I'm Going to State the AI Problem as Eloquently as Possible.

Originally posted by Craterus22


here is a question for you - if you are on the brink of extinction who would you attack first? the most powerful civ in the world? or the guy right above you on the ladder?

or would you make a deal (whatever it takes) to have a stong ally help you attack the guy above you on the food chain...

I have seen the ai pound against the strongest civs with no hope of victory (let alone survival). I have also seen it attack weak civs too, but the stupidest AI should know an impossible battle when it sees it - this one doesn't


Such foolish, hopeless wars have occurred in the real world. Saddam Hussein's "Mother of all Battles." Japan should have known better than to attack a nation that had 10 times its industrial capacity in WWII. Likewise the Confederate states in the American Civil War should have noticed that the North had not only a bigger population, but had a vastly superior navy and industrial capacity. In both cases, the weaker nation thought its soldiers' "honor" would triumph. They had early successes, but the stronger nations ground them down eventually. And sometimes the stronger nation gets "war weariness" as in Vietnam.

I agree, though, that this tendency in the AI is too consistent. The AI should refrain from so many foolish attacks; it should make stupid attacks, but not all the time, not all AI civs. I think we need at least some of the AI civs to be smarter, but not all of them.
 
Originally posted by sumthinelse


Such foolish, hopeless wars have occurred in the real world. Saddam Hussein's "Mother of all Battles." Japan should have known better than to attack a nation that had 10 times its industrial capacity in WWII. Likewise the Confederate states in the American Civil War should have noticed that the North had not only a bigger population, but had a vastly superior navy and industrial capacity. In both cases, the weaker nation thought its soldiers' "honor" would triumph. They had early successes, but the stronger nations ground them down eventually. And sometimes the stronger nation gets "war weariness" as in Vietnam.

I agree, though, that this tendency in the AI is too consistent. The AI should refrain from so many foolish attacks; it should make stupid attacks, but not all the time, not all AI civs. I think we need at least some of the AI civs to be smarter, but not all of them.

Wrong on all counts.

The South had a very good chance of winning (by surviving) in the Civil War, and if Sherman hadn't taken Atlanta when he did in 1864 Lincoln way very well have lost the election to peace candidate (and Southern sympathizer) McClellan.

Japan misjudged America's RESOLVE mostly; even so, if they had the same luck we did at the Battle of Midway in 1942 they may have been able to force the American fleet (what little was left of it) back to the American coast, and perhaps force a negotiated peace if the American casualties became to heavy.

Saddam had no choice but to leave his cannon fodder conscripts to receive the Allied attack. His best armor units also didn't know that the new American fire systems in the Abrams tanks totally outranged the best he had.

I am trying to think of any historical attack as stupid as the AI's attacks. . . in 6,000 years of history the only one I can come up with that is close is the Khmer Rouge of Cambodia attacking Vietnam in the 1970's. But the KR were demented genocidal fanatics. The reality is that the AI is stupider than most any historical civ.
 
History is littered with bad decisions concerning war. But because of the great variability in combat results (randomizer), there is always a chance of victory, justifying many risky adventures. So there are two classes: Obviously bad decisions that resulted in disaster, and obviously bad decisions that result in victory.
 
Zach and others

Yes there are times when civs have made bad decisions in the "real world"

The bad decision shouldn't be the first club out of the bag though...

We sometimes call something a bad decision because of 20/20 hindsight. At the time it wasn't such a bad decision if we can put ourselves in thier shoes with the information available to them at the time.

The ai should have the capability of seeing the map and counting up where units are (or the approximate amount of units in a given region)... just like we do! I used the OGRE example earlier... where points are assigned to each unit so that it can estimate where the worst enemy is at the time.

Occasional harassing troops are good, if you had a plan to attack elsewhere within a couple of turns. Heck a good raiding party wouldn't be a bad idea(a two or three unit stack making its way to pillaging a resource or two).

The one positive thing I can say aobut the ai is that they decided to provide protection for settlers... why can't they do something similar with coordinating attacks?


btw- i bought ctp2 based onsomeones recomendation here on the boards(thanks) - it was $15 at my local EB and with the mods it has been a pleasure to play. It looks like I won't need to buy the expansion for civ3 now - MP and numorous civs to choose from...

