Sept 22 AI Discussion

[Sansa_Stark]

Some of you can just put your fingers in your ears, sing la-la-la and ignore any possible worries, can't you?

I'd say this is a legitimate worry. While the reasons given in this thread could apply, it is a very short time until release, and you'd think if they'd have a proper AI, they'd want to show it. Not a peep has been heard about the AI for all this time, apart from much talk about their strategic agendas, etc. The combat AI has barely been mentioned (afaik; please direct me to a video where they talk about it if it does exist). Worry level: 8/10.

No, I agree it is weird and I can't say I am happy about it. But everything else what devs did with Civ6 seems to be great, so I think there are reasons to believe this is actually not that bad.

AI wasn't good in Civ5 and I did enjoy 1000+ hrs without any mods nevertheless. I don't expect AI to be much better in Civ 6 and people who do are setting up themselves for huge disappointment. But realistically, do you really think that infantry vs warriors is possible on any reasonable level? I have no idea why it happens in these streams but I am going to assume that's really some easy level and not worry about quality of my experience on Emperor+.

 

[CaiusDrewart]

I agree that, based on the streams, the AI looks to be even worse than Civ V's was. And I thought that was such a low bar that they couldn't possibly fail to clear it.

 

[Magil]

Well, it's no better than relying on inductive reasoning to turn one happenstance into an 8/10 concern. Are you too used to playing on Deity to think such discrepancies should *never* transpire?

This isn't the only time though. While it's the only time it's gotten as far as the modern era, it's also happened in the games where the player has advanced into the medieval/renaissance and seen warrior blobs wandering around, particularly for city-states. It's a reasonable concern at this point.

 

[stiiknafuulia]

Well, it's no better than relying on inductive reasoning to turn one happenstance into an 8/10 concern. Are you too used to playing on Deity to think such discrepancies should *never* transpire?

As I said, they've barely made a peep about the combat AI. Given that it was a major complaint for many players in Civ V (for me it ruined the game), you'd think they'd want to reassure us that the AI has been improved. Instead, there's been total silence, and now in this vid (and others before it) we see what looks an awful lot like evidence to the contrary (i.e. brainless AI). If it quacks like a duck, and all that... It doesn't help that I've seen this exact sequence of events unfold multiple times with other games (evidence of bad AI, no evidence for good AI, fan protestations about how everything will be a-ok come release day), and all of those times there were great lamentations to be heard when the actual state of the AI became apparent upon release. I'd be very happy to be wrong about this, but I will excercise great moderation when it comes to my hopes about this.

EDIT: Although to be fair, it's bad strategic AI to not build good units, not combat AI. But we haven't really seen any good examples of the latter, either (at least not that I'm aware of).

 

[Krajzen]

You know it, I know it, all us regulars at Civfanatics know it. The AI will be as good as it was in civ5, civ4, civ3, civ2, civ1 and all the side games.

Its not that i think Firaxis is incompetent (they are very competent) but the AI has been poor in so many civ games (and even x-com which was a seperate team!) that wishing or hoping for better AI in the next game is just a pipe dream.

It has been discussed to death; the players that would apprechiate better AI are a minority. Plenty of players still lose on Prince and below difficulties so AI is good enough. As lazy and annoying as it is, if you want a challenge you will just have to enable AI cheats.

Given that it was a major complaint for many players in Civ V (for me it ruined the game), you'd think they'd want to reassure us that the AI has been improved.

The truth is much simpler. Creating advanced AI actually 'good' at playing complicated games like civ is simply impossible on the current tech development level. No 4X strategy game has genuinely good AI. And we won't see any really impressive good AI unless some technological breakthrough.

Hell, modern games do not even have an artificial intelligence - just weighed lines of stupid blind code programmed to react on certain imput (sometimes with semi random elements) that is prone to the stupid simple bugs like math errors (because programmer was tired typing said code) that is supposed to make the passable impression of any intelligent process behind it.
Do not expect good strategy game AI for next few years.

I mean, what should we demand is AI that can manage game mechanics and create any (if predictable) danger for human player. But genuinely good AI, learning AI, AI using advanced tactics and traps, very adaptive AI, intelligent AI, AI inventing unorthodox ways - this is simply impossible at the moment. And certainly not in the game with extreme amount of variables, such as civ.

Firaxis didn't say anything about combat AI because they know they can't do much about it, and they can't say "we have passable mediocre ai like other games" for marketing reasons.

