AI incompetence deliberate?

[Abraxis]

I'm obviously discussing

obviously not

 

[Archon_Wing]

obviously not

Alright, why don't you try to explain what I'm discussing instead of wasting time with half baked quotes and attempted witticisms?

 

[Abraxis]

I already did, and obviously it was not so obvious. Obviously if this were obvious, there'd be no need to discuss it.

 

[Archon_Wing]

So pray tell, are you trying to argue that I shouldn't be criticizing the AI because said task is hard? Or are you trying to suggest that Civ's AI is good at what it does?

 

[Abraxis]

Me? No, I'm doing the same thing as you, mindlessly deconstructing something for no particular reason. Isn't that obvious?

Is it a coincidence that discussion is falling apart around us? I think the answer to that should be obvious.

Where is the constructive feedback? There isn't any because the solution is beyond the scope of even professionals, and acting like they are at fault is why I pick on you with a certain special derision I do my best to keep separate from my vocabulary.

Oh, so you edited your original post since this...

And you quoted the un-edited piece you were responding to so what's the problem?
I edit everything in the endless pursuit of clarity, until I get distracted by another thing. Maybe you should to, instead of just pinning the status of "obvious" to your statements like a royal seal, as though it carries any authority in lieu of actually saying something of substance or trying to get through to people.
-Contrary to your obvious implication, this does not at all change the message, or the flow of the discussion until you make it the subject. So one wonders, why bring it up?
More mindless deconstruction, this time nested here to die in the crib I hope.

 

[Archon_Wing]

Oh, so you edited your original post since this...

Well then, I don't think me or most casuals even need a so called "learning" AI that would play perfectly and beat pros. Not that was ever an expectation in the first place. We're kinda talking about the AI not escorting its settlers.

 

[Josephias]

As some others have commented it seems it is not a failure in the AI not knowing that a settler has to be escorted, but knowing what to do with plans changes. I've normally seen escorted settlers in open land, and even going unescorted could be logical in some situations (forward settling is a "race" game, and sometimes you need to take the risk to not escort your settler, if you know the zone is "safe").

Problem is, when AI loses this "race" (either you build a city first in the spot), or you just block the targetted spot with a unit, it does not know well what to do, therefore the settler remains "roaming" around as a 0-value asset (which, obviously, he is not). There should be a routine to "rebase" settlers that have lost its target, and reuse them if a new opportunity arises, that i think it is what is missing now.

It must be taken into account that, nevertheless, the AI takes a lot of risk with forward settling. One of my games last week, playing Egypt, I saw Brazil lossing one his escorted settlers to barbarians when trying to settle in my island/continent (yup, I got one for myself only in that game). The scort spearman was not able to survive continued attacks by barbarian cavalry (while my war chariot and spearman were just blocking "good settle spots" from them and watching they being attached). Finally, I was able to rescue the settlers from the heathen, and they founded Amarna instead of Sao Paoulo ... Later in the game, for sumeria forward settler, I was not so lucky, and I had to declare war to kill the escort and "convince" again the settlers they would do better if the city they founded was Egyptian .

 

[Archon_Wing]

Maybe you should to, instead of just pinning the status of "obvious" to your statements like a royal seal, as though it carries any authority in lieu of actually saying something of substance or trying to get through to people.

??? Royal seal? What are you even talking about? Authority on what? I'm not on authority on my own thoughts? lol

See, we were talking about the status of the AI, and I mentioned that the AI's perceived poorness is not unique to Civ 6 as a counterresponse to the idea of the AI pre Civ 5 were better, and thus picking on Civ 6 makes no sense, nor does attaching vague intentions to it makes much sense. Then, you quoted this one sentence and pretty much went off on a tangent that really has nothing to do with what I was saying, ie difficulty of coding said AI, or whoever's at fault is completely irrelevant to the particular branch of discussion. I mean, sure maybe coding a humanlike AI would be impossible or not fun. So? That was never a point of contention. Why do we have to pick between this AI and some idealized one? So I questioned why you were bringing most of this up, and thought that at least half it had nothing to do with what I was saying. So which I claimed was "obvious" though I didn't really know that was such a strong word. I guess I'll not use the O word anymore.

