Do you want an AI that can make reasoned choices?

[HorseshoeHermit]

'Prevent victory' is not the way any agent should play any game.

It's a game. You play it because it has a winner, and you cared about determining who it would be and then honoring that laurel afterward. Playing to tear away a win from the winner - whoever that is, until someone else is the winner and then tearing that down too - is to pooh-pooh the game itself, and defeat your very participation in the thing.

Keep playing with whatever your strategy was to gain progress to the goal and commit. If your tactic does not get to the end first, then confirm it and let yourself value the tactic that did actually get there.
AIs should also play like this.

 

[oSiyeza]

'Prevent victory' is not the way any agent should play any game.

It's a game. You play it because it has a winner, and you cared about determining who it would be and then honoring that laurel afterward. Playing to tear away a win from the winner - whoever that is, until someone else is the winner and then tearing that down too - is to pooh-pooh the game itself, and defeat your very participation in the thing.

Keep playing with whatever your strategy was to gain progress to the goal and commit. If your tactic does not get to the end first, then confirm it and let yourself value the tactic that did actually get there.
AIs should also play like this.

I would put it like this. Actual Leaser personalites thatdetermi e gameplay. I stead of wird pointless bonus and mostly meaningless agendas.

 

[civac]

'Prevent victory' is not the way any agent should play any game.

It's a game. You play it because it has a winner, and you cared about determining who it would be and then honoring that laurel afterward. Playing to tear away a win from the winner - whoever that is, until someone else is the winner and then tearing that down too - is to pooh-pooh the game itself, and defeat your very participation in the thing.

Keep playing with whatever your strategy was to gain progress to the goal and commit. If your tactic does not get to the end first, then confirm it and let yourself value the tactic that did actually get there.
AIs should also play like this.

In MP FFA games you can get a dynamic that contenders will try to spoil each others victory if possible. For instance, if your science victory is a few turns slower than somone else's culture victory it makes sense to raze some of their cities if you can. However, players that are not contenders and can't win themselves anymore are expected to stay out of the victory race.

 

[rgp151]

I never understood the claim that people wouldn't want to play against a more aggressive and capable AI. That's what settings are for!

I think most would rather that the increased difficulty setting increased the intelligence of the AI rather than giving the AI resource advantages. The resource advantages seem much more unfair.

I think it would be much better if increasing difficulty in 4X games was more like increasing difficulty in Chess games as opposed to just a matter of giving the AI players larger resource advantages. And anyway, all of it can be controlled via settings.

 

[reddishrecue]

Compared to the past AI in past versions of civilization; AI has improved a lot. There was no immortal and no deity difficulty AIs in past version. Immortal and Deity appeared in what... between civ 2- civ 4?? I know there is a deity and immortal AI in civ 4 which has deity as almost impossible to beat but it doesn't have AI intelligence since the AI is just spoiled! More settlers, workers and technologies... more of everything! that gets the AI successful asap with about the equal effort from the AI to the rest of the difficulties. The only difference is that the difficulties spoil the AI more.

 

[Leyrann]

I think most would rather that the increased difficulty setting increased the intelligence of the AI rather than giving the AI resource advantages. The resource advantages seem much more unfair.

I would say that in theory, they are the same. That said, resource advantages can feel more unfair, and on the other hand, in reality the resource advantages are probably less conducive to increasing difficulty; you see very strongly with Civ VI that on the highest difficulties, the early game is pretty difficult, and then once you reach the mid-game you catch up to the AI, and by the lategame it's a breeze. Resource advantages are probably also significantly easier to introduce, however.

Compared to the past AI in past versions of civilization; AI has improved a lot. There was no immortal and no deity difficulty AIs in past version. Immortal and Deity appeared in what... between civ 2- civ 4?? I know there is a deity and immortal AI in civ 4 which has deity as almost impossible to beat but it doesn't have AI intelligence since the AI is just spoiled! More settlers, workers and technologies... more of everything! that gets the AI successful asap with about the equal effort from the AI to the rest of the difficulties. The only difference is that the difficulties spoil the AI more.

