Fixing AI should be Easy

[aieeegrunt]

This isn’t an all or nothing thing. It should not be hard to significantly improve an AI as bad as Civ6.

Just making the First District Priority in a city be harbor followed by commercial hub would help a lot

As far as making it militarily more competent, the only way to do that is to ditch 1 UPT

 

[frutiemax]

Anyway, without DLL it's impossible to fix it.

 

[civac]

As far as I understand it self-learning Go AIs have two stages.They have a neural network that evaluates candidate moves based entirely on pattern recognition and a calculation engine that use Monte-Carlo tree search to choose between the candidate moves. The first part requires very few computational resources. There are online go playing programs which use their neuronal network to essentially play instantly (definitely <0.2 sec per move). It's possible to take advantage of the fact that this type of program does almost no tactical calculations but its small-scale pattern recognition (shape in go terms) and large-scale pattern recognition (direction of play in go terms) are so good it can beat quite strong players if they do not target that weakness explicitly.

This kind of AI is probably not suitable for comprehensive Civ game play. The self-learning games to train the neural network are evaluated by playing them to the end and checking who wins. This is not feasible for a game of Civ. Moreover, the idea of playing games to their conclusion was developed for Go explicitly because evaluation of a Go positions (who is winning and by how much) was an unsolved problem. Older AI systems (chess engines) use some value which can be calculated out of the game state to evaluate positions. In chess the comparison of the material of both players is used (probably not just the material value but also the maneuverability and positioning of the pieces). For Civ games, it should similarly be possible to develop a fairly precise and easy to compute measure of the economic capacity and military capabilities of each faction. The Civ AI would then endeavor to increase its economic potential as much as possible which enables partitioning the overwhelming complexity of the entire optimization problem into smaller ones (at least on the time scale).

Go-style AI would be suitable to control the tactical combat in Civ 6 exclusively. The problem is that it would have to take reinforcements on both sides into account which muddies the waters a lot.

 

[frutiemax]

Applying machine learning on civ 6 seems fun but the real problem is that the AI can be a lot improved by some proper behavior tree/probability programming. The goal for an improved ai in civ 6 is to give a proper challenge for experienced users, but not necessarily always beat the pros. Perhaps on the civ 6 development team they lacked good players, hence they didn't know how to make a proper ai.

 

[DizzKneeLand33]

I use a mod that adds 1mp to all units and Gran Columbia AI is absolutely fantastic with that.

Wow, I think I'm going to have to try that. Of course the barbarians could be even scarier there too.

 

[Vargas1]

I wonder if Civ 6 being ported to the Nintendo Switch was the death knell for a strong AI in the franchise. I don't have a reference on hand, but I remember reading or hearing that one of the challenges that Firaxis has faced with the AI is late-game turn times. Players don't want to sit and wait for long AI turns, so the computational complexity that is required for good AI play has been sacrificed to keep turn times low. You can do both if you have enough processing power, but the Switch is not exactly a beast of a machine. If sales were strong on the Switch I don't think Firaxis would sacrifice that market for the sake of making a stronger AI.

I wouldn't be surprised by this - I play on the Switch, and the AI turns get pretty long in a standard game even now (I also have dabbled on a few large maps, but the time needed to actually complete a game is awful). I would imagine with more complex AI it would get fairly unplayable.

 

[frutiemax]

I wouldn't be surprised by this - I play on the Switch, and the AI turns get pretty long in a standard game even now (I also have dabbled on a few large maps, but the time needed to actually complete a game is awful). I would imagine with more complex AI it would get fairly unplayable.

I doubt it, on my computer, on standard map, a turn with 8 AIs takes less than 10 seconds. I wouldn't mind waiting twice the time (20 seconds) and let the AI makes more calculus.

 

[Aristos]

Perhaps on the civ 6 development team they lacked good players

Or good coders.

 

[PiR]

I doubt it, on my computer, on standard map, a turn with 8 AIs takes less than 10 seconds. I wouldn't mind waiting twice the time (20 seconds) and let the AI makes more calculus.

It can be much much worse depending on e.g. your device, the map size, and late-game.
Even some mods can heavily slow down the turns.
This aspect should not be underestimated.

 

[Kwami]

And also, 20 seconds times 250 turns is 83 minutes of just waiting around.

 

[Aristos]

And also, 20 seconds times 250 turns is 83 minutes of just waiting around.

You can pet the dog in the meanwhile.

