Let's discuss AI

Horizons said:

I remember the original design goal with Civ5 was to make a competitive AI that realized it was playing a game and tried to win. But there was an ENORMOUS furore on here about lack of diplomacy?

Click to expand...

I just finished an Emperor level game as Morocco and beat Sejong in a Space Victory by turn 350s with 0 battles or wars against any major Civs because of my extremely high diplo status.

I had DoFs with India / Songhai / Sweden / USA / Korea / Brazil throughout the entire game and everyone hated the Huns.

Askia wanted to backstab me numerous times but I'm guessing he didn't want to face the diplomatic consequences. I had trade links with basically everyone and the USA / Brazil were following my religion. I didn't even build up a large army because I was too concerned with building enough Tourism to fend off Brazil's influence.

I won by 1 spaceship part in the end because pretty much everyone friended another and traded RAs. So there was some diplomacy going on in my game but the AI didn't play to win. I thought ideological differences would create some hostility, but it never materialized

In a sense there is diplomacy, but it was a very passive, boring game and I'm looking forward to the fall patch.

Dux1 said:

No, the AI doesn't care if you're winning. It will try to win itself, but if you are 10 turns away from winning a landslide World Leader vote, for example, the AI doesn't make any attempt to prevent this from happening. It just goes about trying to pursue victory on its own terms, ignoring the fact that it's own victory plans couldn't possibly come together until after the player has already won the game.

I would much rather the AI try to prevent the player from winning. In the above example, the AI should be trying to buy off or destroy City-State allies of the player. If the player were on the verge of cultural victory, the AI should attack the player's cultural centers. Etc., etc.

Click to expand...

This matches my experience, at least when it comes to diplomatic victories. In G&K, the AI would actively contest my pathway towards a diplomatic victory by buying off my city-state allies or conquering them. In BNW, I always coast to a diplomatic victory without spies from this or that civilization rigging elections in my city state allies, or having my allies be invaded, or having them be bought off.

BNW is a great improvement over G&K in many ways, but the above is one of a few examples in which BNW has taken a step backwards.

I am currently watching the lecture now and 14:45 mins into the video, Soren mentions the Turing test. More details found a few minutes later. Should it matter whether civ 5 passes or fails the Turing test? If you're playing single player, this is irrelevant. If you're playing multiplayer and you simulated this test with one other player and one AI, does it still matter? The best simulation I can think of is have one other player choose beteen Monty and Shaka and the AI chooses the other person. Player changes name to their character and has to act aggressively. Then there is environment. Add several other AI's that interrogator knows are AI? Or just have a small controlled environment?

The next thing, at 15:20 I thought Soren may have either oversimplified the "playing to win" and "playing to lose" concepts in fun and good AI or mixed the categories up. It depends on the game itself. So in Sims, he says the AI is fun, but the aim of Sims isn't to win or lose.

35~mins - civ 4 tech trading problem. Hah! No wonder I hate civ 4.

36~ mins - the matter of cheating. I think this is what we should be mainly discussing. I now understand cheating for AI was necessary: namely, to counter us cheating in very simple ways such as reloading. AI has to provide a challenge but not look like cheating.

Vanilla actually did include a "they THINK we are trying to win the game they same way they are (and they don't like it)". What people hated most about it was often times the human hadn't even decided how he was going to try to win. That negative diplomatic modifier was removed with G&K.

All this is very disappointing. I was ready to make the switch from IV to V, but I think I'll wait until the patch to see if the aggression is fixed. I cannot stand the thought of going through the early stages of the game without having to field a defense vs other civs. Makes the opening very monotonous.

I think I'll definitely love V once I get into it, but i'll hold off for its improvements to give it a fair shake. After all, I love IV, but picked up the complete edition a couple years after it had been all patched up to BTS 3.19.

Ankh said:

Is it because the technology isn't there yet, or is it because game developers don't want to devote more resources to improve it when it's good enough to beat "most people"?

Click to expand...

The answer is in bold. It's not that the AI is "good enough to beat most people" either. It's just so random at deity that it's either really hard or easy, it depends on to many variables and the AI's consistent lack of a desire to win but instead just ignore all game mechanics and rush science and dominate you rather than try and beat the other civs.

In 1997 over 10 years ago a game called Age of Empires had an AI that was 10x better than this one and for over 20 years computers have been able to slaughter even the best human chess player 90% o the time via STRATEGY not cheating.

Fraxis obviously made an executive decision to skimp on the AI funding for civ V. From what I understand even previous civs had better AI.

Chess and Civ are not comparable because Chess is inherently solvable. I'm unaware if that has actually happened (I know Checkers has been solved) but one could conceivably create an AI that can counter every possible move in the game without error. Chess game developers have to program mistakes into the AI behavior to keep the game winnable.