That's not true. Civ 2 already had Deity for certain, I don't know about Civ 1, but I think it existed there as well. Civ 4 did have more total difficulty levels, with Noble and Monarch not appearing in any of the other games as far as I know. Also, the AI in Civ 6 gets spoiled just as hard on high difficulty levels. On Deity, they start with 3 settlers, 2 builders and 5 warriors. As well as a 100% production and gold income bonus, and 40% science, culture and faith bonuses. (and a combat bonus stronger than standing on forest)

 

[Gorbles]

It's a game. You play it because it has a winner, and you cared about determining who it would be and then honoring that laurel afterward. Playing to tear away a win from the winner - whoever that is, until someone else is the winner and then tearing that down too - is to pooh-pooh the game itself, and defeat your very participation in the thing.

I mean, I just want to point out that this is absolutely how some players view MP games, particularly matches involving more than two players. Nor is this limited to Civilisation either - it's been a type of play in most strategy or strategy-related games I've played over the past 20 years.

So the question is: should we model the AI on how players act (which would include this), or should we model the AI on how we think players should act? Because the latter, nevermind being difficult to agree on in the first place, also means we're intentionally looking to hobble the AI just because it might make moves that hurt you over other players.

 

[reddishrecue]

That's not true. Civ 2 already had Deity for certain, I don't know about Civ 1, but I think it existed there as well. Civ 4 did have more total difficulty levels, with Noble and Monarch not appearing in any of the other games as far as I know.

Civ 2 had Deity, I would agree with you on that but what you're saying for civ 1 having Deity is not true. Civ 1 had up to emperor.

Also, the AI in Civ 6 gets spoiled just as hard on high difficulty levels. On Deity, they start with 3 settlers, 2 builders and 5 warriors. As well as a 100% production and gold income bonus, and 40% science, culture and faith bonuses. (and a combat bonus stronger than standing on forest)

I said that AI was more spoiled, the only difference is that you showed the increased percentages and the additional units as well as promotions. I see you did careful research on this which is awesome but I was saying spoiled which shows a brief idea of what you're saying.

 

[civac]

Measured by the ability to play the game effectively the Civ6 AI is far weaker than Civ3 or Civ4 AI. Both of the latter are able to field viable armes and mount successful military campaigns. Also Civ4 AI can build a strong and properly developed lategame economy.

 

[TheMeInTeam]

It's a game. You play it because it has a winner, and you cared about determining who it would be and then honoring that laurel afterward. Playing to tear away a win from the winner - whoever that is, until someone else is the winner and then tearing that down too - is to pooh-pooh the game itself, and defeat your very participation in the thing.

Keep playing with whatever your strategy was to gain progress to the goal and commit. If your tactic does not get to the end first, then confirm it and let yourself value the tactic that did actually get there.
AIs should also play like this.

Actually, there is no clear reason the AI should play badly on purpose.

Preventing other(s) from winning and going on to win yourself *is* a tactic that "actually gets there". In fact, if you are capable of executing this tactic consistently, it is guaranteed to get you there.

That's a big if, however.

Measured by the ability to play the game effectively the Civ6 AI is far weaker than Civ3 or Civ4 AI. Both of the latter are able to field viable armes and mount successful military campaigns. Also Civ4 AI can build a strong and properly developed lategame economy.

Civ 4 deity has significantly more time pressure from AI, too. It will win space in 1800s often, and sometimes culture much faster than that. You have to be ready to either win your own VC sooner, or wade through many hundreds of units (in the case of space racing AIs, including the most advanced available + nukes) in order to block their victory long enough to get one yourself.

Other posters are right that this is at least in part due to bonuses at a level it hasn't had since. Stacking also favors AI, since it can concentrate its production bonuses into a threatening force much more easily than 1UPT, which inevitably gets constrained by frontage available.

