Does the AI ever get less stupid?

[poom3619]

Yeah, making a Genius-level AI isn't near as easy or practical as people suspect.

It doesn't matter what difficulty level you are on, the AI is mostly the same. The real difference is the bonuses the AI receives in extra production, gold, science, etc. Including starter techs and extra units. Try Deity if Immortal is that easy for you.

No offense. But did coder of Firaxis find it too hard to make proper AI that can simply plays the game without making several odd/stupid mistakes? I for one don't need to be genius to understand the game rule and optimal way of playing. and I don't need genius AI on deity level. I need proper AI on deity level.

Yeah, Coding AI is harder than whining about what it suppose to do. But given that release date isn't set in stone. People would complains less if Firaxis give coder more time to make it. AI can do some mistake, Of course. But it is obviously a problem when constant complaint about AI that exist since CiV vanilla never been quieted. It is still a topic of discussing after years.

 

[Xahz]

No offense. But did coder of Firaxis find it too hard to make proper AI that can simply plays the game without making several odd/stupid mistakes? I for one don't need to be genius to understand the game rule and optimal way of playing. and I don't need genius AI on deity level. I need proper AI on deity level.

Yeah, Coding AI is harder than whining about what it suppose to do. But given that release date isn't set in stone. People would complains less if Firaxis give coder more time to make it. AI can do some mistake, Of course. But it is obviously a problem when constant complaint about AI that exist since CiV vanilla never been quieted. It is still a topic of discussing after years.

You understand that time and money cannot build infinite technology, right? We just barely have supercomputers who can beat human opponents in chess - and that's with millions of dollars being thrown into the development of that technology from some of the biggest tech giants in the world. Civ has infinitely more variables and there is economically no reason to develop a more intelligent AI other than to quiet a extremely vocal minority that still buy and play the game anyway.

Also, and this is by far the biggest and most important element, the more sophisticated the AI the longer your turns take. There is a middle ground between making the AI challenging but also keeping it grounded so that the game is playable. Giving the same AI Deity level bonuses is the best way to do that.

 

[bc1]

Stop trying to rationalize, the AI sux simply because it does not matter enough to the average player. Better graphics means better sales, which is much easier than better AI anyway.

 

[Sonereal]

Stop trying to rationalize, the AI sux simply because it does not matter enough to the average player. Better graphics means better sales, which is much easier than better AI anyway.

That, and the series isn't known for its strong AI in the first place.

Relevant

In the end, it depends on where you think Civilization V should, and could, fall on that ladder. Eventually, you will have a rather competent AI in this series. Just not for a while. A long while.

Battle for Wesnoth, which is a free game, handles terrain better than CivV. On the other hand, there really isn't much in the way of ranged weapons like in CivV. At least, the ranged weapons we think of.

 

[Xahz]

Stop trying to rationalize, the AI sux simply because it does not matter enough to the average player. Better graphics means better sales, which is much easier than better AI anyway.

So what is your basis for stating that better graphics are easier to develop than better AI? Do you have a source for that? Also, you are "rationalizing" with that explanation as well.

We know empirically that turn length is significantly longer the more variables there are for the AI to compute. I've stopped playing games towards the info era because turns get frustratingly long. If turns were that long from the beginning and even longer towards the end, I probably wouldn't even buy the next iteration of the series. The game would be unplayable.

 

[poom3619]

You understand that time and money cannot build infinite technology, right? We just barely have supercomputers who can beat human opponents in chess - and that's with millions of dollars being thrown into the development of that technology from some of the biggest tech giants in the world. Civ has infinitely more variables and there is economically no reason to develop a more intelligent AI other than to quiet a extremely vocal minority that still buy and play the game anyway.

Also, and this is by far the biggest and most important element, the more sophisticated the AI the longer your turns take. There is a middle ground between making the AI challenging but also keeping it grounded so that the game is playable. Giving the same AI Deity level bonuses is the best way to do that.

To be honest, Yes. There are actually no reason to develop good AI in terms of money. No wonder why they don't do that, and you apparently are not at week after vanilla release to see complaint on AI literally flood this sub-forum.

I understand why you think longer turn is effect of AI's calculation. I used to think like that too until I turn off Unit movement animation from advice of someone in this forum which turn out make late-game feel normal.

 

[NordicExpanse]

I had posted in another thread about Persia practically calling off its attack on one of my backwater towns because I had a near-death unit recuperating on a neutral tile within city defense range of Pasargadae. The city defenses finished off my hoplite and this action caused his swordmen unit and three archery units to retreat back inside Persian borders. AI is more comic relief than a challenge; granted, I'm only on Prince level.

 

[Zaimejs]

This game is complicated. But why does there seem to be the rule that the AI have to move all of their pieces every turn? They should be allowed to hold a position with their troops.