 

[homan1983]

Last stream when ED played with Norway Japan AI has 0 ships and 0 units on turn 150.

In order to be fair to them, that was an advanced start game (i.e. the game started in the later eras). It's much more forgivable when the AI can't perfectly handle wonky settings. Not so, when it can't play the base game on standard settings (i.e. quick/normal on continent/pangea)

On the other hand, I have consistently seen, in every single lets play, from youtubers to firaxis presentations, that the AI does not upgrade its units on time and/or is behind significantly.

In the earlier times I would just put it down to: "refining the AI is the last step after all features are implemented". Now it's much more concerning. The game has probably gone gold and early versions have been given to youtubers who are playing it privately until the NDA is lifted next week.

One of the worst feelings I had in civ5 was that, when playing a game, I felt that AT ANY TIME, FOR ANY REASON, I could take out anyone I wished, regardless of how much stronger they were. It felt like I was just playing pretend-civilization, going through the motions and keeping myself amused.

 

[Hajee]

huh whats the issue?

One civ was be hide a bit on unit upgrades?

hmm, seems legit, I mean only if the Native Americans had Guns and the Euro invaders did not, what would history be like today

 

[Brutus2]

Let's keep in mind that these games are set up to purposefully show some elements, maybe they aren't played straight.

They said in the stream the game was set up by Pete. They also seemed surprised by how much faith Scythia was producing and then showed a bunch of Kurgan all built next to each other even though that should not be possible.

So this is not a real game play.

 

[salty mud]

I've seen countless game releases and I've seen countless apologists desperately justifying every bad detail in game trailers. No different here. If anyone thinks that the material we see here, a month from release, will be significantly different to the product released next month then you're simply delusional and ignorant of the video game industry.

 

[BrokenBokken]

The truth is much simpler. Creating advanced AI actually 'good' at playing complicated games like civ is simply impossible on the current tech development level. No 4X strategy game has genuinely good AI. And we won't see any really impressive good AI unless some technological breakthrough.

Hell, modern games do not even have an artificial intelligence - just weighed lines of stupid blind code programmed to react on certain imput (sometimes with semi random elements) that is prone to the stupid simple bugs like math errors (because programmer was tired typing said code) that is supposed to make the passable impression of any intelligent process behind it.
Do not expect good strategy game AI for next few years.

I mean, what should we demand is AI that can manage game mechanics and create any (if predictable) danger for human player. But genuinely good AI, learning AI, AI using advanced tactics and traps, very adaptive AI, intelligent AI, AI inventing unorthodox ways - this is simply impossible at the moment. And certainly not in the game with extreme amount of variables, such as civ.

Firaxis didn't say anything about combat AI because they know they can't do much about it, and they can't say "we have passable mediocre ai like other games" for marketing reasons.

I think this needs to be understood. It sucks and it can definitely hurt game play, but I don't know what you expect to be done about it. I'm sure they could make it better than it is currently, but how much better... who knows? If the AI still worthless even when they put a lot of resources in to increasing it, they might be a larger return per dollar putting resources elsewhere. Hopefully patches and mods will at least let it handle getting the correct strength units on the board even if it won't help tactically.

 

[anandus]

You know it, I know it, all us regulars at Civfanatics know it. The AI will be as good as it was in civ5, civ4, civ3, civ2, civ1 and all the side games.

Its not that i think Firaxis is incompetent (they are very competent) but the AI has been poor in so many civ games (and even x-com which was a seperate team!) that wishing or hoping for better AI in the next game is just a pipe dream.

It has been discussed to death; the players that would apprechiate better AI are a minority. Plenty of players still lose on Prince and below difficulties so AI is good enough. As lazy and annoying as it is, if you want a challenge you will just have to enable AI cheats.

Or play against humans

 

[Human Crouton]

They set these games up so that they have a 0% chance of losing the game and a good chance of showing what they want to show during their stream. When the press people got to play several weeks ago they were not having it as easy.

 

[Magil]

I think this needs to be understood. It sucks and it can definitely hurt game play, but I don't know what you expect to be done about it. I'm sure they could make it better than it is currently, but how much better... who knows? If the AI still worthless even when they put a lot of resources in to increasing it, they might be a larger return per dollar putting resources elsewhere. Hopefully patches and mods will at least let it handle getting the correct strength units on the board even if it won't help tactically.

