The AI will never be fixed, so stop hoping... :(

[CivFart]

I don't think the AI will suddenly become a genius after the next patch (as it never was in the civ series), but I expect slow and steady progress in patches and expansions.

Sid Meier have specifically outspoken his philosophy about the AI in civilization. In his own experience, in all the feedback he had over the years and in all his knowledge he have come to the conclusion that people dont want to have a smart AI, even if they say so. They want an ai that confirms what they are doing, not an ai that is smarter than the player. A player dont like feeling like the less smart guy.

So according to Sid Meier his philosophy is that the AI shouldnt be outsmarting the player he should rather be reflecting what the player is doing, give the player confirmation of the situation and the player will be satisfied.

This is based on sid's experience and feedback over the years. It works.
(Even if people complain and say they want a genius ai, they dont really mean it, according to Sid)

 

[Slowpoke]

It's sadly true really. But people really do want a mildly competent A.I. at the very least. I mean I can beat an advance wars A.i. with very high odd against it, but at least it has the sense to put infantry in front of artillery.

 

[CivFart]

We can agree that this AI is not finished, thats a whole different story, an unfinished (and buggy) AI vs a bad AI are two different stories.

 

[PieceOfMind]

It's sadly true really. But people really do want a mildly competent A.I. at the very least. I mean I can beat an advance wars A.i. with very high odd against it, but at least it has the sense to put infantry in front of artillery.

Ah, but you can spin those words to Meier's intention anyway. If an AI is at least mildly competent in the tactical military minigame, then it allows the tactics used by the player to overcome that AI to have meaning. Confirm the player's superiority by putting up somewhat of a fight.
If all a player has to do is build four horsemen and go to town, attacking and retreating constantly, the experience is unrewarding even if the player does try a more interesting tactical approach at another time. Essentially people want tactics to be necessary, not voluntary.

 

[Zydor]

The role of an AI in Civ is not to win a battle, win a war, or even display any kind of tactical or strategic competance. Its role is to play its part in giving the player an enjoyable experience.

As the difficulty level rises, so does the difficulty in getting the AI to carry out that role. The tactical and strategic options that present themselves and considered by a human player rise exponentially at a subtantial rate, the higher the difficuilty level. They have to, else bordom sets in.

The grading of the AI response to the player at each level is therefore massively complex. An AI dependent on scripted logic will never carry out that role in a Strategy Game such as Civ - not in a million years. So it needs to have deductive logic programmed in to make its own decisions. Few programming tasks are more difficult .... but are much enhanced when players point out glitches/horrors over time, that can be used to fine tune the AI responses by patch et al.

I sometimes think we expect too much of AIs, we tend to simplify the task of getting them to work. Although undoubtedly artillery moving forward without Infantry cover can legitamately be described, at best, as "unwise"

So will it be fixed, in a purist military tactical & strategic sense I doubt it, not in our lifetimes anyway, the task is far too complex, the human brain will in the end run rings round it no matter how clever the pure logic is. We can all help by constructively pointing out the glaring horrors that may surface so they can try to plug holes, work out balance consquencies yaddie yadda. Not much else we or they can do.

The current AI will undoubtedly improve, as Civ 5 in its make up and game flow settles. AIs will always be behind the curve when rapid development change occurs because until all change is complete - or at best very near complete - programming a proper AI response is nigh on impossible. To say there is change in Civ 5 is the understatement of the Decade, the game was far from finished on first release, so the AI is undoubtedly going to be wanting for a while yet with the Guargantuan "Patch" thats on it way.

Regards
Zy

 

[anti_strunt]

Folks, it's entirely possible to give computers problems that are too complex to solve in a reasonable time. There isn't a good computer Go program, or computer Diplomacy program, and both of those have simple rules and maps.

Well, Diplomacy also has the whole "social/diplomatic interaction" thing, pretty hard to properly encode that sort of thing...

 

[freeluos]

Before improving the tactical AI, improve pathfinding and formations. Look at supreme commander (for those that played it) in the beginning it was a mess. But after introducing proper formations the AI became a lot more formidable.

But SC had some luck. Units could "clip". Civ units are bound to pretty large hexes. Clipping is not recommended. I fear for a challenging AI just because of this. Often there is just not enough room.