Moderator Action: Please do not troll the forums in this way. leif

 

[ThunderLizard2]

There is also the problem that I would never play any game in which the AI would beat me every time. The AI can't be too good if we want the game to be enjoyable for most players, and most of players play only casually.

Have you heard of a difficulty slider? If you are getting beat then you should be playing Chieftain or Settler not asking that the AI be dumbed down for those who know how to play.

Civ6 AI is so terrible it does things like slam campuses to the point it runs out of money and has to disband it’s army

It would be hard to make an AI worse than this one

True this.

Big issue is Firaxis refuse to release dll code so modders could fix the AI like they did for Civ V with vox populi.

 

[Gedemon]

Have you heard of a difficulty slider? If you are getting beat then you should be playing Chieftain or Settler not asking that the AI be dumbed down for those who know how to play.

there are two issues with that from my point of view (supposing we could have an AI good enough to force the human player to use bonuses), both somehow related to immersion, so maybe not an issue from the point of view of a pure competitive player:

1. if the AI is invincible (as in Go or Chess) you may need to use bonuses in a way similar to what we have with Deity: you and the AI won't play the same game.
2. that AI will use every possible exploit to try to win the game, and that not something you want to play against, even if you have bonuses to counter them (on the pro side, such an AI would help the devs to identify possible exploits and fix MP)

So yes, IMO, that hypothetical AI would have to be "dumbed down" for the fun of at least a portion of the humans players (and it's already the case with the current AI, for some specific and potentially frustrating situations, at least in civ5 code)

 

[DizzKneeLand33]

So far the extra movement point is working out well in so many ways. Now to adjust that science obsession that they programmed in. Then if the AI could be programmed to conduct wars past 10BC, this game would be a lot better.

I sadly think the war logic is hard coded, but maybe there is a setting that can be tweaked. I have tried for years adjusting other numbers in diplomacy, but to no avail.

EDIT: Now that we have been told that the final patch has been issued (basically), I'm inspired again!

 

[Futumch]

Something I have seen a lot lately is the Ai losing its cities to revolt in the mid to late game. I won't notice it and suddenly I see 3 or more of their cities turn red and my trade routes get chopped to pieces. The revolting cities are not even close to my borders most of the time. The Ai shouldn't have cities revolt unless the player is within 10 hexes. And another thing....free city units shouldn't pillage trade routes.

They seem to settle out in the sticks, far away from their "homeland" (so to speak), then when a new era comes along - bang, city revolts. Don't even get me started on the very poor settlement decisions of Kupe.

 

[frutiemax]

Civ 5 ai is so much better than Civ 6, I just played a game and I had some zulu resistance, while playing vs the AI in civ 6 is often a city simulator...

 

[Fluphen Azine]

I don't think it has anything to do with generations. I've been playing computer games for 35 years and I've never enjoyed an "unbeatable" challenge. Perhaps the proportion of casual gamers within the gamer population is growing with the rising accessibility of gaming platforms (there are many more devices that allow gaming, per capita, now than in the past), but without research we can only speculate about it.

Just because you or I couldn't beat the game doesn't mean the game was unbeatable and I've been playing video games for 42 years now, starting with pong.

The truth is that AlphaGo and AlphaZero are real and you all are telling me that they can't fix the Civ 6 AI??

 

[Aristos]

The truth is that AlphaGo and AlphaZero are real and you all are telling me that they can't fix the Civ 6 AI??

Because they really CANNOT. The last very good Firaxis coder was Soren Johnson. He was also the one who had enough self esteem and reassurance about his own skills to come up with the idea of releasing the dll source code.

The Soren spirit (and skillset) is all but gone from Firaxis. And IMHO, for ever.

 

[Archon_Wing]

Just making the First District Priority in a city be harbor followed by commercial hub would help a lot

Why build both? Commercial Hub is pretty useless if there's already a lighthouse there. All that really does is make it easier to steal gold from them.

It's not like the AI lacks in gold either.

 

[Gedemon]

The truth is that AlphaGo and AlphaZero are real and you all are telling me that they can't fix the Civ 6 AI??

The truth is you can't "fix" Civ6 AI with those kind of AI, so it's completely unrelated.

Now, could civ6 AI be "fixed" ?

Depending of the definition of "fixing" that would take more or less time, if we had access to its code. But we don't.

Could it have been better ?

Yes, absolutely. And I don't think it's a lack of competence, but a lack of interest for Firaxis. Civ6 has sold millions of copies, Civ7 will do the same, even with a similar AI.