Is AI still brain dead?

[sonicmyst]

I don't know if you can call me a casual gamer, but I still cannot win an Immortal or Deity game and I find those levels stressful. I usually play on Emperor or King.

The only issue for me about this game is how the AI handles some of its combat, as well as them trading free cities when they are losing. Other than that, I am still enjoying the game (do I qualify as an apologist? I don't know)

 

[CivLuvah]

Agreed although I in no way apologise, as far as I am concerned it has never been a chess game.

Apologists are people who defend for something using arguments. Apologies might be part of such defenses but they are not necessarily the main focus of apologists' arguments, whatever the arguments might be.

 

[ggalindo001]

As I have mentioned in another thread -- I will be lucky in my lifetime if a complex game (such as Civ) develops an AI that remotely approaches a human in terms of capabilities. So when I think of the AI, I am not expecting playing against a human being, HOWEVER, I am wanting an AI that isn't "brain dead".

Some of the quicker fixes to the AI which would go a long way to enjoyment.

1. I need to feel vulnerable (esp. in the higher levels). At the moment, even and especially at the higher levels, once I develop a cohesive army that starts to conquer, it will not be stopped. My full army could be half a world away, and I am not concerned about counter attacks from even the most militaristic of Civs near my homeland. It is as tho the AI looks at my army strength rating and without knowledge of -where- it is located, it sits tight.

Fixing this item alone would provide significant challenge. And back to the poster that discussed the fear of a SOD showing up -- a coordinated attack when I have very little back in the mother country to defend, at worst slows down or eliminates the snowball, and at best, changes the balance of power.

2. Deal related AI -- and esp. the capitulation of the AI when winning a war, has to be fixed. I took Victoria's advice and actually went down to Emperor level -- and actually found the game to be fun. Rome DOWed me, and even tho I had defensive units, they were able to take one of my cities quickly with a coordinated attack using superior ranged units -- and was marching on my capital. As I was preparing my last ditch effort to defend my capital, they suddenly asked for peace. And in the peace deal, in exchange for gold, I got my captured city back!

I have argued elsewhere that the warmonger penalty is so steep and so quick that it really hampers the AI's ability to effectively wage a war to really create a balance of power change.

3. Diplomacy -- not sure that going from "lifelong friends" to denounced on a dime feels immersive -- and really penalizes the player in trying to play a balanced game. But if you are going to denounce, you really should be willing to go to war over it....at the moment, denounciations are simply ways to make me click thru screens and not a true game mechanic.

I'm optimistic about R&F, and again, not looking for incredibly tactical AI -- but the 3 items above feel like they could be easier "quicker" fixes. This, along with the AI using Air power, would go a long way to moving the AI from "brain dead" to very reasonable. And like the other posters have said, the AI has become a lot better from initial release -- and when focused, can provide a decent challenge into the industrial eras.

 

[ThunderLizard2]

I will be lucky in my lifetime if a complex game (such as Civ) develops an AI that remotely approaches a human in terms of capabilities

I'd suggest 2K stop adding overly complex features so they can publish a game that the AI can reasonably handle. It seems now they add a bunch of interesting features without considering how the AI can handle them - and the one AI guy has to figure it out after the game's been designed. A game that takes 8 to 30 hours to play is not a great title for MP (getting 6 people synced up for 8 hours???) so you need to have a decent SP which means decent AI. I've tried Civ5 MP and it was horrible with frequent de-syncs and constant quitters and my understanding is Civ6 MP is not much better. So that leaves the AI. If the AI is not good, then it seems the game devolves into a glorified SimCity type of game where you can build your empire as you like - but that's not what the Civ titles have been about at least not until Civ5 and the 1UPT mess.

 

[drubell]

The Civ6 AI still doesn't know the meaning of "retreat." The AI will predictably and consistently send an army that has a 0% chance of taking a city and still suicide the units into walls.

Even Civ5 AI knew how to retreat its units a year after launch.