As far as I have read/seen about civ 5 AI they can't make it worse. That's a good start.

 

[scratchthepitch]

Sid Meier have specifically outspoken his philosophy about the AI in civilization. In his own experience, in all the feedback he had over the years and in all his knowledge he have come to the conclusion that people dont want to have a smart AI, even if they say so. They want an ai that confirms what they are doing, not an ai that is smarter than the player. A player dont like feeling like the less smart guy.

So according to Sid Meier his philosophy is that the AI shouldnt be outsmarting the player he should rather be reflecting what the player is doing, give the player confirmation of the situation and the player will be satisfied.

This is based on sid's experience and feedback over the years. It works.
(Even if people complain and say they want a genius ai, they dont really mean it, according to Sid)

Since the Civ games have been using the same basic AI throughout and this was developed early, before home computers were very powerful, I strongly suspect Sid is doing a bit of marketing-speak there. I don't buy a word of it. To significantly improve the AI, to bring it up to speed to match modern computing, would entail a complete redesign. They are obviously not willing to do that, and rely upon the much less expensive route of patching tweeks into the old dinosaur. But you cant tell your customers that, you need to spin it so they think an obsolete AI is really what the customer wants. If what Sid claims is actually true, there would be nothing stopping them having a dumb AI at low play levels and gradually having this AI employ smarter routines at the higher play levels. Others do this, and quite successfully.

 

[PieceOfMind]

Before improving the tactical AI, improve pathfinding and formations. Look at supreme commander (for those that played it) in the beginning it was a mess. But after introducing proper formations the AI became a lot more formidable.

But SC had some luck. Units could "clip". Civ units are bound to pretty large hexes. Clipping is not recommended. I fear for a challenging AI just because of this. Often there is just not enough room.

As far as I have read/seen about civ 5 AI they can't make it worse. That's a good start.

I'm unsure what you mean by 'clipping'. Is it developer speak or gamer speak?

 

[JLoZeppeli]

Ah, but you can spin those words to Meier's intention anyway. If an AI is at least mildly competent in the tactical military minigame, then it allows the tactics used by the player to overcome that AI to have meaning. Confirm the player's superiority by putting up somewhat of a fight.
If all a player has to do is build four horsemen and go to town, attacking and retreating constantly, the experience is unrewarding even if the player does try a more interesting tactical approach at another time. Essentially people want tactics to be necessary, not voluntary.

Ok: don't play tactical wargame, if you are a Sid follower....

Boys Civ V AI is awful, maybe they can improve it a little, but after they had delivered the game we have i think that is impossible they can fix it properly....

 

[freeluos]

I'm unsure what you mean by 'clipping'. Is it developer speak or gamer speak?

In the wiki link it is well explained. Although it is mainly a graphical explanation there it often is also a pathfinding problem (no overlapping units/polygons means more complex pathfinding algoritms in order to allow smooth unit movement). Civ V is not an rts but it can be compared. No clipping in an rts would have a great impact on performance and unit behaviour. And in most cases it is not be implemented well (units that get stuck, take too long to reach a destination). And often it is much worse if they need to move in a formation (which is essential for CIV V for a good war AI).

Before I forget:
http://en.wikipedia.org/wiki/Clipping_(computer_graphics)

But luckily you don't have to be very smart to fix warriors running in fortified positions triple times stronger. Bad thing is they need backup...

Edit: after reading the wiki link above the proper term should be collision
http://en.wikipedia.org/wiki/Collision_detection

 

[Öjevind Lång]

Or you cheat and give them +1 or +2 line of sight.




I doubt the expansion packs will sell very well if they don't have AI upgrades.

They will, according to the update information. The information about all the new stuff is very long by now, but it's there among the "various tweaks and fixes".

I can't wait for the patch. But I will rather than get a broken patch like the infamous second-to-last one in BtS.

 

[PieceOfMind]

In the wiki link it is well explained. Although it is mainly a graphical explanation there it often is also a pathfinding problem (no overlapping units/polygons means more complex pathfinding algoritms in order to allow smooth unit movement). Civ V is not an rts but it can be compared. No clipping in an rts would have a great impact on performance and unit behaviour. And in most cases it is not be implemented well (units that get stuck, take too long to reach a destination). And often it is much worse if they need to move in a formation (which is essential for CIV V for a good war AI).

