Cheating AI.......

Spec · February 20, 2002, 15:36

God I'm tired of hearing ppl complain about a cheating AIs. Is there one game in the world where the AI doesn't cheat?!?!?
Personnaly I never saw one, may it be on a PC game or a platform game. Let me explain:

Racing games: The harder the level the harder it is to win. Why? Not cuz the AI is suddenly a Michael Shumacher, no! Cuz the AI has a catch up!! Cuz you learn to get better, not the AI so it has to have an advantage or there's no reason to call it "level HARD"

First person shooters: At hard levels, the AI either has wallhacks (CS bots), Increased energy and armor, increased damage, increased reaction time (Halo, Half-Life, Soldier of fortune....) and kills you with one shot compared to 15 for you to kill it. (that's Halo for you...)

Strategy games. Of course the AI knows were you are all the time. If anyone has ever played Starcraft here, you just know the AI has a map hack! It has 5 islands it has to check before finding you but somehow, it always does it on the first try. (Red alert, Age of empires and more, it's all the samething)

And I could go on and on....So just accept it, AIs cheat. That how they get harder and harder. They dont LEARN, they are given advantages over you. And this is not specific to Civ3!!!So suck it up!

Spec.

The Andy-Man · February 20, 2002, 16:10

yeah, but some AI advantages just piss some ppl off cos they are just TO unfair (like in Civ2, triremes could sail anywhere, and in Civ3, the AI goes straight for your week cities even though it dont know where they are.)

of course the AI knows the mao already, but it would be nice if it didnt abuse the fact the way it does...... and production bonuses suck, aswell as the fact that i have never notice the AI get war weariness.

Spec · February 20, 2002, 16:18

Quote:

Originally posted by The Andy-Man

of course the AI knows the mao already, but it would be nice if it didnt abuse the fact the way it does...... and production bonuses suck, aswell as the fact that i have never notice the AI get war weariness.

Cheesy advantages just like in anyother game. The most annoying AI cheat I ever saw is in Starcraft. It knows the map and where any of your units are. That makes the game a lot harder but I like it anyways. And of what I know, artificial intelligence is not yet available for video games so we have a long wait until it is.....

Spec.

Encomium · February 20, 2002, 16:23

I guess some of you wouldn't mind if you started playing Chessmaster - and the AI decided to suddenly give a bishop the power of a queen so it could put you in check! I WOULD mind.

Rules are rules, and common sense and reality should not change. If the AI beats me by some outrageous cheats it has achieved nothing, except to encourage me to find ways to cheat back.

I want a SMARTER AI - not an AI that cheats more and breaks its own rules.

Spec · February 20, 2002, 16:52

Quote:

Originally posted by Encomium

I want a SMARTER AI - not an AI that cheats more and breaks its own rules.

Be my guest OH MASTER ENCOMIUM, please develope the thecnology that every company has now failed to achieve!! Surely you can help us OH MASTER KNOW IT ALL!

Every AI breaks it's own rules. As I mentionned above, Map hacks, Wallhacks, Catchups, bonuses, name it! If you dont get that than I can't help you.

And btw, I never saw an AI suddenly change a Warrior into an ICBM or whatever. A chess games is composed of 36 squares only, you dont have to irrigate/mine, build cities/units, control an empire for crying outloud!! And the AI in a chess game only works with mathematical strategies. Nowhere NEER the kind of AI for a videogame. And by videogame , you know what I mean.
This is not a chess game you...............

Spec.

The Andy-Man · February 21, 2002, 11:13

i've mever seen the AI in Europa universalis 1 cheat... i dunno if it does, but i never seen it do anything stupid and unfair, also, it dosnt seem to know where the best provinces to colonize are straight away and so on. the AI dont even gang up on human unless he get taking provinces.

and there are plenty of games where the AI cheats, but not to the extent of destroying gameplay.

woody · February 21, 2002, 11:47

Quote:

Originally posted by Encomium
I want a SMARTER AI - not an AI that cheats more and breaks its own rules.

Then why don't you just code one up, and make yourself a million bucks, genius?

Ironikinit · February 21, 2002, 12:11

Europa Universalis. It is said by some that the AI doesn't cheat in that game. The AI is a pushover IMO, but supposedly it doesn't cheat.

There are other games I'm pretty sure didn't have the AI cheating, but EU is the only game with complexity comparible to Civ.

Personally, I have no problem with the AI cheating. It can't save the game and reload, that's all the philosophical justification needed right there. Saving the game is the biggest cheat of all. Well, other than playing in god mode or so.