 

[Karpius]

I'd suggest 2K stop adding overly complex features so they can publish a game that the AI can reasonably handle. It seems now they add a bunch of interesting features without considering how the AI can handle them - and the one AI guy has to figure it out after the game's been designed.

I am sure 2K will stop adding such complexity when people stop wanting it. But you just have to read through the various threads here to see that most Civ players can't wait for new features to be added. There are numerous 'wishlist' threads where people express their ardent desire to have their favorite Leader/Civ/UniqueUnit added. So long as the market is there, I doubt 2K and Firaxis will worry much about the AI.

I might offer one suggestion, however, for a very good military game. Pick any level, any map size, but choose to play against only one other Civ, with no City States. I have only tried this once myself, so the sampling is too small to declare it always provides the desired results, but in the one game I did, It turned into quite the military contest.

I played England against Rome on a TSL map with all victory conditions checked, on King level. Perhaps it was because
the AI had nothing else to focus on, but those Legions are nasty when they have nothing better to do. It was the first time I made a serious attempt at a religious victory, but failed because Rome and England were at war most of the time. Rome/AI actually placed some key forts, kept the Mediterranean closed and developed a navy to rival England's. I, playing England, quickly settled the Americas, but it was real bloody business trying to break into Europe.

My concentration on religion right from the outset, I believe, put me at a slight production disadvantage that I never quite recovered from. By the time Rome launched its satellite, I was too far gone from any other victory aside from Domination. Unfortunately, the cities with the spaceports were too deep into Roman territory to reach in time. I believe the game ended around turn 220. For me, thats a really fast game, but it was fascinating. I intend on trying other one-on-one match ups in the future.

 

[DWilson]

Also, turning barbs off boosts the AI, because the barb AI is so much better than the normal AI. Subsequently, the AI get set back worse than you do when barbs are active.

 

[Japper007]

Ugh again people saying the game should be dumbed down so the AI can "handle" it. What is the point of a sequel if not to add new content? We might as well keep playing civ II then, the AI can sure handle that game! ( By the way it couldn't... Civ 2 was just mindless ICSing all the time and the AI still couldn't be competitive without cheats)

 

[ThunderLizard2]

What is the point of a sequel if not to add new content?

What's the point of playing a game against the equivalent of a 5-year old? Doesn't seem very interesting.

 

[Japper007]

What's the point of playing a game against the equivalent of a 5-year old? Doesn't seem very interesting.

So go play multiplayer, it's what I did, and what I've been doing since civ 4.(4 and 5 AI really wasn't much above playing a "5 year old", even on Deity) Sure ingame lobby is atrocious but installing the PYDT client takes 5 minutes.

 

[Gorbles]

We've gone two pages and I don't think anyone has pointed out that realistic AI is something some computer scientists have devoted their lives and careers too, and it still isn't something that we could reliably call "solved". While game-related AI theory is obviously a bit of a different kettle of fish, I think people in general vastly underestimate the difficulty in reaching some kind of competence with these kinds of things.

It's a difficult thing to create. And like everything else that goes into a video game, it's subject to resource constraints and how important it is relative to literally everything else going in the game. Would you prefer better AI at the cost of less bugs being covered? What about worse map generation? Less mechanical improvements being made? As this is something of a specialist subject, I'm trying to pick features that would directly and immediately suffer if the people responsible were tasked with focusing on AI above said things.

And before folks start with "lol the game already has bugs", that's not a solid argument for introducing more.

 

[ThunderLizard2]

So go play multiplayer, it's what I did, and what I've been doing since civ 4

Civ4 multiplayer wasn't bad except for quitters. Civ5 multiplayer was unplayable even on a LAN - always de-synced before a game ever finished and 2K/Firaxis never bother too fix it. Heard Civ6 still has de-sync and other issues

Would you prefer better AI at the cost of less bugs being covered?

I'd prefer a decent AI to new fluffy features. Maybe they need to have more than one "AI guy"

 

[pgm123]

The Civ6 AI still doesn't know the meaning of "retreat." The AI will predictably and consistently send an army that has a 0% chance of taking a city and still suicide the units into walls.