Civ isn't like that because of the wide number of randomized pieces for the AI to figure out. The board in Civ is different every time and while a human player can understand it intuitively intuition is not and will not be in the foreseeable future a tool that a computer can utilize.

OldBoy318 said:

I just finished an Emperor level game as Morocco and beat Sejong in a Space Victory by turn 350s with 0 battles or wars against any major Civs because of my extremely high diplo status.

I had DoFs with India / Songhai / Sweden / USA / Korea / Brazil throughout the entire game and everyone hated the Huns.

Askia wanted to backstab me numerous times but I'm guessing he didn't want to face the diplomatic consequences. I had trade links with basically everyone and the USA / Brazil were following my religion. I didn't even build up a large army because I was too concerned with building enough Tourism to fend off Brazil's influence.

I won by 1 spaceship part in the end because pretty much everyone friended another and traded RAs. So there was some diplomacy going on in my game but the AI didn't play to win. I thought ideological differences would create some hostility, but it never materialized

In a sense there is diplomacy, but it was a very passive, boring game and I'm looking forward to the fall patch.

Click to expand...

Not sure what your complaint here is, you barely won it sounds like a close fun game....

Have you tried on deity yet?

Arachnofiend said:

Chess and Civ are not comparable because Chess is inherently solvable. I'm unaware if that has actually happened (I know Checkers has been solved) but one could conceivably create an AI that can counter every possible move in the game without error. Chess game developers have to program mistakes into the AI behavior to keep the game winnable.

Civ isn't like that because of the wide number of randomized pieces for the AI to figure out. The board in Civ is different every time and while a human player can understand it intuitively intuition is not and will not be in the foreseeable future a tool that a computer can utilize.

Click to expand...

I see what you mean sortof, but civ has certain rules just as chess does given there are many many more rules but I believe it isdoable, just as flight simulator AI can act human with multiple rule sets, I guess it's the random aspect that makes it the most different from chess.

That does not explain how Microsoft can publish a TBS game over 10 years ago with a much better AI than civ though.

Uh, you're talking about Age of Empires still, right? It's been a very long time since I last played that but if I recall it is a real time strategy game, not turn based. Entirely different beast than Civ. Dexters addressed how the same arguments don't really work last page (he was talking about Total War but it applies to AoE too).

Ankh said:

Let's take a look at the global steam achievement to beat King difficulty:
The Once and Future King: Beat the game on the King difficulty level - 5.6%.

Click to expand...

Well I'm probably in the minority, but I have dozens of wins on immortal and deity, yet I don't have this achievement. I've simply never played on king.

I'm sure a lot of civ 4 veterans went straight to the higher levels since civ 5 is pretty easy in comparison.

Espírito;12651912 said:

So what you're saying is that by making Civilization more viable as a franchise (starting with C-IV it was intended to appeal to a broader range of players) they are able to make the AI worse. Looks like those of us who want a good AI have been screwed by the market.

Click to expand...

With my 20+ years of gaming analysis this sort of behavior also signals the death of franchises as it alienates the dedicated fans in favor of some fly by nights.

Arachnofiend said:

Uh, you're talking about Age of Empires still, right? It's been a very long time since I last played that but if I recall it is a real time strategy game, not turn based. Entirely different beast than Civ. Dexters addressed how the same arguments don't really work last page (he was talking about Total War but it applies to AoE too).

Click to expand...

This is inaccurate, the exact same rules apply on every front except time.

elijahfire said:

In 1997 over 10 years ago a game called Age of Empires had an AI that was 10x better than this one and for over 20 years computers have been able to slaughter even the best human chess player 90% o the time via STRATEGY not cheating.

Click to expand...

That's complete, complete bull. The AI in AoE was nowhere near that good. I used to set up single player skirmishes where I played against two AIs on the hardest level in order to have a challenge. I wasn't even very good at the game, I had friends who beat me in multiplayer quite thoroughly.

This type of lying about past games is what's making Civ V look bad. Yes, the AI in Civ V is not as good as the AI in your lies. But then, that would be impossible!

And no, you are not right that things aren't different for a real time game. Every RTS that I can remember, AoE included, has always relied on the same things: spam units and move them fast with an inhuman ability to micromanage movement. Even so, all of them are beatable by an average level player. And all RTSs, including AoE, have very simple requirements for the AIs beyond combat, because the economic system is not as sophisticated and there's no diplomacy.

Ankh said:

Some people have already touched on the topic: Do we want a good AI or a fun AI?

Today many people don't want a cutthroat AI that plays to win, especially in a sandbox game like civ. Actually, game design nowadays leans more towards always letting the player win, otherwise he will get bored and play something else. Gone are the days when games were made to challenge the player (think old school arcade shooters). Today, the focus is on a "fun" AI that doesn't actually play the game but is there as an entertainer for the player. This is another reason why the AI isn't improving.

