's less about "new thing bad" and more that brand loyalty can create over-entitled customers that easily feel betrayed if they aren't perfectly catered to at all times. I feel like what was going on with a lot of these crowds in question, is that they thought their favourite installation still had some fatal flaws, but not only did Firaxis not address a single of those flaws for the next installation, the devs removed some key features that made those love the previous installation so much, replacing those with features they thought only made the game worse. This is a narrative that repeats itself in pretty much every single long-running game series, and I think this is where a lot of the Toxic Gamer Mindset has its seeds planted.

If you feel in any way called out by the paragraph above, let me tell you: Firaxis is not a human being who can betray you for reasons individuals would betray each other over, it's a subsidiary of a subsidiary of a multi-billion dollar corporation, built on the backs of workers who make way less money than you'd expect for an industry of this scale, skill & prestige. Any game of any series, that you had looked forward to for a long time but ended up being disappointed at once you got your hands on it, must've been truly agonizing to actually work on (not to mention the amount of times developers must've wailed in despair at how the marketing team spun the upcoming product; now you gotta worry about rent AND possibly disappointing millions of fans)

 

[agonistes]

How likely do people think it is that Firaxis delivers a decent AI. In the age of Chat GPT, is it too much to expect? And if no decent AI is provided, will they at least open up dll so we can get modders to fix the AI like in CIv V?

Close to zero.

Most companies aren't investing in decent ai. I think there is one grand stategy game out there that consistently beats the players (i forget the name, meinteam knows it).

Maybe when ai can create game ai on the cheap, then we will get decent ai, Right now, there is a surge in players who care more about good graphics and being able to win easily than they do about good ai, and that is who games are being designed for.

 

[Leyrann]

Right now, there is a surge in players who care more about good graphics and being able to win easily than they do about good ai, and that is who games are being designed for.

2007 called, they want their complaints about game development back.

 

[Glassmage]

The Civ 5 AI to this day cannot move an archer and shoot with it on the same turn.

People have ridiculously rosetinted glasses about all prior civs. I’m sure we’ll be hearing about how much better Civ 6 is than Civ 7 for quite some time as well.

Civ5 Vox Populi literally fixed this. Civ6 World Congress is crap and will never get fixed because they won't release DLL

 

[pokiehl]

Civ5 Vox Populi literally fixed this.

You’re talking about a mod and I’m talking about the game itself.

Civ6 World Congress is crap and will never get fixed because they won't release DLL

What does bad AI have to do with the Civ 6 World Congress? The Civ 6 World Congress is bad and annoying chiefly because it’s so random. Nothing to do with bad AI at all. Sounds like you haven’t played it?

 

[GeneralZIft]

What does bad AI have to do with the Civ 6 World Congress? The Civ 6 World Congress is bad and annoying chiefly because it’s so random. Nothing to do with bad AI at all. Sounds like you haven’t played it?

I think he's talking about how easy to game the AI in the World Congress. My guess

 

[pokiehl]

I think he's talking about how easy to game the AI in the World Congress. My guess

I don’t really agree with that. The AI in the World Congress isn’t particularly bad—each entity just votes to its own benefit. Obviously you lose the nuance of the Civ 5 World Congress system of being able to “buy votes” from others.

 

[Leyrann]

I think he's talking about how easy to game the AI in the World Congress. My guess

You can somewhat game it, but that doesn't necessarily mean you can always get your desired outcome. Only that you can predict which way the AI will vote.