It would be cool if the civ team opened up scripting to the masses similar to ctp2 - I believe collectively we could come up with some better solutions for the ai (also-it would be nice if we could make a different ai for each civ)
 
Originally posted by Zouave


Wrong on all counts.

The South had a very good chance of winning (by surviving) in the Civil War, and if Sherman hadn't taken Atlanta when he did in 1864 Lincoln way very well have lost the election to peace candidate (and Southern sympathizer) McClellan.

If Jefferson Davis hadn't put Hood in charge of the Confederate forces at Atlanta, Sherman may well have not taken it in time...



Japan misjudged America's RESOLVE mostly; even so, if they had the same luck we did at the Battle of Midway in 1942 they may have been able to force the American fleet (what little was left of it) back to the American coast, and perhaps force a negotiated peace if the American casualties became to heavy.

Had they had better luck at Pearl Harbour (Harbor!?) and found the US carrier fleet there as they had anticipated, Midway probably would not have happened at all




I am trying to think of any historical attack as stupid as the AI's attacks. . . in 6,000 years of history the only one I can come up with that is close is the Khmer Rouge of Cambodia attacking Vietnam in the 1970's. But the KR were demented genocidal fanatics. The reality is that the AI is stupider than most any historical civ.

Its not that difficult if you look for individual battles/events within a wider war. The American Civil War is littered with them, for instance: Pickett's Charge at Gettysburg, Cold Harbor (even Grant regretted that one), Franklin & Nashville (John Bell Hood again) to name but a few
Just to even things up and poke "fun" at my side of the pond: what about almost anything done on Western Front in WW1
(Lions led by Donkeys, anyone?) or Gallipolli (well, it was hardly the ANZACs' fault was it...?)
 
Let's take a look at how stupid the AI really is. See the attached image.

I, as the Persians, are putting my Immortals to use as I left the crappy island I was stuck on and start invading the French and moving south before the Japanese start getting too powerful.

There is little left of the French except Lyons. I have a stack of Immortals and catapults ready for the final assault.

So what are the French doing? See the image. Not one but TWO settler/foot soldier combos is wandering north. WHY? Because I earlier had to raze Paris and I built a new town there - Gordium. But because there were open tiles the braindead AI, with its Settler Diarrhea, feels COMPELLED to send settlers there. That obsession overrides everything including self-preservation. :crazyeye:

So I wiped them out, got four free workers, and then stormed Lyons.

Later in that same game, Rome, one third the size at best of China, declared war on China. Fifteen turns later the Romans were exterminated.

That braindead AI. :crazyeye:
 

Attachments

  • yyyyy.jpg
    yyyyy.jpg
    104.2 KB · Views: 109
Originally posted by Zouave


Wrong on all counts.

The South had a very good chance of winning (by surviving) in the Civil War, and if Sherman hadn't taken Atlanta when he did in 1864 Lincoln way very well have lost the election to peace candidate (and Southern sympathizer) McClellan.

Japan misjudged America's RESOLVE mostly; even so, if they had the same luck we did at the Battle of Midway in 1942 they may have been able to force the American fleet (what little was left of it) back to the American coast, and perhaps force a negotiated peace if the American casualties became to heavy.

Saddam had no choice but to leave his cannon fodder conscripts to receive the Allied attack. His best armor units also didn't know that the new American fire systems in the Abrams tanks totally outranged the best he had.

I am trying to think of any historical attack as stupid as the AI's attacks. . . in 6,000 years of history the only one I can come up with that is close is the Khmer Rouge of Cambodia attacking Vietnam in the 1970's. But the KR were demented genocidal fanatics. The reality is that the AI is stupider than most any historical civ.

Zoave, thanks for the comment. It's interesting to debate with someone who knows something about history.

As I said, I agree with you that the AI is too *consistently* stupid about declaring war and incompetent in tactics. Infogames/firaxis may have dumbed down the AI to keep players from getting unhappy about the difficulty. I think most of us would welcome more configuration options to control the AI's competence.