Someday maybe programers will be able to run code strategy game AI that doesn't have to cheat to challenge good players. In the meantime, pity the paranoid types such as Enco. He thinks Firaxis pays people to make favorable posts here and at CivFanatics. Imagine what he thinks the AI is up to.

"I didn't get that big promotion at work and it's all because of that no good cheating AI in Civ 3!"

"My candidate lost the election. Damn you, Firaxis!"

Etc.

Grumbold · February 21, 2002, 12:26

In EU1/2 the AI follows almost all the rules exactly like the player does. There are a couple of notable exceptions in EU1 which I *believe* they ironed out in EU2. This does partly mean that the AI is predictable and can be beaten in the long run. This doesn't mean that you can win every game though. NPC France is still quite capable of whooping player England simply because its so much bigger and stronger in 1492. You have to play the diplomatic game well to avoid destruction as a weak country. Become too powwerful and almost everyone will declare war against you when they see you've got your hands full. Try to fight an army led by a superlative historical leader and you'll get whupped even if you outnumber them 4:1. Much like Civ once you own a third of the world the hard part is over.

Rimpy · February 21, 2002, 14:15

Quote:

Originally posted by Ironikinit
Someday maybe programers will be able to run code strategy game AI that doesn't have to cheat to challenge good players.

I think that this is probably the Holy Grail of AI programming in video games. Ideally, the AI would be as smart as it could be at the highest difficulty level, and purposefully dumbed down at lower difficulty levels.

The only "bonus" that the AI should get is the fact that it can make a million calculations per second and it's able to consider every piece of information that it has. The human, on the other hand, can only reasonably consider a limited amount of future scenarios, and he/she can't process the huge amount of data that's available. In other words, the AI can micromanage to a degree that's just not feasible for a human to do.

The ideal AI is like that usually found in chess games. An increase in difficulty levels simply means that the AI gets more time to play out future scenarios. At lower difficulty levels, it only gets a limited amount of time to try out these different scenarios, and ends up making dumber decisions.

Unfortunately, because the universe of future scenarios is so much more complex in a game like Civilization than in Chess, we're stuck with hacks like research and production bonuses. I don't blame the AI programmers for resorting to hacks. It's the only option that's feasible given current computing power.

Rimpy

Zouave · February 21, 2002, 14:54

Quote:

Originally posted by Encomium
I guess some of you wouldn't mind if you started playing Chessmaster - and the AI decided to suddenly give a bishop the power of a queen so it could put you in check! I WOULD mind.

Rules are rules, and common sense and reality should not change. If the AI beats me by some outrageous cheats it has achieved nothing, except to encourage me to find ways to cheat back.

I want a SMARTER AI - not an AI that cheats more and breaks its own rules.

The AI plays by slightly different rules to make up for its supposed intellectual inferiority. Now if you are having trouble .....

Leonidas · February 21, 2002, 15:45

Perhaps a way around this is to design an AI that "learns" from the strategies of the human player.

I know this has been tried before in one other game I know of using a neural grid path.

I'm not sure if this can be done, but it would work something like this:

1) The player pillages the AI terrain, the AI then starts to pillage. This strategy is added to its database.

2) The player uses a ship to load units behind the lines. The AI starts to do the same thing.

Even though the AI would have built-in strategies and decision-making, it would continue to "learn" from the human player.

Another possible method, would be to have the very best players' strategies for the game in question be designed to be used by the AI. This would work well in a game that has many sequels like the civ series.

In this instance the very best players in Civ3 could have all the loop-holes/strategies they have devised to be incorporated into the AI decision-making tree. . .

Not sure if this is possible, but just a couple of ideas to kick around

This would involve a lot of playtesting though, so the early release of games would be a no-no

Alex · February 21, 2002, 15:47

"Anti-Encomium"? You are a DL, aren't you?

Zouave · February 21, 2002, 15:54

Quote:

Originally posted by Alexnm
"Anti-Encomium"? You are a DL, aren't you?

Double Login? No. I just never bothered to post here before.

If you go visit civfanatics.com, there is a poster named Zouave, who I am 99.999999% sure is also Encomium. Either that, or one of them simply cuts and pastes posts. Anyway, the repetitive and unconstructive nature of his posts have managed to infuriate me enough to flame him on another forum.

Spec · February 21, 2002, 16:06

Quote:

Originally posted by Anti-Encomium

Anyway, the repetitive and unconstructive nature of his posts (Encomium) have managed to infuriate me enough to flame him on another forum.