Not expecting a miracle when it comes to AI. But I think "don't have warriors in the Medieval/Modern era" is a pretty low bar to clear.

 

[stiiknafuulia]

Not expecting a miracle when it comes to AI. But I think "don't have warriors in the Medieval/Modern era" is a pretty low bar to clear.

This. It was Prince and they were showing off the Legions, but still... Since they know about these concerns among the more strategically-minded, why not alleviate them if they can? Why not show a test-game where an AI has built 20 Medieval cities, has tech parity and is wielding its units in at least a non-Civ V braindead kind of way (embarking in the water and the like)? If they really can't do that in 2016, hell, I might as well just migrate to playing against only human players.

 

[TheMarshmallowBear]

This. It was Prince and they were showing off the Legions, but still... Since they know about these concerns among the more strategically-minded, why not alleviate them if they can? Why not show a test-game where an AI has built 20 Medieval cities, has tech parity and is wielding its units in at least a non-Civ V braindead kind of way (embarking in the water and the like)? If they really can't do that in 2016, hell, I might as well just migrate to playing against only human players.

If this is to be believed then the build is fairly recent, and likely to have had the pacing of the game done correctly in terms of age vs turn times.

Each era in Civ 5 takes about 50 turns on Quick.

He was researching PLASTICS in 1650 AD. which struck me as odd, so I presume Pete may have cheated a little. Or there might've been some sort of bug, since Scythia had a crazy faith bonus.

As far as I know, the Ai was on par, because Japan had Crossbowmen in 1650 AD.

Remember that unlike Civ 5 God and Kings/BNW, many units are missing, so there's HUGE gaps (once again) in the line units (Archer > Crossbowmen instead of Archer > Composite Bowmen > Crossbowmen

Even on Diety, players are able to beat the game and win a Science victory in crazy amount of time.

 

[homan1983]

The 2nd game was also played and not set up. For one thing, they said that one of Congo's cities had more war weariness because it was a conquered city, and Ed Beach said he can explain it because Kristiansan is not a congolese name... they laugh and say Pete has been a naughty person.

https://www.twitch.tv/firaxisgames/v/90778502?t=1h03m34s

They also take a look and discuss how the game has developed (Which they wouldn't if it was set up).

With regards to the Kurgan next to eachother: either the rules were changed, or the AI is not following the restriction which would be a bug.

Also interesting, I don't know if we'd seen this before but the Kurgan gives +2g +1faith, and if next to pasture +2g +2faith. This seems pretty good, interesting to see them have a non-militaristic option.

 

[CaiusDrewart]

I actually don't mind the thought of huge military technology disparities being possible. There were plenty of times in real world history where rifles and cannons clashed with bows and spears. I wouldn't mind seeing this happen in-game every so often.

With that said, if AIs are consistently not upgrading their units and not keeping up, that's a big problem. There needs to be challenge.

Generally speaking, I'm very pessimistic about the Civ VI AI. Tactically, 1UPT is as complicated as ever and we all know how that went over in Civ VI. Now the economic side of the game has a complexity to match, with the AI needing to choose which districts to build and where best to place them. That's going to be a mess.

 

[criZp]

Isn't the youtube dudes releasing videos in a week? It's just to wait and see then.

 

[MistyRonin]

You know it, I know it, all us regulars at Civfanatics know it. The AI will be as good as it was in civ5, civ4, civ3, civ2, civ1 and all the side games.

Its not that i think Firaxis is incompetent (they are very competent) but the AI has been poor in so many civ games (and even x-com which was a seperate team!) that wishing or hoping for better AI in the next game is just a pipe dream.

It has been discussed to death; the players that would apprechiate better AI are a minority. Plenty of players still lose on Prince and below difficulties so AI is good enough. As lazy and annoying as it is, if you want a challenge you will just have to enable AI cheats.

I couldn't express it better.

Besides with all the complex features that Civ 6 will add, it's really hard to expect an incredibly good AI. To truly enjoy a proper and intelligent opponent that takes advantages of the features, you have to fight other humans in MP.

 

[Metecury]

If you want a real challenge play with real people in multiplayer, as others have said the vast majority of players (cusal non "fanatics2 of civ) are just fine with this type of AI. So either one settles for cheating AI on higher difficulties or play with friends (which honestly makes for great diplomacy, war and generally lots of fun)