Before I forget:
http://en.wikipedia.org/wiki/Clipping_(computer_graphics)

But luckily you don't have to be very smart to fix warriors running in fortified positions triple times stronger. Bad thing is they need backup...

Edit: after reading the wiki link above the proper term should be collision
http://en.wikipedia.org/wiki/Collision_detection

Ok, since I see you now meant to speak of collision detection, I would guess you are interested (perhaps already following?) the pre-production news on Kings and Castles. In it, and in SC2 as well, a much speedier pathfinding algorithm is used. I think they call it flowfield, but the point is that units are permitted to get closer together for a short time, to navigate obstacles while keeping formation for example. Fewer headaches with units bumping into each other (extremely annoying in SC1).

I'm even more confused when you speak of 'no clipping' as that is even different again. The wiki page you provided mentions that 'no clipping' is a different concept to the clipping it describes. Indeed, 'no clipping' is a more familiar concept to most gamers.

Honestly though, I don't see how collision detection is relevant to civ5. Were you making a general analogy? Getting units to maintain a formation despite meeting obstacles (including terrain and other units) is a tricky task for sure, and I don't understand what you're suggesting to do to improve it.

 

[Mac2411]

They will, according to the update information. The information about all the new stuff is very long by now, but it's there among the "various tweaks and fixes".

I can't wait for the patch. But I will rather than get a broken patch like the infamous second-to-last one in BtS.

Are you saying that they are going to give the AI a default sight bonus? While it might make the AI more competent, I don't like that idea at all. For one, it greatly diminishes the value of the American UA. Secondly, it is just another layer of "give the AI more bonuses" to make it better, which I don't like to begin with. This is just a sign of capitulation that the AI can't be actually improved in my opinion. That makes me think that what some of the posters in this thread are saying regarding the same basic AI still being used from Civ I days is true.

 

[Öjevind Lång]

Are you saying that they are going to give the AI a default sight bonus? While it might make the AI more competent, I don't like that idea at all. For one, it greatly diminishes the value of the American UA. Secondly, it is just another layer of "give the AI more bonuses" to make it better, which I don't like to begin with. This is just a sign of capitulation that the AI can't be actually improved in my opinion. That makes me think that what some of the posters in this thread are saying regarding the same basic AI still being used from Civ I days is true.

As I read it, they are working hard at making the AI smarter both when it comes to warfare and to diplomacy. They are also going to nerf the Horseman unit.

 

[man-erg]

Civ, is more complex then chess by a huge order. The two big problems is the number of varibles:

I'm not so sure that chess really *is* that much simpler. Could it just be that the basic rules have had only minor amendments over centuries, when with Civ, they throw the old one out and start again with each version? If chess was invented today, the AI would be no less difficult than for Civ5. But for chess, we have centuries of books about strategy, opening moves etc. Imagine if that existed for Civ?

Personally, I think starting again each time is a flawed design strategy. In contrast, I'm very much enjoying Europa Universalis 3 with it's 3 full explansions, several years of AI improvements - and another huge expansion coming soon! And EU3 on release was a lot worse than Civ5 on release....

The underlying question is "What is a sequel for?". A total re-invention of a series or an organic improvement? Why did Civ5 need to be created anyway? I believe the answers are down to marketing and business, rather than technical. Hence any debate about the AI's technical capabilities - and there are many interesting comments here- is just irrelevant anyway. With organic improvement, the AI can improve too. However Civ5 is probably the biggest single rejection of previous conceprs in the series so far. A big risk, but perhaps with the right marketing and names behind it, it could never fail anyway?

If only 1% of the effort put into graphical baubles was put into AI. Compare the thousands of hours spent by huge teams of graphic and audio designers with that spent by a couple of devs on the AI.