So do you want to talk about the Civ franchise? Or what?

 

[empresskiova]

I think the AI is trying, but without “tactical thinking” it will only ever do so well. But, if this thing actually has bait tactics, it’s doing a hell of a lot better than I thought.

Plus, now it doesn’t give up Relics for 1 gold coin.

 

[Archon_Wing]

Yea, trade is one thing to get down nicely.

I don't ever expect the AI to be tactically sound..Tthere are certain simple things that have helped so much and didn't require some impossible form of technology. The AI seeking friendship in order to settle aggressively was one of the better changes . The fact that it pissed off a number of people means that it is working as intended. The AI being very willing to dogpile on a target has been the highlight of the most recent patch. So it's not all doom and gloom as it has been getting better, and getting down this foundations means it'll only get better.

Right now it suffers more from failing to upgrade units but tbh this is more of a game mechanic flaw rather something to really hold the AI accountable if it just had the bad luck of not getting resources.

 

[Cerilis]

As some others have commented it seems it is not a failure in the AI not knowing that a settler has to be escorted, but knowing what to do with plans changes. I've normally seen escorted settlers in open land, and even going unescorted could be logical in some situations (forward settling is a "race" game, and sometimes you need to take the risk to not escort your settler, if you know the zone is "safe").

Problem is, when AI loses this "race" (either you build a city first in the spot), or you just block the targetted spot with a unit, it does not know well what to do, therefore the settler remains "roaming" around as a 0-value asset (which, obviously, he is not). There should be a routine to "rebase" settlers that have lost its target, and reuse them if a new opportunity arises, that i think it is what is missing now.

It must be taken into account that, nevertheless, the AI takes a lot of risk with forward settling. One of my games last week, playing Egypt, I saw Brazil lossing one his escorted settlers to barbarians when trying to settle in my island/continent (yup, I got one for myself only in that game). The scort spearman was not able to survive continued attacks by barbarian cavalry (while my war chariot and spearman were just blocking "good settle spots" from them and watching they being attached). Finally, I was able to rescue the settlers from the heathen, and they founded Amarna instead of Sao Paoulo ... Later in the game, for sumeria forward settler, I was not so lucky, and I had to declare war to kill the escort and "convince" again the settlers they would do better if the city they founded was Egyptian .

I think one of the problems is that the AI doesnt consider you are deliberately blocking it. As far as it is concerned, your unit may as well only be passing through and free the tile later again.

 

[trev1972]

Ive seen the AI escorting settlers lately so thats changed?, however the AI is incredibly dumb. The big issue with it for me is that i know when i play that anything but taking advantage of dumb ai for early conquest is sub optimal play.

I dont think it is deliberate, i just think good AI is low down the priority list, only 8.4% of players have won an emperor game, so their target market is extremely casual- which means cool looking features but no big disaster if the games AI doesnt stand up to scrutiny.

I wish they would sort out the AI city placement though, it is unbelievably bad.

One thing i sometimes think is deliberate though is forward settling. Ive been playing island plates maps lately- and the AI sailed past two beautiful empty islands in order to plant a city on the few squares i handnt occupied on my landmass.

 

[Victoria]

The hard part is coding an AI that plays the game (itself) like a human would, with the same limited scope, understanding, and reactions the user has

I can only half argue as I am half a programmer but I respect your view @Abraxis and the line above is quite telling, for example in the case of a settler moving

The AI needs to react correctly to the fog of war and so has no visibility of units we do not. I see strong evidence of this.
The issue then become memory and this I feel is where things get quite complex. Do you think they have programmed in a memory that they saw a knight unit at location X with unit Y 3 turns ago and now unit Z will not move that way with a settler because it is too risky?

Our minds remember things and make assumptions and risks based on memory. I do wonder how difficult it is to retain such a memory from an AI perspective and to react well based on such facts.

