Whatever they do I hope the AI is substantially upgraded

Wrong. It matters. I did not denounce the accuracy of their statement, but the complete nonsensical nature of it. You have the "right" to not see it, but to me it's pretty clear. Watch the video again, listen to the wording, and if you really want to go into details, watch the body language. I can clearly see them saying exactly that: "We purposely made a mediocre AI because the air units were too powerful in civ 5, and we did not want them to be dominant in civ 6". How is that not complete nonsense to you??? :crazyeye:
Your subjective reading of somebody elses' body language through an Internet stream isn't conclusive, sorry. You twisting their words also doesn't make me think you actually want a discussion on this. I mean, you're very clear on telling me I didn't pay enough attention, but then you come out with the humdinger of "we purposely made a mediocre AI" as some kind of quote by Firaxis. No thanks!
 
Your subjective reading of somebody elses' body language through an Internet stream isn't conclusive, sorry. You twisting their words also doesn't make me think you actually want a discussion on this. I mean, you're very clear on telling me I didn't pay enough attention, but then you come out with the humdinger of "we purposely made a mediocre AI" as some kind of quote by Firaxis. No thanks!

Freedom of thought. You are welcome to your own opinion. So am I.
 
Now, I am curious what an AI created by feeding civfanatics games into a neural net would be like. I bet it would create a pretty good AI. The problem would not be a lack of game logs clearly. When I said it would not be cost effective, the issue would be hiring someone or taking someone else off task, just to do machine learning to improve the AI when the benefits would probably not be worth it.

I think something of the sort could be used to help guide specific aspects of the AI (like tactical combat). But from my understanding it would be of limited use for the 'overall' AI as the AI needs to remain somewhat 'high-level' (weighted choices and the like) in order to function with mods (rule changing ones). The more tightly the AI has specific steps/rules, the more it will fall apart with a simple rule change.

I definitely agree it's primarily a resourcing/priority issue (there's a lot they could do to improve the AI without cutting edge machine learning techniques!)

Wrong. It matters. I did not denounce the accuracy of their statement, but the complete nonsensical nature of it. You have the "right" to not see it, but to me it's pretty clear. Watch the video again, listen to the wording, and if you really want to go into details, watch the body language. I can clearly see them saying exactly that: "We purposely made a mediocre AI because the air units were too powerful in civ 5, and we did not want them to be dominant in civ 6". How is that not complete nonsense to you??? :crazyeye:

They weren't talking about the AI strength - they were talking about the literal strength of the air units - they've buffed them by like an additional 20 strength in Gathering Storm. I.e. the fact that human players barely build them (and don't considers them a good 'value proposition') in 6 outside of 'roleplaying', versus them being much more vauled in 5. Granted, most people think the issue is the aerodrome requirement rather than their strength.

Frankly, I'm not sure anyone on the development team can speak to the AI 'design decisions' outside of the one credited AI developer (which neither of them were). Maybe I'm wrong, but that seems to be one of the major issues - the design process seems to be 'here's a neat idea, would it be fun and how could we implement it' and then later "okay it's implemented, now hand it off to the AI dev to deal with" - rather than "here's an idea - will it be fun? How will the AI be able to use it? etc." right at the point of design.
 
Honestly, that statement yesterday sounded, to me, like a lame excuse for the mediocre Air AI. They think we are all stupid, and would believe such a nonsense. "Yeah, we downplayed the air forces on purpose, because they were overpowered in civ 5" blablabla... complete nonsense!
I think I know where real problem is. When you make air fighter class defensively attack ground units it becomes virtually impossible for AI to wage offensive wars and capture cities. The late game overall balance is so off mark they need to literally gimp certain combat areas for this game even remotely look balanced. But it is not. And it is very dissapointing, because when you fix one problem it creates another one and so on, and so on...
 
Last edited:
Cooling off period. Could have used that in 1914.

Nothing would have stopped the European Civil War that started in 1914 (and ended, apparently temporarily, in 1945). The intricacies of the alliance system that was created, ironically, after the Napoleonic Wars to ensure that a continental conflict would never happen again, was the main culprit, not the assassination.