I worked in games development many years ago, and was told the same thing as people are saying here, that "gamers don't want an invincible AI that crushes them". But they don't want one that rolls over and lets them tickle it's belly, either!!! And that is the current state of Civ5 AI. The aim is to make the gamer *think* that they are some kind of tactical genius to win, so the AI has to put up a fight, before bowing to the intellect of the Mighty Human. Sure, the AI 'brain' will never be as sophisticated as the humans, but this is easily balanced by the fact that the whole game is in *its* memory, not yours! It knows where you are and what you do at exactly the moment you do it. It especially applies to FPS, as it's actually easier to code enemies that instantly home in on the human every time, than those that behave more 'life like' and make mistakes. The former would result in nobody ever playing a game for more than 10 mins though, before quitting in frustation.

They need to get rid of the obvious AI flaws in Civ5 (but no more than that) and I'm sure they will do that as people keep saying, there are some very trivial things that would make huge improvements.

 

[Zydor]

..... Personally, I think starting again each time is a flawed design strategy .....

Gets my vote ..... "not invented here syndrome" is the cause of many screaming disasters.

Civ Franchise has been hugely successful, and the "complaints" on Version changes have been within expected bounds. Until Civ 5. The complaints this time are on an unprecedented scale, far far beyond "new version syndrome". Its no co-incidence it happened when wholesale changes were made to the very core fabric of the Franchise, the latter had never happened before - previously, the core Franchise ethos and make up had been incrementaly enhanced - unlike this time when it was wholesale destruction of what makes up the very core essence of Civ.

I think most traditional fans are waiting for this upcoming Patch, at present its a kind of "the jury is still out" feeling. Fingers & toes crossed, but I still have a gut feeling that the Franchise we've known for 25 years is about to go down the pan; and some wacky reincarnation, unrecognisable from the Franchise we knew, is about to surface and borrow the Civilisation Franchise name as a marketing label.

Regards
Zy

 

[ohioastronomy]

It's hard to develop a good AI for a fatally flawed game design, and I think that the lack of stacking qualifies as fatally flawed.

 

[pazmacats]

Panzer General only worked as a fun game because the player was THINKING he played vs the AI, where he was playing against a pre-designed setup.
That's why playing Warcraft of Starcraft is pointless in single player after completing the campaign.

Civ5 designer's didn't realize that the only challenge in a single player game is
a) The map setup (it's like a puzzle basically)
or
b) the overwhelming odds (e.g. space invaders/tetris or Civ1)
or
c) mini games (which are mini a's or mini b's)

Nobody will understand what I am trying to say, but nevermind, it's just a computer game.

====
bottom line: They fkked up multiplayer. THAT's where they really failed. It's the year 2010 afterall...

 

[freeluos]

Ok, since I see you now meant to speak of collision detection, I would guess you are interested (perhaps already following?) the pre-production news on Kings and Castles. In it, and in SC2 as well, a much speedier pathfinding algorithm is used. I think they call it flowfield, but the point is that units are permitted to get closer together for a short time, to navigate obstacles while keeping formation for example. Fewer headaches with units bumping into each other (extremely annoying in SC1).

Well I am not really into it yet. When the game is finished I will take a look at it. But it's good that smart cpu friendly pathfinding is mainstream in RTS.

I'm even more confused when you speak of 'no clipping' as that is even different again. The wiki page you provided mentions that 'no clipping' is a different concept to the clipping it describes. Indeed, 'no clipping' is a more familiar concept to most gamers.

Sorry, forgot to update that. But what I meant was when collission detection is applied performance will drop but the behaviour of the units would be much more logical.

Honestly though, I don't see how collision detection is relevant to civ5. Were you making a general analogy? Getting units to maintain a formation despite meeting obstacles (including terrain and other units) is a tricky task for sure, and I don't understand what you're suggesting to do to improve it.

If the computer calculates the path of the units it can determine the best formation. May be even with empty hexes between the units during the travel to its destination. Freedom of movement and placement

Another solution could be to devide a hex into smaller hexes. Normally there would be only 1 unit per big hex. But under circumstances units could make use of those parts of the hex. Next move units are forced to move aside to create the space that was nescessary. Thus maintaining 1upt but more flexible. I believe I am not the first mentioning this.

Less graphical power but more energy towards the basic functions of the game. But again, I think people care more for graphics. That sells is what sells the game nowdays...

I hope I have clarified myself and my ideas about unit AI/behaviour