As far as the South winning the Civil War, maybe it was possible for the South to hold out. And my great-great-great grandfather fought for the South, was captured when they surrendered at Vicksburg, was released in a prisoner exchange, joined the Confederate Army again, and was captured a second time! Still, most people would agree it was a foolish war for the South to get into, for many reasons, including the ones I mentioned previously.

America also had extremely bad luck in the Pacific. There was a telegraph that somehow never arrived in time at Pearl Harbor. It would have notified Admiral Kimmel that the Japanese were about to attack. At Midway, the American torpedoes released prematurely from many of the torpedo bombers, but even so, the number of torpedo bombers sent at the Japanese carriers pulled the Japanese fighters too low to stop the dive bombers. Tolstoy speculated about cause and effect in history when writing about the battle of Borodino in War and Peace. The luck factor usually evens out over time. If the Japanese had lost only 1 or 2 carriers at Midway, the war might have lasted longer. But in that case, another Midway-style ambush surely would have happened somewhere else later on, because America had broken their code. Japan underestimated not only America's resolve, but also:

- The difference America's wealth and industry would make

- America's advantage in science, which would grow as the war progressed.

- I am probably omitting the "practical" side of American warmaking, which, unlike Japan, did not bet everything on a small, irreplaceable, elite group of naval pilots.

Yamamoto did not underestimate, and did not want a war with, America, for very good reasons.

"Saddam had no choice..." Yes he did. And if he didn't know about the Abrams tanks, why does the AI have to know how good your units are?

I think the most one-sided wars don't get as much historical coverage simply because they tend to be less interesting. But they occurred. For example, the little island Melos (Thucycides, the Pelopenesian War XVII) fought the vastly stronger Athenians with the result that their men were all killed and the women and children enslaved. And this chapter might never have been written, but Thucydides wanted a story to demonstrate a principle of "justice," which was, for the Athenians, "might makes right."
 
Okay, time to play a little devil's advocate here.

A good example of a stupid war would be when Argentina tried to take the Falklands away from Great Britain. Was there any doubt how that was going to turn out?

Settler diarheea (sp?) has occurred in history. Great Britain and Spain did the same basic thing when they built their empires. The basic motto was "go out and suck up land and resources, wherever they may be." A little territory in the New World, a little in Africa, a little in Asia, this island group in the Pacific, that group, oops, that one is already taken, but the next one is free.

Enough devil's advocate.

The two examples I just gave are exceptions, and in the case of the empires, they weren't playing those games close to home, like you're forced to do in the game.

I've seen a lot of comments about performance hits if the AI was improved. Performance shouldn't even be an issue. There is a LOT of wasted time (from the CPU's point of view) while the player is moving. If some fairly strait-forward multi-threading was done, the computer could analyze what needs to be done during that time. Adjustments could be made based on the players moves, but many of the moves that are made either a) shouldn't be visible to the other civs, or b) aren't of use to the computer. My order to build a bank in an interior city shouldn't have any affect on what the computer does. The difficulty levels could be determined by how much of this dead time is used for analysis.

Many of the issues we gripe about have nothing to do with AI. For example, the diplomacy issues usually aren't AI-related. If a unit crosses my border, even if it is just trying to get from point A to point B, it is an invasion force. Giving me an option to smack around that invasion force, without hitting my reputation for the rest of the game, is NOT AI-related. What would make it an AI issue would be if the computer had to decide whether the repurcussions of sending that force into my territory was worth it anyhow. Since there are no repurcussions to consider, there is no AI.
 
Originally posted by talen
Okay, time to play a little devil's advocate here.

A good example of a stupid war would be when Argentina tried to take the Falklands away from Great Britain. Was there any doubt how that was going to turn out?


It may have worked though. If they had held out for longer and been luckier then we may have ended up letting them keep te islands. Also the Argentina's wanted the islands alot so attacking the Falklands helped to keep their government in power...
 
Originally posted by Dell19


It may have worked though. If they had held out for longer and been luckier then we may have ended up letting them keep te islands. Also the Argentina's wanted the islands alot so attacking the Falklands helped to keep their government in power...