Even Civ5 AI knew how to retreat its units a year after launch.

Weirdly, I have the opposite problem where the AI will send a lot of force against me that would pose a challenge and I kill one unit, so it retreats (mild exaggeration for effect).

 

[Zuizgond]

I'd prefer a decent AI to new fluffy features. Maybe they need to have more than one "AI guy"

Fully agree. I can't believe they only have one guy working on this... If only they put all the budget of the useless animations and voice acting (that I turned off on day 1) into AI improvement.

 

[Gorbles]

I'd prefer a decent AI to new fluffy features. Maybe they need to have more than one "AI guy"

And I'd prefer new fluffy features. Neither of us are wrong, it's just personal preference.

And preference extrapolated out to the general userbase is what drives expectations, which is how developers decide on prioritisation of such. So, logically, the desire for such is obviously weighted against prioritising AI. Nevermind the issues in believing hiring more people magically fixes a problem. It can improve a situation, but it's not as simple as "hiring more than one person".

 

[salty mud]

@Gorbles

what's the point in fluffy features if they just don't work properly?

 

[Krajzen]

We've gone two pages and I don't think anyone has pointed out that realistic AI is something some computer scientists have devoted their lives and careers too, and it still isn't something that we could reliably call "solved".

No, just no.

Civ5 got FANMADE community patch mod which made ai really good when compared with base game. I actually played with it and its ai was actually okay and actually much better than vanilla civ5 ai. It was not humanlike but nobody here wants genius or humanlike ai, just ai which doesn't look like a complete immersion-breaking idiot.

It is possible. It can be done. It is just Firaxis incompetence or unwillingness.

Civ6 didnt get such mod yet (due to lack of released dll code) but it already got simple basic ai upgrade mod (which just tweaks values without actual source code work) which is, indeed, markedly better than civ6 ai.

No, just no. We don't need neural networks and turing machines for civ 1upt ai to not be depressingly moronic.

 

[bbbt]

@Gorbles

what's the point in fluffy features if they just don't work properly?

If 'fluffy features' is referring to game mechanics, they work fine (balance issues non-withstanding). The AI can't work properly with them, which is a different thing. I'm not a fan of changing game mechanics based on the AI's ability to use them. And from Firaxis's perspective: From some (obviously very rough) extrapolation on the achievement percents: It looks like about twice as many players play on prince or lower than king or higher. Also significantly more players have played mutiplayer than king or higher.

However, I think initially 'fluffy features' was referring to things like the leader animation and the like. As someone who disables leader animations pretty quickly, I'm definitely in the theoretical 'AI engineer over more graphical enhancements' camp. Unfortunately, from a marketing perspective, 'we have a great AI' doesn't really sell copies. And frankly, I think they figure a number of people complaining about the AI will complain about the AI regardless of how much resources they throw at it.

On an optimistic note, I'd say the general gamer consciousness of 'AI' has grown, making it increasingly more of a selling/sore point for the broader audience - they might not want a fully challenging AI, but they want AI that at least looks like it's putting up a fight when they beat it

 

[ThunderLizard2]

Civ5 got FANMADE community patch mod

I played the Delnar AI mod in Civ5 and it was an improvement but because of Civ5 design (1 UPT) there were still issues.

Civ6 didnt get such mod yet (due to lack of released dll code) but it already got simple basic ai upgrade mod (which just tweaks values without actual source code work) which is, indeed, markedly better than civ6 ai.

It's even more absurd that 2K/Firaxis won't release the dll code so at least the modder community can try and fix their mess. It's ridiculous that the mod community has to spend time making the game work. Modders should be spending their efforts on adding new and interesting content not fixing 2K/Firaxis's poorly written code.

I'd say the general gamer consciousness of 'AI' has grown

Yes let's hope so. Maybe if the community refused to buy more DLCs until the game was fixed 2K/Firaxis would take notice. Sadly, until that happens change is unlikely.

 

[liv]

Someone wanted to make a petition for more DCLs after the expansion. So yeah do not count on a buyers boycott