I understand and agree withyou all the way.

Minuteman · February 21, 2002, 16:11

Quote:

Originally posted by Anti-Encomium

Double Login? No. I just never bothered to post here before.

If you go visit civfanatics.com, there is a poster named Zouave, who I am 99.999999% sure is also Encomium. Either that, or one of them simply cuts and pastes posts. Anyway, the repetitive and unconstructive nature of his posts have managed to infuriate me enough to flame him on another forum.

Oh Christ! I can't believe I'm sticking up for Encomium here, but I feel compelled to.

Not that I've been keeping score or anything, but Encomium/Troyens hasn't posted anything at civfanatics in awhile. I guess he found the "Fanboy friendly" environment there not to his liking. So I don't think anything he posted there recently drove you to come here and take shots at him.

Oh well! So much for my good deed for today...

Alex · February 21, 2002, 16:21

Quote:

Originally posted by Anti-Encomium

Double Login? No. I just never bothered to post here before.

If you go visit civfanatics.com, there is a poster named Zouave, who I am 99.999999% sure is also Encomium. Either that, or one of them simply cuts and pastes posts. Anyway, the repetitive and unconstructive nature of his posts have managed to infuriate me enough to flame him on another forum.

Well, I couldn't agree more with you...

Back on topic, before Ming closes this...

I think that it's simply impossible to have an AI that doesn't cheat, given our current level of computing power.

Now... sometimes I'm also pissed off with the way that the AI cheats. The way it goes right after your weakest city, or the (post-patch) tech-whoring that happens among AI civs. Other than that (and those things will not spoil the game for me, by any means), I don't mind it. Cheats are a necessary evil.

ACooper · February 21, 2002, 16:25

Quote:

Originally posted by Encomium
If the AI beats me by some outrageous cheats it has achieved nothing, except to encourage me to find ways to cheat back.

Reminder: It's a game.

BTW, Tell me of an "outrageous cheat". I'll be waiting.

Ironikinit · February 21, 2002, 16:29

The AI tech trading might not be the best solution, and it still discriminates against the human in trade in a quite obvious manner. However, the current set up seems to be generating fewer complaints than the earlier way the AI had of trading techs and resources on the player's turn. That was frustrating and prompted me to reload sometimes, one of those instances Enco talks about when the AI cheat leads to the player cheating. I don't like to reload on principle, but I like it even less when it takes a full minute to do it.

Yadda yadda. In short, I like the new way better.

Zouave · February 21, 2002, 16:30

Many of the AI 'cheats' directly balance a flaw in the AI. For example, the AI judges military strength purely on quantity, so it often ends up in wars that a human player would never start. But, when the war starts, it has the advantage of knowing which cities are poorly defended, giving it a slight advantage. Also, the AI is not nearly as good as the player at developing cities via improvements and terraforming, so it is harder for the AI to create good 'science' cities. To balance this, the AI is programmed to trade technologies aggressively, to keep up with the human.

Zouave/Troyens/Encomium still posts at Civfanatics, although maybe not in the same quantity. And still he posts the exact same things he posted a month ago, in almost the same words. These same things are posted here at Apolyton as well. Occasionally, he calls people fools who do not agree with him.

Spec · February 21, 2002, 19:25

Quote:

Originally posted by Ironikinit
...I like it even less when it takes a full minute to do it.
"reloading that is"

Now there's a nice piece of machinery!

Captain · February 21, 2002, 19:31

Quote:

Originally posted by Anti-Encomium
Many of the AI 'cheats' directly balance a flaw in the AI. For example, the AI judges military strength purely on quantity, so it often ends up in wars that a human player would never start.

hmm, there was a really good thread on the inherent AI limitations and how gameplay was designed around those limitations... it's been a while and my memory is crap so I can't remember which thread that was or even the name of the person who explained it (maybe Analyst Redux?) anyhow, it made a whole lot of sense.

still, in this case, I don't see how judging its military strength by quantity is an inherent limitation. that's just poor coding. wouldn't using the A/D/M + HP values of the units be much better? it seems simple enough. But maybe I'm missing something...?

Spec · February 21, 2002, 19:56

Quote:

Originally posted by Captain

still, in this case, I don't see how judging its military strength by quantity is an inherent limitation. that's just poor coding. wouldn't using the A/D/M + HP values of the units be much better? it seems simple enough. But maybe I'm missing something...?