 

[aieeegrunt]

Measured by the ability to play the game effectively the Civ6 AI is far weaker than Civ3 or Civ4 AI. Both of the latter are able to field viable armes and mount successful military campaigns. Also Civ4 AI can build a strong and properly developed lategame economy.

The Civ Rev games the AI is more threatening

 

[uhu]

Playing to tear away a win from the winner - whoever that is, until someone else is the winner and then tearing that down too - is to pooh-pooh the game itself

I just want to point out that this is absolutely how some players view MP games, particularly matches involving more than two players. Nor is this limited to Civilisation either - it's been a type of play in most strategy or strategy-related games I've played over the past 20 years.

Yep! Already Francis Tresham's original Civilization board game (published in 1980 by Hartland Trefoil) played also like this. When the crowd reached the winning zone, it was bad to be too obviously in front, because then most played against you, but also bad to be too much behind ... [Goal was to reach a point limit; resource cards, units and cities counting VERY few, civilizing achievements ("techs&civics") very much]

Civ 2 already had Deity for certain, I don't know about Civ 1, but I think it existed there as well.

Civ 2 had Deity, I would agree with you on that but what you're saying for civ 1 having Deity is not true. Civ 1 had up to emperor.

Yes. "Deity" in civ1 would have been (nearly?) unplayable, because the main difference in difficulty was given by the number of content citizens per city: emperor=1, king=2, prince=3 = AI level, chieftain=4, settler=5 ... all following citizens were born discontent.
Starting with "Deity"=0 content citizens would have meant, that already the very first citizen would refuse to work a city tile (and MUST be turned into an entertainer immediately to avoid the city going into rebellion = producing NOTHING at all) until a temple (needs ceremonial burial) is built from the sole production of the city tile ...

 

[Noble Zarkon]

I will be shocked if Civ 7 does not have a neural-net-based AI that is capable of a superhuman level of play.

Prepare to be shocked! The market for such a game would be very small, most people regard the AIs are there to provide flavour.

 

[Leyrann]

Yes. "Deity" in civ1 would have been (nearly?) unplayable, because the main difference in difficulty was given by the number of content citizens per city: emperor=1, king=2, prince=3 = AI level, chieftain=4, settler=5 ... all following citizens were born discontent.
Starting with "Deity"=0 content citizens would have meant, that already the very first citizen would refuse to work a city tile (and MUST be turned into an entertainer immediately to avoid the city going into rebellion = producing NOTHING at all) until a temple (needs ceremonial burial) is built from the sole production of the city tile ...

I remember that in Civ 2 Deity, once you started growing bigger, this in fact started happening. However, there was also the 'very unhappy' citizens, and for some reason (most likely a bug) those only cost 2 luxuries to make happy, rather than 6 (regular discontent cost 4 luxuries, content cost 2 luxuries). That led to an exploit where it was incredibly happy, so long as you played it right, to keep expanding rapidly and then build a very happy empire.

 

[Alaindor]

Do you think then, that it's feasible to train an AI with machine learning for Civilization?

Because I've seen multiple people in this thread argue that it isn't.

I've posted this link in another thread but never got any comments on it... I'd love to hear some!

https://techcrunch.com/2016/12/06/a...-human-players-at-complex-civ-strategy-games/

That was Freeciv, probably less complex than Civ6, but it was in 2016, a massive six years ago and technology has moved forward quite a bit since in that domain! So I'd argue that yes, it can be done.

 

[Alaindor]

I've posted this link in another thread but never got any comments on it... I'd love to hear some!

https://techcrunch.com/2016/12/06/a...-human-players-at-complex-civ-strategy-games/

That was Freeciv, probably less complex than Civ6, but it was in 2016, a massive six years ago and technology has moved forward quite a bit since in that domain! So I'd argue that yes, it can be done.