Click to expand...

A game of Civ is far too long to be just about seeing the victory screen. The way there is where the fun is, and when I feel like I am getting a piggybackride to the victory screen, I'll just stop playing. Just as I stopped my science victory game because no one would DoW me, even though I had 2 or 3 units only and when Shaka finally did declare, he decided to take on the city state that I was allied with on my small continent first before moving for my capital. By that time, I had enough ranged units built up to wipe out his tech-equal forces

He also failed to take the city state with his 20 units.

KillingMeSoftly said:

This is a semi-new development since G&K expansion. In vanilla, the AI would actively hate you for trying to win the game. A lot of people complained about this and it was removed. Thus now the AI doesn't really mind at all if you're going to win.

Click to expand...

If people want to play a game of Civ with n bystanders just so it's not too boring as they progress towards victory, they can play Chieftain.

The AI no longer trying to prevent your victory on higher difficulties is really, really bad. It appears the WoW-ing of this game has arrived.

Any aggressive warmonger leader should DoW you and try to prevent your victory if you're getting too close. I can live with Ghandi being satisfied of surviving the game but Attila or Napoleon should raise hell.

This game has a total of 7 different difficulties (not counting settler because that's basically the tutorial game where you can't ever be DoW ad will win every war because the AI is set to only build 1 unit), even if you're playing much worse than the computer, you should still be able to win on Chieftain. Instead, those who actually understand the game and play it well have to play it on the absurd cheating levels of immortal / diety because the AI needs massive, massive economic advantages to overcome its incompetence.


and all of that wouldn't be that big a deal if multiplayer was actually playable :/

The QC said:

That's complete, complete bull. The AI in AoE was nowhere near that good. I used to set up single player skirmishes where I played against two AIs on the hardest level in order to have a challenge. I wasn't even very good at the game, I had friends who beat me in multiplayer quite thoroughly.

Click to expand...

That's your poor opinion, I never said it was good, I said it was 10x better than civ V which is quantitatively true in all aspects.

The QC said:

This type of lying about past games is what's making Civ V look bad. Yes, the AI in Civ V is not as good as the AI in your lies. But then, that would be impossible!

Click to expand...

umm huh?

The QC said:

And no, you are not right that things aren't different for a real time game. Every RTS that I can remember, AoE included, has always relied on the same things: spam units and move them fast with an inhuman ability to micromanage movement. Even so, all of them are beatable by an average level player. And all RTSs, including AoE, have very simple requirements for the AIs beyond combat, because the economic system is not as sophisticated and there's no diplomacy.

Click to expand...

Learn to articulate your fan boi rage a bit better and get back to me.

Quineloe said:

Instead, those who actually understand the game and play it well have to play it on the absurd cheating levels of immortal / diety because the AI needs massive, massive economic advantages to overcome its incompetence.

Click to expand...

^ this.

elijahfire said:

That's your poor opinion, I never said it was good, I said it was 10x better than civ V which is quantitatively true in all aspects.

Click to expand...

I have to agree with The QC, you're both lying and dreaming, or the other way around. And even if you were not, you really cannot compare TBS and RTS in that matter, as it's been explained before. In RTS you don't need to have a real smart AI as it can compensate by speed (where you have no chance to beat it). And come on in AoE, you don't have as many variables as in Civ (far from that), it's also much easier to program.

Kordanor posted a great podcast with the developers on the topic of aggressive AI.

http://www.idlethumbs.net/3ma/episodes/firaxis-revisionists (start at 15 mins)


They basically said that they did nerf the early game aggressiveness from the AI because of the new economic system. They proposed that the early rushes would cripple the AI. They used the term "throttled down" when it comes to aggressiveness. Thoughts?

alpha wolf 64 said:

The AI also has no concept of defense. In a war using infantry and artillery, I gave the defending AI 3 x-com. All 3 were dead in 3 turns because they attacked until they committed suicide. Also, the attackers are failing to take a city with 0 hit points left because they wasted all of their melee units attacking the city with no hope of taking it that turn, instead of allowing their ranged units time to weaken the defenses.

Click to expand...

This. I wish the AI could be taught to play defensively. Better yet ... when at war could the AI be programmed to either play At War Offensively where it is trying to invade your land and take your cities, or play at War Defensively where it takes up defensive positions and is only trying to thwart your attack. Even if the AI does not succeed it would at least make it tougher for the human opponent and make you suffer as many casualties as possible increasing the opportunity cost of going to war. This to me would be a major improvement.

I thought I read somewhere that the problem they had with the size of the map and 1UPT in particular with CS is that the units would occupy every space so they had to make them move or shuffle around so the map would not get gridlocked. Isn't this why we see some of the bad AI moves that take them out of the position or end up with a ranged next to a melee where they should just fortify up?