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

[KingMorgan]

Nice of the OP to tell us to abandon all hope.

The AI can, and should be fixed. There is a minimum level of programing and Civ 5 does not appear to have met it. The current issues are an insult to fans, Firaxis (at best) have not delivered value for money, at worst they are fraudsters.

 

[LordTC]

Nice of the OP to tell us to abandon all hope.

The AI can, and should be fixed. There is a minimum level of programing and Civ 5 does not appear to have met it. The current issues are an insult to fans, Firaxis (at best) have not delivered value for money, at worst they are fraudsters.

The real problem is that they've made combat several orders of magnitude more complicated, and the AI is going to be hopeless at it. Combat was far easier in previous versions of civ, and the AI was still pretty bad, making it an even harder problem for the AI is the wrong way to go if you want the game to be more interesting.

 

[KingMorgan]

The real problem is that they've made combat several orders of magnitude more complicated, and the AI is going to be hopeless at it. Combat was far easier in previous versions of civ, and the AI was still pretty bad, making it an even harder problem for the AI is the wrong way to go if you want the game to be more interesting.

I'm reading your reply as:

Firaxis desgined gameplay that cannot be achieved by the programmed sub routines.

Are your sure this is what you mean? Has Firaxis desgined a game that is impossible for the developers to code? Or are the devs incompetant? Which ever way I interpt your reply it seems some department of Firaxis is at fault.


My view point is that if you can travel to the moon on the processing power of a modern scientifc calculator, then making an AI that is fit for purpose using the power of an average computer should be no hard task. Assuming that is your prepared to allocate a sufficent amount of compentant resources.

 

[elthrasher]

This is my biggest problem with the AI, which should be very easy to fix. The pre-attack calculation is already there, the AI should know when an attack is a horrible idea.

I've wondered about this myself as I see 2hp longswords suicide themselves into my fortified samurai. Could it be there's some kind of "well, you're gonna die anyway" clause that tells the AI to suicide units that can't really get to safety?

The worst is not taking the killshot. Same game I had to end my turn with a 1hp samurai right next to a city. He's dead, right? No. The city bombarded my full health guy. Duh. This is deity, by the way.

And one more. Same game. I surrounded a city that had a fully-healed longsword in it. I didn't have enough juice to get the city in one turn, so I had to end my turn with two lightly-injured samurai on open terrain. The longsword stayed put and was destroyed when I took the city with a single attack next turn. Sure, he was dead and I'd have had the city anyway, but he could've taken a samurai with him.

So is there any rhyme or reason to this or does the AI just not know what the hell to do?

 

[KingMorgan]

I've seen 1hp warriors go for LS fortified on a forest hill numerous times, AI cities bombarding full health units whilst other 1hp units are ignored.

Your right when you say the AI decision algorthyms are FUBAR.

 

[LordTC]

I'm reading your reply as:

Firaxis desgined gameplay that cannot be achieved by the programmed sub routines.

Are your sure this is what you mean? Has Firaxis desgined a game that is impossible for the developers to code? Or are the devs incompetant? Which ever way I interpt your reply it seems some department of Firaxis is at fault.


My view point is that if you can travel to the moon on the processing power of a modern scientific calculator, then making an AI that is fit for purpose using the power of an average computer should be no hard task. Assuming that is your prepared to allocate a sufficent amount of compentant resources.

I'm not disagreeing that its the developers fault, I'm just saying its the design of the game that makes the AI bad at combat (military decisions are too difficult in terms of unit placement to do on the amount of resources available). Your moon analogy is seriously flawed as the processing power was used to operate equipment and take measurements not to do powerful reasoning tasks, which we used humans for. If the equipment NASA used to land on the moon had to figure out to do so itself and could then we would have passed the Turing Test by now, as it is we are nowhere close, and most of AI is more interested in some form of the Limited Turing Test.

 

[KingMorgan]

You may be right, the analogy is probally unsuitable.

Generally speaking I think it illustrates my point that with the modern processing power of computers it not too much to ask or to expect that Firaxis with more care/resources could have done a better job. Certainly for Civ 4 BTS, Firaxis hired a member of the Modding community to improve the AI after he proved a better developer (without the experience that Firs employees had)

Bottom line is Civ 5 AI needs alot of work.

 

[Paeanblack]

Yeah the bonuses on deity are pretty small comparatively. Like I said before, I'm really surprised the mod I expected to do well the most, extra difficulties, nobody really downloads. Doesn't steamrolling deity for the 50th time in a row get boring?

Lately I've been playing with map size, map type, and civ all set to random, and it's actually quite fun. You can roll some pretty harsh starts compared to the Deity AI, and occasionally I lose. I haven't yet rolled a start that I think would be unwinnable if played a second time, but the first time through without knowledge of the map specs is enjoyable. I also avoid scouting in a manner that is specifically designed to reveal the map size.

 

[bryanw1995]

just read patch notes, the AI appears to have gotten a LOT of attention. wonder if the OP will come back and eat his words...

 

[Mutineer]

Processor power for brute force forced combat calculation to dept. of 3 or 5 is not that hard on contemporary hardware.

But system (terrain/unit information) had to be presented in a way that is fast accessible, ie.. should be well design form ground up.


