A simple truth about the so called "AI"....

[Hellenes]

During my years in these forums (since the times of Legiontotalwar) I warned people about the OBVIOUS FACT that there is no and NEVER will be any ARtificial Intelligence (just the notion of this term lowers my IQ) in our lifetime....
Right before RTW's release I said that without a MP campaign (WITH BATTLES) the TW series were living in a PURE PIPEDREAM....
The PROGRAMMED OPONNENT (as I said no "AI" until 2900ad) will be ALWAYS stupid in a game of the complexity of the TW series....
I just cannot comprehend how intelligent and mature people can even think of an "AI" in a commercial piece of software designed to run on CPUs that cannot even dream the power needed.....
THANKS GOD CA is coming with a MP campaign with this the TW series are FINALLY complete.... thus any "AI" pipedreamers THAT HAVE internet access can be swiftly pointed towards MP....

[upsettingshorts]

It's called heuristics, and it can mimic intelligence until you figure out what its actions and reactions are. ETW has done a better job of disguising them and varying them than previous TW series. In general, I do agree that multiplayer campaigns are ideal and I'm looking forward to them.

...or maybe I'm confusing it with fuzzy logic. I haven't had my caffeine today...

[Laoch]

I completely agree with what you're saying. However, I think the main gripe people have with the AI is that it's very poor not that we want it to be of the same intelligence as ourselves. We understand it will never be able to take on a human, but the standard it's at now is not good enough.

Just my two cents.

[sdomi_cabsav@Hotmail]

You seem to miss the definition of AI. One terrible example, a bird has intelligence but come on! it isn't very intelligent. We have AI but yes it is very terrible compared to humans but it is still called AI. They do not mean AI as the incarnation of the human mind in a machine. I mean yes that is quite ridiculous. Yes machines can be intelligent but at an incredibly basic and primitive level. Programmers just try and code AI so it creates an illusion of human like intelligence.

[Phier]

And yet, a computer can beat the best human chess player. All I'm asking for is for some if/then programming, I'm not looking to discuss philosophy with the game.

[upsettingshorts]

And if you have Deep Blue in your study, you have the right to expect that kind of artificial intelligence too.

Granted, an AI wouldn't need to be that good to defeat your average TW player - Kasparov is like a Napoleon of Chess. We're just armchair generals.

[WeardLan]

And if you have Deep Blue in your study, you have the right to expect that kind of artificial intelligence too.

Granted, an AI wouldn't need to be that good to defeat your average TW player - Kasparov is like a Napoleon of Chess. We're just armchair generals.

Deep Blue, as I understand it is not 'artificially intelligent' as such it's just that it's 'book opening' database is calculated out to end-game, such that it is always able to choose a route through the tree that leads in the worst case scenario to a draw.
Thus it only beats Grand Masters because it can't lose, but you can.

That's about as artificially intelligent as a 6 year old girl throwing darts at the financial times and 'beating' the stock market.

[Hellenes]

And yet, a computer can beat the best human chess player. All I'm asking for is for some if/then programming, I'm not looking to discuss philosophy with the game.

Take a board of chess and calculate the PER SECOND needed variables....(just a hint there arent any there is a thinking time)
And then look at the TW battles and calculate the per second variables (hint hint REAL TIME)....
Plus I doubt that Kasparov would be as good at a TW game as the best TW player....
Chess is FAAAR simpler to OPERATE than a TW game....

[Nebuchadnezzar]

Take a board of chess and calculate the PER SECOND needed variables....(just a hint there arent any there is a thinking time)
And then look at the TW battles and calculate the per second variables (hint hint REAL TIME)....
Plus I doubt that Kasparov would be as good at a TW game as the best TW player....
Chess is FAAAR simpler to OPERATE than a TW game....

Now thats funny. . Have you ever played chess?

Kasparov can't play TW game better then most b/c of operational complexity? Wouldn't the reason more likely be b/c we are more familiar with it and able to use all its exploits lol
....so funny....

[the_mango55]

Now thats funny. . Have you ever played chess?

Chess is indeed far simpler than a total war game. In chess there are only a couple of dozen moves you can make at any given time. In a TW game there are hundreds of thousands.


Look at it this way, STW and MTW had a risk style campaign map, much simpler. People praise the way the AI worked in the campaign of that game, it sent larger stacks and was often more challenging than AI in the newer games.

Did CA just decide to make their AI stupider for the hell of it? No, the map got more complex, thus harder for the AI to handle.

[Phier]

Take a board of chess and calculate the PER SECOND needed variables....(just a hint there arent any there is a thinking time)
And then look at the TW battles and calculate the per second variables (hint hint REAL TIME)....
Plus I doubt that Kasparov would be as good at a TW game as the best TW player....
Chess is FAAAR simpler to OPERATE than a TW game....

Take a look at what happens often in TW battles (hint the AI does suicidal things)

TODAY I was fighting an equal sized army (hint that means the number of men was approximately the same on both sides)

EXAMPLE I was assaulting buildings with his men in them, his cavalry can not fight in buildings but the AI sent them anyways (hint, this is stupid)

RESULT my men took the building and then SHOT the CAVALRY which did not move until they ROUTED (hint, this is really really stupid)

IF they had a simple algorithm such as IF under attack AND can not reach enemy THEN MOVE OUT OF THE LINE OF FIRE it would have been much better.



