Prince AI is pathetic

[Mathias Rex]

To be clear, the main problem here isn't the AI per se, but that the AI don't build and upgrade military units. That should be fixable.

But I'm exited where prince level is on the difficulty ladder. It can make a huge difference if it's nr 3/4 out of 6, or nr 3/4 out of 8. AFAIK we don't know that yet.

And at the same time city states can produce huge amount of units, so the main AI players maybe just have to prioritize better.

 

[Krajzen]

Yep, it is pathetic. And don't excuse it with difficulty level or mechanics, it would be similarly pathetic on Deity.

On a bright side, I have been a long time tester of a particular massive strategy game last year, and AI was like very last thing to be done - 1,5 months before release of said game its AI was braindead moron which could grasp maybe 30% of the game's features and often didn't develop technology at all. In the remaining 6-7 weeks whole AI was build from "barely existant tragedy" to "decent" level.

We have 3 weeks to civ6 release and the version they are playing is few weeks old.

 

[Mustakrakish]

Yep, it is pathetic. And don't excuse it with difficulty level or mechanics, it would be similarly pathetic on Deity.

On a bright side, I have been a long time tester of a particular massive strategy game last year, and AI was like very last thing to be done - 1,5 months before release of said game its AI was braindead moron which could grasp maybe 30% of the game's features and often didn't develop technology at all. In the remaining 6-7 weeks whole AI was build from "barely existant tragedy" to "decent" level.

We have 3 weeks to civ6 release and the version they are playing is few weeks old.

At this point I heard this so many times I'm starting to believe it. We'll see just how "decent" it will be...

 

[Krajzen]

At this point I heard this so many times I'm starting to believe it. We'll see just how "decent" it will be...

By "decent" I mean of course "not enough bad to be abnormal for stragegy games" because I know some stuff about AI and I am still convinced doing genuinely good AI in so complicated games will be impossible for years, until technological breakthrough.

Modern games don't actually have any "real" AI to speak of, just a lot of weighed lines of code typed by tired programmer overnight that try to make passable impression of any kind of intelligence. Self learning AI, adaptive AI, creative AI - as far as I know thode things are impossible ATM, definitely outside of laboratories.

So, expect this AI to be as mediocre as in all strategy games where you wasn't complaining too much - that will be very good for modern standards.

The fact civ6 AI can move and shoot in one turn is already great improvement over civ5

 

[lietkynes]

Can't comment on the tactical AI yet but the strategical AI is a mess, it clearly has no idea of what it's doing.

And game balance is all over the place. Cities are too weak, Eureka boosts are too powerful, Unique units not requiring resources are overpowered.

 

[Mustakrakish]

By "decent" I mean of course "not enough bad to be abnormal for stragegy games" because I know some stuff about AI and I am still convinced doing genuinely good AI in so complicated games will be impossible for years, until technological breakthrough.

Modern games don't actually have any "real" AI to speak of, just a lot of weighed lines of code typed by tired programmer overnight that try to make passable impression of any kind of intelligence. Self learning AI, adaptive AI, creative AI - as far as I know thode things are impossible ATM, definitely outside of laboratories.

So, expect this AI to be as mediocre as in all strategy games where you wasn't complaining too much - that will be very good for modern standards.

The fact civ6 AI can move and shoot in one turn is already great improvement over civ5

You don't say!

I don't think anyone (at least half in their minds) expects "self learning AI" a.k.a real AI

What we need is a proper list of "weighed lines of code" that are NOT typed by a tired programmer overnight THAT with a lot of testing and re-writing of said "weighed lines of code" by well hydrated and not sleep deprived coders

 

[Evs]

Nobody expects a self learning, adaptive, creative AI.

Just an AI that can play the base game competently.

This was never achieved imo in CIV5 until the community patch. So if the community can code a "decent" AI, why can't the Devs whose job it is ?

The link above at 6:25 is where Ed Beach tells us how amazing it is to watch the AI battles. This video is from May (5 months ago). Where is the evidence of this in the recent YouTube videos ?

 

[jjkrause84]

It's happening all over again by the looks of it. I'm getting the feeling of Dejavu....

If you think the AI will be much better on release, remember the AI upon release of CIV5 ?

The game was unplayable for me for months. There was patch after patch after patch to improve the appalling AI so don't think Firaxis will not release the game without the AI being competent.

This is because certain pillars of Civ 5, like 1UPT, are simply game-breaking. By carrying them over into Civ 6 it ensures that Civ 6 will also be inherently broken. An AI that can't handle it is par for the course, as are traffic jams that are more frustrating than fun.

Return unit stacking (preferably limited stacking with ZOCs), make units defend cities, and either grow the size of the map of make cities smaller and then Civ can begin to heal. I love a lot of the new ideas Civ 6 is bringing to the table (especially the incredible flexibility the government and casus belli systems offer). Unfortunately, in my opinion, 1UPT means a broken game no matter what you do. The AI cannot handle it

 

[Evs]

This is because certain pillars of Civ 5, like 1UPT, are simply game-breaking. By carrying them over into Civ 6 it ensures that Civ 6 will also be inherently broken. An AI that can't handle it is par for the course, as are traffic jams that are more frustrating than fun.