Devide big map on local combat theaters as high end function is not that hard either, in other to speed up computation. But again, data has to be design for that. Basically you write AI first, then put additiona stuff around it, not other way around.

 

[troytheface]

lol

if i read the op correctly then the AI is broken and they have to find someone to program it and Civ5 is not chess

these all culminate into "the AI won't be fixed"

In essence the op is magic and can see the future and knows about how to program games

doubtful

 

[scratchthepitch]

The developers of Civ have essentially been using the same AI since Civ1. This is why the same lame problems keep appearing in each version.

 

[Tabarnak]

The developers of Civ have essentially been using the same AI since Civ1. This is why the same lame problems keep appearing in each version.

..... and the only way to boost AI is to give them cheats. Deep Blue (or Deeper Blue) should be imported.

 

[PieceOfMind]

Moderator Action: Moved to General Discussions

(it doesn't fit in Strategy and Tips)

 

[ImperialGuard]

for Civ5 they announced how they split the AI into three "contexts" (I'm sure there is a more suitable word).

Tactical, ???, Strategic.

This seemed like it was an evolution of the AI in Civ, and would help to improve the AI in all game terms.

The game was rushed out the door, but I think with subsequent patches, the AI will improve.

Will it be as smart as a person? Never. Can some of the "low hanging fruit" be improved? Definitely.

I am waiting for them to release the core SDK, so people can "get under the hood".

I've already seen some good discussions on how to improve the AI (there was a really good, thoughout article by Bibor on how the AI could better determine defensive positions). Alot of smart people on the forums ... there was a guy who wrote an AI improvement mod for Civ4 that was so good it got incorporated into one of the patches. So who knows what the future will hold?

 

[scratchthepitch]

..... and the only way to boost AI is to give them cheats. Deep Blue (or Deeper Blue) should be imported.

Never heard of those. Are they AI programs?

 

[ohioastronomy]

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.

It also doesn't help that the game is at the wrong scale (too few hexes for the combat system to work well) and that the lack of stacking makes pathing terrible.

It isn't magic to say that it's not feasible to change this, it is common sense. That's one game design rule that they apparently forgot: a clever idea in a computer game is only a good idea if the computer knows what to do with it.

 

[CivFart]

meant to post this in general.

I really don't think they can fix the AI. It would take an enourmous effect to do it right. The first problem I see if finding the right people to actually do the job we want. In the world of Computer Science AI is still among those problems that have moved relatively slowly compared to other fields, I mean, compare the AI to any of ther other Civs, are the leaps in those fields comparable to the other sections of the game.

Civ, is more complex then chess by a huge order. The two big problems is the number of varibles:
The number of different units, their abilites, their use in combination with other units and how they should be used in their enviroment.

Second, in chess you have perfect information, this is not the case in Civ. To compound things the AI seems to have no memory as seen in a post of PlayWithMe (Rome). The AI doesn't know that your units were on the hill 1 turn ago if they lost vision of them. So, you implement memory and then prediction properly.

Given how simple Chess is compared to Civ and the time its taken for the AI to provide a challenge that we can expect from a human player its not going to happen for Civ.

Sure there are seemingly simple things that could be done to improve the AI:
1 Use your Generals... god. Its that hard. Increase the spawn rate of the generals for the AI on harder difficulties in additon or in exchange for the amount of units they can produce. Reducing the number of units the AI has will probably make the game go faster as well.
1b If your going to lose your City its ok for your General to GTFO.

2 We get numbers that say, hey, you will die if you attack. I would suggest the AI use these numbers.

3 Build proper terrain upgrades, seriously, even late game, I take someone's capital and I have to rework the terrain so that the city can grow. Let them have super workers or extra workers so that they can develop thier land properly.

4 Teach them how to play defensively, while its more difficult for the AI to do they need to build forts adjacent to their city so that it makes it harder to approach the city due to sphere of influence penality to movement.
4b There simple defensive things that can be done to just make it that much difficult to take cities such as clearing out forests around the city. Not attack my units on the over side of the river, rather waiting for my units to attack your fortified unit across the river.

5 Focus fire? Seriously...
6 Don't level my troops uncessarily. If the damage you can deal is < then the heal rate of my unit stop.
7 Don't make ranged units if your ******ed. Just make Warrior, Spearmen and Horsemen class if you can't handle the responsibilities.
8 Retreat your guys if they are damaged... you normally have like 10 other dudes behind you so you don't have to worry about getting run down.



In the end in comes down to being a hard problem to solve and the amount of money required to make an improvement. I don't know how sales for a game like Civ goes but most games will have 90% of their life time revenue generated in the first 3 months of their release. So... if I was a money guy, I wouldn't exactly be concerned about putting much more money into AI rather than producing DLC and explansion packs to solve problems and implement features that should of been including the game from the get go.

I'm sorry to disappoint you, the upcoming patch has several/multiple AI fixes.

 

[PawelS]

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.

 

[pickle]

the AI will never be as smart as a person, if it was then anything above noble/king would be unbeatable.
but there are many small problems with the AI that we can reasonably expect to get fixed and from what I read in the notes for the upcoming patch it looks as though that is what is happening.
also someone said that they will not bother fixing the AI because most of the profit is made early on. they still want to sell dlc, expansions and eventually civ 6. and if they don't fix some of the problems then these extras later on will fail.