Never got any reply to this link although it's the third time I mention it in three different posts over the last months!

Ok, let me reply myself with a question. Is this too disturbing, or underwhelming, or off-topic, or hard evidence that yes, machine-learning is possible and affordable for Civ6? Any expert opinion here?

 

[Naokaukodem]

Never got any reply to this link although it's the third time I mention it in three different posts over the last months!

Ok, let me reply myself with a question. Is this too disturbing, or underwhelming, or off-topic, or hard evidence that yes, machine-learning is possible and affordable for Civ6? Any expert opinion here?

Well AIs are designed to be beaten. I mean for video games, not against Kasparov. Otherwise no need of machine learning, just give Deity AI 10 starting settlers.

 

[vorlon_mi]

Never got any reply to this link although it's the third time I mention it in three different posts over the last months!

Ok, let me reply myself with a question. Is this too disturbing, or underwhelming, or off-topic, or hard evidence that yes, machine-learning is possible and affordable for Civ6? Any expert opinion here?

I think that I read the article back when it came out, but didn't remember some key details. So i re-read it today. In my day job, I work in a department that supports teams doing AI, but I don't work in it myself. I played a handful of games in FreeCiv more than 15 years ago.

1. In the accompanying video, the ML logic first settles cities, constructs improvements, and builds its empire. Later in the video, it is shown defending itself from an invasion. It pursues a space race victory, since that is one of the two victory types in Civ2.
2. In the text, the learning aspect is described as using machine reasoning, where experience from humans is entered as words and phrases. Put another way, the company could have fed in one of the CivFanatics strategy guides recommending techniques in the early game, combat, and achieving victory.
3. FreeCiv, like Civ2 and Civ3, allows unit stacking and traversing over mountains. Mountains became impassable in Civ4, and remained impassable (with limited exceptions) in Civ5 and Civ6. Thus, the logic for moving units is much less complex than Civ6.
4. The article makes no reference to the computing resources needed to run their ML logic while it plays the game.

So, I believe it would be possible for Arago (the company) to train HIRO (their software) to play Civ6. Since they would ingest human-generated experience about gameplay, such as how to use siege weapons to assault a city, I would expect HIRO to do that competently. I would expect that HIRO would pursue a space victory in Civ6 competently. HIRO might even pursue a culture victory effectively.

Key questions:
- What would be the economic model for making money off such an AI? Perhaps online play against HIRO, where a server in the cloud is running the logic? Would players pay a subscription fee to have a HIRO AI opponent? Would it be profitable for Firaxis to include a surcharge in each game purchase to have that AI available? Given what we've read about the number of hardcore fans and the number of casual players, I'm not sure how profitable it would be to charge all purchasers for the HIRO AI?
- What would be the impact on fun for the human players by using such an AI as an opponent? This logic would tend to pursue a path it believes to be optimal to pursue certain victory conditions. Might its metanarrative "always go for space", "always go for XXX" become predictable? How would difficulty levels be implemented? What happens when 2, 3, 4 HIRO AIs all start on the same landmass?
- If the logic is implemented on the player's PC, what are the HW requirements for that PC? How would that be different from the current GPU requirements for gaming PCs today?

 

[CppMaster]

Never got any reply to this link although it's the third time I mention it in three different posts over the last months!

Ok, let me reply myself with a question. Is this too disturbing, or underwhelming, or off-topic, or hard evidence that yes, machine-learning is possible and affordable for Civ6? Any expert opinion here?

ML Engineer here. Yes, it's possible, but it'd be very hard to do, so it might not be affordable. Also it'd require much more computing power than just algorithm.

 

[rgp151]

Prepare to be shocked! The market for such a game would be very small, most people regard the AIs are there to provide flavour.

I just don't agree. I think most people would prefer better AI logic in place of bonus resources for the AI. Bonus resources feels more unfair than an AI that makes effective decisions. And again, all of this can easily be controlled with setting anyway.