I think I know where real problem is. When you make air fighter class defensively attack ground units it becomes virtually impossible for AI to wage offensive wars and capture cities. The late game overall balance is so off mark they need to literally gimp certain combat areas for this game even remotely look balanced. But it is not. And it is very dissapointing, because when you fix one problem it creates another one and so on, and so on...

I smell the Behavior Trees, once again, to blame for this fact you are pointing out.
 
Well, you won't be buying it then.

What do you mean there is nothing you can do to stop civs who attack city states? Why can't you march your own troops up there to defend them?
Because that means declaring war. You could be allies with someone and not even that option! Another thing that needs to change. There isn't an option in the diplomacy to ask an AI to stop attacking a city state. Why not?
 
Because that means declaring war. You could be allies with someone and not even that option! Another thing that needs to change. There isn't an option in the diplomacy to ask an AI to stop attacking a city state. Why not?

Because it's unfinished? Seriously.
 
Has there been any word from Firaxis yet on whether the AI in Gathering Storm has had a substantial upgrade?
 
Has there been any word from Firaxis yet on whether the AI in Gathering Storm has had a substantial upgrade?

As of the build used in the Sweden livestream, there's no immediately obvious improvement in the AI's ability to defend itself or attack. No idea as to whether there's been an improvement in the AI's ability to pursue a victory condition, or if interactions with the AI have become more interesting or nuanced (although there is a new grievance system which may result in more nuanced diplomacy). The new world congress seems to be set up to avoid any semblance of negotiation with other leaders, so presumably that's a recognition that the AI couldn't do that effectively so better to let each leader cast their votes in isolation.
 
Some expectations are way too high here and tbh delusional. CIV AI is already pretty good. Look at other strategy games where the AI is either totally failing or barely able to play the game. Like PDX games HOI4 and Stellaris are a mess (even if I love them) and Total War also has big issues. CIV has always been very playable. And of course with more options and higher complexity the differences between human and AI are bigger. That can't be an argument against making the game better and releasing more features and content. Deity players who even want a bigger challenge have to put added artificial restrictions in their games, play mods or do multiplayer. We are only a tiny fraction of the playerbase anyway.
There are no strategy games out there of CIV or higher complexity with an AI that can keep up. And it won't change anytime soon. It might be achievable but you would to have to put dozens of millions of dollars, thousands of manhours and hire additional programmers of a calibre that often work in other fields not in gaming. Nobody is going to do that for any strategy game so you get the basic scripted "AI" which can't really think very well over extended number of turns its almost always only in the moment.
 
Why would they upgrade the AI? They can make money just by putting out new civs.
Take a look at how long the threads are on this site for each new civ in GS. Sure, people might be miffed about the AI, but not enough to refuse to buy the product.
 
As of the build used in the Sweden livestream, there's no immediately obvious improvement in the AI's ability to defend itself or attack.

The Norway AI actually built a sizable navy! A naval battle that turn several turns to play out actually happened! Carl lost a few ships! I honestly can't remember that ever happening in any of my games. Sure the AI still lacked any sort of strategy and just kind of threw units in there one at a time but at least it was something.
 
I was impressed with Norway's navy, I've never lost a naval unit to the AI in a naval battle in all the tine I have played Civ VI.

Though Norway didn't defend well on land. But I suppose the game wasn't on the highest difficulty.
 
Dream on!
 
was impressed with Norway's navy, I've never lost a naval unit to the AI in a naval battle in all the tine I have played Civ VI.

Though Norway didn't defend well on land. But I suppose the game wasn't on the highest difficulty

They do get double production on caravels, so makes sense they had a lot of them. I do like that they went fairly aggressively after their target. Makes me wonder if the land tiles are what's causing the AI so many issues. Being unimpeded on the sea may actually help the AI.

I'm beginning to wonder if the AI can't get any better. There are too many complex mechanics the AI can't handle.
 
Back
Top Bottom