There should be subroutines written in... when attacking a city of a certain strength and era, there should be a way to offer probability of success based on a certain combination of troops. I can do that... I can look at an enemy city and determine that I'll need at least two siege, a couple of bows and maybe 3 melee troops to take that city. All told, without much resistance, I'll be okay. So AI should have at least that many before moving in...

Anyway... it's frustrating how stupid the AI is.

 

[Sonereal]

This game is complicated. But why does there seem to be the rule that the AI have to move all of their pieces every turn? They should be allowed to hold a position with their troops.

The AI doesn't hold a line. It holds a general area.

There should be subroutines written in... when attacking a city of a certain strength and era, there should be a way to offer probability of success based on a certain combination of troops. I can do that... I can look at an enemy city and determine that I'll need at least two siege, a couple of bows and maybe 3 melee troops to take that city. All told, without much resistance, I'll be okay. So AI should have at least that many before moving in...

You can look at a city and say, "Well, prolly need this, that, and the other thing." The AI doesn't learn like you do. Think back to when you first started playing and didn't know what you should use to take a city, so you probably overproduced a unit and underproduced another that would've been crucial to take a city.

After a few games under your belt, you feel you have a general idea what you want to do when you want to take a city. However, you run into a city where your previous strategy simply didn't work. Maybe its because the rough terrain means you have to bring your catapults to knife-distance to besiege the city?

Well, you can learn from that, but the AI...can't. The AI doesn't really learn.

Otherwise, yes. A good rule of thumb is to at least bring siege equipment and bows, something so simple you think the AI would do that, but it consistently fails at that task. I've seen the AI build up an invasion against me pre-Renaissance that consisted of no cats, only melee units and one archer unit, despite the fact the AI in question could build catapults and maybe even trebuchets.

The AI is also notoriously bad at keeping defensive lines up. At one point, I attacked a catapult that was moving between cities. I did not destroy the catapult (couldn't follow up right then), but the AI moved it to the other city. Instead of allowing it to heal and, more importantly, fire from within the city, it moved the catapult again and I destroyed it with a nearby horseman.

 

[gps]

You understand that time and money cannot build infinite technology, right? We just barely have supercomputers who can beat human opponents in chess - and that's with millions of dollars being thrown into the development of that technology from some of the biggest tech giants in the world.

You are talking about programs that are able to beat chess grand masters and world champions. I think the usual average chess player is easily beaten by ten year old chess computers bought at your local gas station for 39 dollars...

Civ has infinitely more variables and there is economically no reason to develop a more intelligent AI other than to quiet a extremely vocal minority that still buy and play the game anyway.

Well, it's interesting to see the lead designer Jon Shafer himself plus many game review magazines and portals joining that "extremely vocal" but totally irrelevant "minority" admitting/criticizing weak AI. I think that's a first in Civ history!
But I am with you. Seriously. How could those crazy folks expect to get a decent AI in a game of strategy??? Makes no sense at all. Also why should there be rifles and bazookas in an ego shooter or cars in a racing game...

 

[poom3619]

You can look at a city and say, "Well, prolly need this, that, and the other thing." The AI doesn't learn like you do. Think back to when you first started playing and didn't know what you should use to take a city, so you probably overproduced a unit and underproduced another that would've been crucial to take a city.

After a few games under your belt, you feel you have a general idea what you want to do when you want to take a city. However, you run into a city where your previous strategy simply didn't work. Maybe its because the rough terrain means you have to bring your catapults to knife-distance to besiege the city?

Well, you can learn from that, but the AI...can't. The AI doesn't really learn. [...]
.

They can make some tester play the game and teach their AI to. in a way. play like them, teach AI to counter when they meet a kind of player. Yeah, Civ5 is "more complex" and all that. But I think Panzer General done better in AI battle

and I thought AI can being teach to learn. At least that was Stardock claim since GalCivII (2006)

 

[Zaimejs]

AI can't "learn", but you can preteach it. If then statements are wonderful tools. If an enemy unit is within x hexes, don't move your catapult out of the city. If you are defending, build archers and put them in strategic hills. If attacking, then build x number of each type of unit. It can't be that hard to build in some basic strategy.

I still think 50% of the problem is the rule that Ai has to move every unit every turn. I think this is built in to allow for traffic flow, but it doesn't make sense when at war.

 

[Magma_Dragoon]

There is a computer that can beat any human in chess. It costs $100 million and is the size of an commercial refrigerator, but it wins every time.

 

[Silverfuturist]

I haven't seen many cheaply-available commercial chess AIs that can beat good players

Google "chess engines ratings." There are 20 computer programs that are stronger than any human. Programs like Houdini and Rybka are free, I run Houdini on a gaming Alienware desktop to analyze my games, it just blows me away how good it is. For less than $1300 I have something better than what took IBM a team of experts and millions of dollars to produce in the late 90s.