I'm not looking for Kasparov, I'm just trying not to get Cletus.

[Cayo_Julio_Cesar]

OP ppl just want a damn decent AI, obviously is not the damn AI of "I Robot", they meant lots of alogrithms (sp?) and sutff like that.

MTW1 battle AI and sometimes cmapaign AI (switzerland) was a pain in the there, people want that back.

[Multiball]

And yet, a computer can beat the best human chess player. All I'm asking for is for some if/then programming, I'm not looking to discuss philosophy with the game.

chess is limited in its scope, each player is equal at the start and can only make so many moves, If you program in every single move and program in any move that can follow it and make the computer select the best move every time, the computer will min most times, if you then give the computer to plan ahead many moves including what it will do depending what the possible moves the player will make you can make the computer near imposssible to beat...

Also all the chess computers have been supercomputers.. not 3ghz single cores with 1gb of ram and a 256mb graphics card which is the minimum for this game. So all the AI has to work based on this spec... And you hear people complaining about AI turns taking too long? do you people want your AI to do anything? As it has alot more to think about than a chess computer.

PS... Im not saying the AI is good in Empire Im just destroying your chess argument.. its not AI its a program that picks the best move based on cold hard inputed facts. The "AI" in Empire needs work but it would take weeks of processor time to make it react in even a half competent manor half the time as it accessess everything it has to before deciding its move.

[Phier]

chess is limited in its scope, each player is equal at the start and can only make so many moves, If you program in every single move and program in any move that can follow it and make the computer select the best move every time, the computer will min most times, if you then give the computer to plan ahead many moves including what it will do depending what the possible moves the player will make you can make the computer near imposssible to beat...

Also all the chess computers have been supercomputers.. not 3ghz single cores with 1gb of ram and a 256mb graphics card which is the minimum for this game. So all the AI has to work based on this spec... And you hear people complaining about AI turns taking too long? do you people want your AI to do anything? As it has alot more to think about than a chess computer.

PS... Im not saying the AI is good in Empire Im just destroying your chess argument.. its not AI its a program that picks the best move based on cold hard inputed facts. The "AI" in Empire needs work but it would take weeks of processor time to make it react in even a half competent manor half the time as it accessess everything it has to before deciding its move.

See my rebutal post to this idea. I'm aware that chess is different than TW, again I'm just expecting some logical reactions by the AI. Also a modern computer can do quite a bit. Back when chess was all the rage in computer gaming my desktop would have been a super computer.

[pdanders]

chess is limited in its scope, each player is equal at the start and can only make so many moves, If you program in every single move and program in any move that can follow it and make the computer select the best move every time, the computer will min most times, if you then give the computer to plan ahead many moves including what it will do depending what the possible moves the player will make you can make the computer near imposssible to beat...

Also all the chess computers have been supercomputers.. not 3ghz single cores with 1gb of ram and a 256mb graphics card which is the minimum for this game. So all the AI has to work based on this spec... And you hear people complaining about AI turns taking too long? do you people want your AI to do anything? As it has alot more to think about than a chess computer.

PS... Im not saying the AI is good in Empire Im just destroying your chess argument.. its not AI its a program that picks the best move based on cold hard inputed facts. The "AI" in Empire needs work but it would take weeks of processor time to make it react in even a half competent manor half the time as it accessess everything it has to before deciding its move.

That's part of the secret of Chess AI.. but it's more complicated than that... the trick is the algorithm for evaluating how good each move is... which is of course non-trivial

[akinkhoo]

the problem with AI atm is: 1. difficult to teach the program to "see". how do you descript it in logic is a 'left flank' for example? however it can be done as show in chess, you just need to divide the battlefield. 2. how to descript tactics, often it is dumb down to said, if cannon, charge cavalry into it. of course if you really want the AI to be smart you must explain to the AI to account of infantry near the cannons. it's doable, but it is expensive to hire a hundred AI programmer to descript in detail the tactics.

[Phase]

I am still waiting to hear something that really isn't obvious from the start.

OK, let's just play a game where the opposing armies just sit there because the AI doesn't compete with a human! Yay!

[upsettingshorts]

There are also limits to the amount of possibilities the AI can calculate in a second.

I think the AI is likely to be at its best when it has the initiative, when the human has to react to the AI's movements. But I can't say I've seen any particular difference between offense and defensive AI skill.

Maybe in naval battles, with the weather gauge they seem pretty solid - if unremarkable.

[Sacerdos]

I think the AI would be better if we were still on the Risk style map that Medieval and Shogun had, just updated with 3D graphics.

[GeorgiaPeanuts]

Uh real AI is a reality in research institutions. My university GaTech made an AI demonstration that featured two AI entities that could interact with a player based on his words alone. The premise was to either get them to break up or to solve their marriage problems whatever you decide to do. http://www.interactivestory.net/

Now the AI in TW is AI as well, because AI is a blanket term for programming done to mimic human intelligence. It is not very good because it can't learn new possibilities based on what happened before.

The AI has at its fingertips the general arrangement of the player's army, composition, and how it was positioned on the battlefield. It should be able to draw conclusions after the fight is done and store them in some sort of database. Even as simple as trying one of its AI formations and if they get destroyed it could note that and anytime it encounters similar player armies it would try a new AI attack routine.