That's right! Many wars are fought for domestic political reasons without regard for the general welfare of the state. The government in Argentina was using this as a ploy to justify their military oppression, though it backfired.

On the Persian v. French situation posted above, was there a better strategy? Of course not. They were going to be overrun no matter what they did. Making a last ditch attempt to escape was as good as any plan.
 
Originally posted by Zachriel


That's right! Many wars are fought for domestic political reasons without regard for the general welfare of the state. .

I agree. And some wars are fought out of ignorance/blindness. "God is on our side...." "We hate you so much, we don't care if this war destroys our country."
 
my main problem with the AI is the way it calculates power. Some little dinky civ with 120 units looks at me, and goes, I ccan take him, he only has 80 units. So they declare war, and by the time they move their army to me, a couple turns later, they go holy sh*%! he now has 140 units! maybe we should have considered that he has 35 citys, and is able to build about 20 tanks a turn! The AI has no comprehension of quality of units (taking out Zulu UU in modern age is fun though) and they don't acknowledge the fact that although the army is small now, when the enemy mobilizes for war, he can build 10 tanks and 10 artillery a turn, with a couple infantry to boot.
 
Argentina had a pretty good chance of winning, actually, but they had a number of things go wrong too.

In their favor, the Falklands are pretty far away from England. They're also pretty small and insignificant. There are not too many islands around that can support British land-based bombers. It took quite a while for troops to sail the 10,000 miles from England to the Islands. The US has this handy "keep Europe out of the Western Hemisphere" policy (the Monroe Doctrine, I believe), so the US should've been on their side. Britain's carriers are not nearly as powerful as America's, and Argentina had a similar carrier. If the war could be dragged out for only a couple years, Britain would just give up the islands.

But going against them was the fact that America sided with Britain and lent support with tankers and electronic intelligence. Britain had a couple subs already in the area to harrass the Argentine fleet. The new French fighters and Exocet missiles they'd bought were not all delivered in time. The invasion was moved up 6 months. The Harrier was a much better dogfighter than anyone thought. The Argentines had very little guided ordinance, and they had an awful lot of dud bombs.

The Argentines had some pretty good pilots, and apparently anti-aircraft missiles aren't all they're cracked up to be. There was one encounter where a pilot on a bombing run over the British fleet had all his bombs fail to drop or explode. So he did the only thing a frustrated pilot could: he dropped his drop tank on the ship! Argentine pilots had to overfly their targets to drop their bombs; I sure wouldn't wanna do that!

The Brits made good use of their subs and damaged the Argentine carrier, preventing it from launching several operations that might have severely hampered the British surface fleet, and they sank the Argentine battle cruiser, which might have been important in defending the islands from the British raid that finally retook Port Stanley.

The Argentines had a lot of bad luck. But perhaps some of that could have been avaoided with proper maintenance or something? A lot of the problems were dud bombs or failed releases. Not completing the transactions with the French for the advanced weapons was also a bad idea. If they had finished the sale, and sunk a few British ships, or if more bombs had worked, British public opinion would likely have swayed toward "we're taking too many losses, and for what? a couple islands?" and the Argentines would have won.
 
Long, pointless rant:

I noticed someone commenting on the game of Go here. Nice. FOr the players who dont know the game, it is pretty simple:
You play on the crosses, not in the squares.
Each player puts down 1 stone a turn.
The stones never move.
A stone is taken off the board when surrounded by enemy stones.
Black goes first or black gets extra stones and white goes first.



Oh- and the board is 19x19=361 crosses. Starting from the beginning, with no stones placed as handicap, the first 6 moves can happen in any of 2 122 781 978 399 040 ways.

In one game that I saw reported on the net somewhere, a computer was given over 15 stones handicap, and lost. The first white stone looked like a paratrooper dropping into enemy territory.

I dont know much about AI programming except the fact that AI bases itself on checking each move and each move sequence to see if that can lead to or towrds victory. A person playing just looks for good moves. A Chess Grandmaster considers 2 moves per turn, I read. Chess is nothing like civ3, but the same principle applies.

<End long pointless discussion>
 
Back
Top Bottom