 

[Sascha77]

I haven't played Civ VI in a while but I'm trying to hype myself back into it by watching a bunch of quill18's let's-plays. And in those, unescorted settlers are still very much a thing. As is the AI's general incompetence. All the play-throughs I watched were on Deity, which, as difficulty-settings go, is just one gigantic crutch to make up for the AI's inability to declare/wage war. Plus the trades they still go for (or even offer on their own) are nothing less than robbery (if the player accepts the trades, that is).

I'm kinda hopeful that all the new mechanics announced for R&F will change the game for the better - but if the dumb AI remains unchanged, I'm not too sure even those additions will help.
The one thing that helps keep my hope up is the way that Civ V's expansions made that game so much better and transformed it into the game it should've been on the day of vanilla-release. If they can pull off something similar with VI, I'm more than willing to forgive the state vanilla-VI was released in.

S.

 

[Tatran]

Following TheGameMechanic playing Indonesia; the military AI is in a very sad state.

 

[Gedemon]

I can only half argue as I am half a programmer but I respect your view @Abraxis and the line above is quite telling, for example in the case of a settler moving

The AI needs to react correctly to the fog of war and so has no visibility of units we do not. I see strong evidence of this.
The issue then become memory and this I feel is where things get quite complex. Do you think they have programmed in a memory that they saw a knight unit at location X with unit Y 3 turns ago and now unit Z will not move that way with a settler because it is too risky?

Our minds remember things and make assumptions and risks based on memory. I do wonder how difficult it is to retain such a memory from an AI perspective and to react well based on such facts.

We don't have access to civ6 code but civ5 AI had no "memory"

I don't think it would be impossible to do (I had given a basic memory to the AI in my WWII mod by hardcoding some decisions based on a combat log of the previous turns for example), but it's as always a matter of development time in relation to what need to be improved.

 

[sonicmyst]

Let me chime in and stir the pot of discussion by asking if you believe that improving the AI might significantly affect turn times? I am not sure if this has been discussed before.

 

[Sascha77]

if you believe that improving the AI might significantly affect turn times? I am not sure if this has been discussed before.

I'm not a programmer, so let me just throw this idea out:

One of the reasons why turns get long - especially in the later game - has got to be the fact that the AI is rocking carpets of doom. Many units = many decisions to make/actions to perform.
A "smarter" AI that is more competent at warfare wouldn't have to rely on insanely huge armies, so wouldn't that alone mean that AI turns would go quicker overall?

E.

 

[Josephias]

I'm not a programmer, so let me just throw this idea out:

One of the reasons why turns get long - especially in the later game - has got to be the fact that the AI is rocking carpets of doom. Many units = many decisions to make/actions to perform.
A "smarter" AI that is more competent at warfare wouldn't have to rely on insanely huge armies, so wouldn't that alone mean that AI turns would go quicker overall?

E.

Really, no if all of these lesser in number unit has to keep records of "what were my last turn results and situations" that have to be checked to decide what to do next. You may reduce units to 1/2, but maybe increase computation time per unit by 3, therefore you keep increasing.

Human intelligence is very good at discarding what is considering uniseful data for the calculation (this causes us to incur in dumb errors sometimes, in the other hand). Discarding options to investigate in-depth is also part of computer AI, but it is not nearly as flexible as the human brain regarding this - they are better at calculating multiple potential outcomes, tough, but this works better with a sensible number of options (chess, go...), and without missing information (no fow)

 

[Rosty K]

I'm not a programmer, so let me just throw this idea out:

One of the reasons why turns get long - especially in the later game - has got to be the fact that the AI is rocking carpets of doom. Many units = many decisions to make/actions to perform.
A "smarter" AI that is more competent at warfare wouldn't have to rely on insanely huge armies, so wouldn't that alone mean that AI turns would go quicker overall?

E.

Actually it seems to me that in the late game the greater part of the turn time is taken by City-States/Barbs, not by major civs.