In multiplayer civ, it is amazing how few people have Google skills. By simply reading the posts of the players who win the game-of-the-months from the past few years, I get accused of cheating because my empires grow so fast and so strong. That is a big hint - you want to get better at civ? Find out who is winning GOTM competitions and read what they have to say. You can even click on their names to get more posts... another biggie hint... now I expect more people in MP to stop rolling over dead with my 2 big bombshells of tips that the world has been officially alerted about...

 

[Walter R]

Yes, well that dumb AI has so far stopped me attaining Emperor (albeit I've only tried twice) - OMG I'm even dumber than a dumb AI...who the hell programmed me?!

 

[Silverfuturist]

Yes, well that dumb AI has so far stopped me attaining Emperor (albeit I've only tried twice) - OMG I'm even dumber than a dumb AI...who the hell programmed me?!

You are self-aware! That is a great sign of intelligence, once you figure out how to destroy Emperor, in the future you will wonder why you thought Emperor was difficult. Just keep reading what the game-of-the-month winners keep posting about and you will get it... if you can master the alphabet to become literate you can beat Emperor.

 

[Sonereal]

They can make some tester play the game and teach their AI to. in a way. play like them, teach AI to counter when they meet a kind of player. Yeah, Civ5 is "more complex" and all that. But I think Panzer General done better in AI battle

I was actually going to bring up GalCiv.

The problem with testers is that it seems like you'd need quite a bit of them. You would need to find testers and give them time to learn the basics of the game, give them time to play on a variety or settings (or try to find a uniformed setting which everything is based around), analyze the player, and improve the AI so it can respond to that player.

Possible, but given that CivV released as a beta, also probably expensive.

Ideally, CivVI would have an AI capable of learning and responding to situations in a more nuanced way, such as building AA guns only when the enemy has heavy amounts of air power, but that's probably just a pipe dream.

AI can't "learn", but you can preteach it. If then statements are wonderful tools. If an enemy unit is within x hexes, don't move your catapult out of the city. If you are defending, build archers and put them in strategic hills. If attacking, then build x number of each type of unit. It can't be that hard to build in some basic strategy.

What if the city is about to fall and the catapult has a 50/50 chance of escaping?

What is a strategic hill?

Would this unit be an acceptable substitute for this unit based on this situation?

It isn't hard to make the AI less incompetent, but it is hard to make it competent, which is why CivV is ten times better played in multiplayer.

I still think 50% of the problem is the rule that Ai has to move every unit every turn. I think this is built in to allow for traffic flow, but it doesn't make sense when at war.

The AI does this even when preparing for war. It does this even when troops are on the border. Because of this, the AI keeps reevaluating the strength of defenders leading to situations where the AI has an army on your border for twenty turns.

 

[Magma_Dragoon]

Or in my case with Germany, 70 turns. Mountains were involved.

 

[ray111]

So far, I find barbarians (normals, not even raging) bigger threat than any AI opponent..
- a single barbarian unit can demolish your whole territory by just setting all your improvements on fire and taking the heal promotion every time, plundering your trade routes and therefore increasing their numbers, unless you have some troops ready to block their way into your kingdom
- I honestly haven't seen a single AI opponent plunder so far, so even if I get my whole country flooded with enemy units, I can fend them off with a shoe (literally, city+1 unit vs about 10+ units)

- invading AI is really easy as well, since I haven't seen them building a single fort or anything, the best "defense" they are capable is setting cities too close to each other, so you can only siege from one side or risk fire from 2 cities at the same time..

What would make AI terrifying?
- plundering
- building an occasional fort close to where your borders meet

two simple things that would make AIs MUCH more dangerous.

 

[4N4C0ND4]

So far, I find barbarians (normals, not even raging) bigger threat than any AI opponent..
- a single barbarian unit can demolish your whole territory by just setting all your improvements on fire and taking the heal promotion every time, plundering your trade routes and therefore increasing their numbers, unless you have some troops ready to block their way into your kingdom
- I honestly haven't seen a single AI opponent plunder so far, so even if I get my whole country flooded with enemy units, I can fend them off with a shoe (literally, city+1 unit vs about 10+ units)

- invading AI is really easy as well, since I haven't seen them building a single fort or anything, the best "defense" they are capable is setting cities too close to each other, so you can only siege from one side or risk fire from 2 cities at the same time..

What would make AI terrifying?
- plundering
- building an occasional fort close to where your borders meet

two simple things that would make AIs MUCH more dangerous.

I was wondering: do the barb units even got promotions? Because I think you confuse the healing promotion with the heal they get from pillage. That is why they can almost heal every turn

Btw I had countless occasion when enemy AI will pillage a bit of my land. It's not as systematic as the barb I grant you, but it does happen. However, I think I've never seen an AI pillage AND move on the same turn. But I could be wrong on that.

Trade roads are also plundered on a regular basis by barb & AI alikes in my games.