I'd say that the AI should also judge the terrain def and movement value because Civs that expand in mountains/hills/jungles are harder to exterminate than the one's on grasslands/deserts/plains. But the AI doesn't seem to care about that. It would be cool to be able to use the terrain as kind of a psycological fortress against the AI.

Spec.

Seghillian · February 21, 2002, 21:09

In Chess programs I understand that the best AI has now achieved grandmaster status and can rival the world champion in a game. The work on this project began more than 30 years ago and must have involved hundreds of chess players and programmers during that time. So effectively more than 1000 human years have gone into developing the AI for a game with a tiny fraction of the complexity of CIV III.

To develop an AI in CIV III that could rival a human player would surely demand vast resources and would make the game unaffordable, for a game whose shelf life will be measured in months whereas that for chess is measured in centuries. Personally I don't see the problem. If you're not enjoying the game because the AI is too tough then just turn down the difficulty level. At the lower levels the player actually has advantages over the AI. Or cheat yourself. As long as it's not GOTM or MP no-one is looking.

February 22, 2002, 04:28

Just a quicky here...

I was playing a game of Baldur's Gate 2 TOB, and the funnest thing I had ever seen in a game happened. During one event my charecter saved these 3 NPCs from a stone curse caused by a beholder, after some dialog, etc. they attacked me. Of course the game allowed me to kill them, but then it happend... they RELOADED and came back a LOT nicer. One of the best easter eggs in a game I ever did see =)

Ironikinit · February 22, 2002, 05:33

I'm pretty sure that the AI can take the quality of units into consideration as well as quantity. For example, if you have a localized weakness, such as the fabled city guarded with a spearman in the modern era, the AI will possibly attack with that area as its goal. Further, I've reloaded games after getting a declaration of war (strictly for scientific purposes of course), changed my troops around, mainly to beef up my borders, and the declaration didn't occur the second time.

I'm not sure how important the unit count really is. People talk on it a lot, but disbanding units doesn't bring automatic AI aggression.

Grumbold · February 22, 2002, 06:21

I think it does some basic math but whatever calculations it makes are different to a humans. Having 45 points of AF when the defender has only 30 points of defence sounds good until you add in fortification, terrain and city bonuses. Still, I'd rather have the AI getting belligerent once in a while because I'm ignoring my military or leaving certain areas undefended than sit cowering in fear. Some games are already too peaceful because if you do strike the right defensive balance it is quite possible to remain on good terms with all your neighbours (caveat - from pre-1.17f experience).

DrFell · February 22, 2002, 08:35

I doubt this thread will get closed.

I can handle an AI that cheats to make it more challenging. Some cheats are ridiculous, but production and science bonuses are ok. It's the only way to make a difficult AI with present computer technology. Sure, they could make it smarter, but how long do you want the AI to take on it's turns? Some of you should try coding an AI or even designing one some day. It's no cakewalk.

The Viceroy · February 22, 2002, 09:15

There can be no perfect AI, simply as there is no perfect human player !! The trick is to have an AI which is challenging, unique (in different approaches to similar issues), makes occational booboo's when exploring as we humans do ..

If we go for an AI which learns, we have to start with laws .. basic rules which we humans always follow .. as otherwise you have an AI which really doesn't know what its trying to acheive .. survival is the most basic instinct .. greed, anger, love, fear, interest in the unknown..... and so on, and so on .. are very very difficult to provide the AI before it starts on its path of learning.. We can add memory, so it remembers who to hate, but it needs a reason to hate .. maybe a threat on its basic survival .. but hate can also be driven by other things like greed ... Its not a matter of coding, more one of thinking of every basic driving force which makes you the human do what you do .. map them out into a set of rules, which always apply (or use fuzzy logic where some apply sometimes) .. and your on your way to having a much better human simulation ..

To make it worse, now you have to teach it .. It can learn from the human, as we do .. but experience is the best master .. each Ai also would need to be different .. following different experiences .. Before the AI can be released into the world of Civ .. it needs to be good, smart, able to cope with other humans ..

Before anybody tells me to goaway and code it, I will tell you .. do you not think I would have if I could ? .. obviously, id love to write this .. but I really can't imagine the multitude of scenario's we'd have to cover to get this working .. maybe somebody wants to start an Apoyton AI group who bravely maps out the ideal AI .. rather like the Human genome project .. it would take a damn long time ..

ps .. I have done AI programming before, and this is a billion time s more complicated than anything ive touched before (if you want Learning AI that is, rather than basic decision tree stuff).