Return unit stacking (preferably limited stacking with ZOCs), make units defend cities, and either grow the size of the map of make cities smaller and then Civ can begin to heal. I love a lot of the new ideas Civ 6 is bringing to the table (especially the incredible flexibility the government and casus belli systems offer). Unfortunately, in my opinion, 1UPT means a broken game no matter what you do. The AI cannot handle it

Agreed. They should never have carried 1UPT over without a solution to the broken AI we got with CIV5.

I also like a lot of the new ideas and features but they are all meaningless imo if what you are playing against is incapable of using them.

 

[Mathias Rex]

Another thing which the AI is good at: faith production, and religious unit spam. It shows too that its maybe a prioritizing problem.

 

[Jabulani]

Sorry to say but thr AI level equals the guy/s civ skill who developed it......

If the AI was being tuned by a bunch of expert civ players instead....we would certainly see less bonus and much much more A I !

 

[Martin Alvito]

If the AI was being tuned by a bunch of expert civ players instead....we would certainly see less bonus and much much more A I !

It'll still take some time to figure out what the hard and fast rules are and which options need to be considered when in order to play optimally. And every time you tweak the base parameters, the equilibria change.

That said, the CIV AI was a beast on Deity in vanilla in the early game so it's certainly possible to ship with functional AI. I'm not surprised that it plays badly at this point; with all the new moving parts it seemed all but certain that the AI would play very suboptimally at least for the first few patch cycles.

Seems like my concerns that late game science costs were too low were correct. Easily fixed, though.

 

[Jabulani]

It'll still take some time to figure out what the hard and fast rules are and which options need to be considered when in order to play optimally. And every time you tweak the base parameters, the equilibria change.

That said, the CiV AI was a beast on Deity in vanilla in the early game so it's certainly possible to ship with functional AI. I'm not surprised that it plays badly at this point; with all the new moving parts it seemed all but certain that the AI would play very suboptimally at least for the first few patch cycles.

Seems like my concerns that late game science costs were too low were correct. Easily fixed, though.

Imo the fact that the pathetic AI could be bettered in the next months doesn't justify the fact that now is horribly pathetic.

Who are civ vi betatesters ? Are they perhaps traffic lights ?

 

[gfeier]

The biggest problem I had with the AI didn't involve units. The AI civs were WAY too generous in negotiations. In one of Marbozir's videos, the AI civ would not make peace even up but would if they paid Marbozir. The AI also seems to place little value on great works.

 

[CaiusDrewart]

This is because certain pillars of Civ 5, like 1UPT, are simply game-breaking. By carrying them over into Civ 6 it ensures that Civ 6 will also be inherently broken. An AI that can't handle it is par for the course, as are traffic jams that are more frustrating than fun.

Return unit stacking (preferably limited stacking with ZOCs), make units defend cities, and either grow the size of the map of make cities smaller and then Civ can begin to heal. I love a lot of the new ideas Civ 6 is bringing to the table (especially the incredible flexibility the government and casus belli systems offer). Unfortunately, in my opinion, 1UPT means a broken game no matter what you do. The AI cannot handle it

Unit stacking would make me so, so happy... I fondly remember the days of Civ IV, when it was actually scary if an AI with a big army and lot of tech declared war on you. You couldn't just pick off their entire army one by one with a group of archers while the AI's units pathetically milled around. The AI would actually (albeit slowly) take your cities! Imagine that.

 

[Ulthwithian]

I can say that creating adaptive AI is not that difficult of a challenge. (System Dynamics is a modeling method that has it as its 'schtick'.) The issue is ensuring it works. You may consider this a 'laboratory' example, as it is primarily an academic method, but it's also used world-wide by large consulting firms to analyze and improve systems.

I will also point out that applying such methods to computer games is generally much easier than applying them to 'real life', as you have clean data and, more importantly, you have a complete knowledge of what is possible.

I would imagine that the greatest difficulty in applying SD specifically here is the amount of time it might take to perform.

 

[Big J Money]

If you think 1UPT ruined civ military AI, just wait to see what unstacking cities will do. I hope it won't be as game-breaking to people who love strong AI.

I'm one of those lucky people that never gets quite good enough in civ to leave the King-Immortal range, so even with crap AI my opponents can still out produce me and make for a challenging game. I actually kind of like that in order to win the game I have to notice and exploit the AI's mistakes with my strategic play being superior. If I were constantly winning on Deity I could see a cause for complaint though.

 

[Ebrim]

Yeah, the bit of the one let's play I watched was a bit unsettling. It's true that Prince was always about how you want to win or just playing around but there shouldn't be no pushback from the AI at all. I agree with what many are saying that it seems like a priorities issue. The AI just isn't building an army. Even if it's a peaceful civ it needs to maintain some sort of military force for unexpected events (like surprise wars). Cities near the borders should always have some sort of garrison (way more important in VI than V).

 

[juanpavo]

The fact civ6 AI can move and shoot in one turn is already great improvement over civ5

Great improvement, yes. Pathetic and embarrassing that it took so long, also yes.

 

[EricTheGreat12]

The build is probably several weeks old, if not more (its not like they were going to send it out without testing it thoroughly).

But there are plenty of (other) reasons to expect the AI to be terrible.

The problem though stems from the fact that the AI in this situation is just terrible. It's extremely doubtful that a miracle occurs and that the AI is completely overhauled. As far as I'm guessing, the only way the AI will be improved is through patches.