An AI that learns from its mistakes.

[Guest crd_polaris]

Wouldn't it be cool if the AI in RTS games could learn from you? Maybe even add a trainer mode where you could teach it statagies. That way each game against the computer would be unique and never stop challenging your skills.

[Gaming Nut]

This has been talked about for years and still has yet to be shown in any game their are just to many thing to write into the program. a few examples if the A.I. decides to rush it must decide when but not wait long. It’s need to decide what the best units to use are dose the enemy have troopers, mechs, Air. Are their fixed defences should he go for his army or drive for the workers? Should he expand or build D should his D be made of units or buildings. There are a million variables and programming an A.I. like that might be do able but would take a long time. thinking about this really makes you appercate how powerful your brain is.

[Artoo]

programming something like that would take so long because you would have the computer go recall what you did and try to correct itself. That would be an immense program to write. The computer would have to have a fast processor in order to really compete. It would be like the chess game where the opponent knows exactly what to do to counter what you were doing. Really, Really, Really, Really, Really ANNOYING, because you would never win unless the computer was programmed to make a mistake.

[Tie Guy]

Originally posted by Artoo

It would be like the chess game where the opponent knows exactly what to do to counter what you were doing. Really, Really, Really, Really, Really ANNOYING, because you would never win unless the computer was programmed to make a mistake.

 

But even Chess Supercomputers can be beaten. A computer knows exactly what it needs to counter everything in the game, and even if it could learn humans would always out manuver it. Why? because even though a computer may know what to do instantly, it is still bound by the rules of the game, and the AI code it runs off of. There will always be a new strategy that the comp doesn't know, or is incapable of doing, and, even though it may build the perfect army, it cannot do anything that you cannot, and the game rules by which it must obey insures that you can always be at least on level terrain with the comp. And after that it is simply chance in a 1 on 1 fight.

[Artoo]

but waht if it remembers over the games:eek: a learning/remembering computer! Also what about Deep Blue?

[Rogue15]

Originally posted by Tie Guy

 

it cannot do anything that you cannot.

 

ehehe try telling that to my Magsealed Assassin Droid who fires rapidly and has alot of health, while the player is only armed with a lightsaber? mwahahahahahahahahahahahahahahahahah

[Tie Guy]

Originally posted by Rogue15

 

ehehe try telling that to my Magsealed Assassin Droid who fires rapidly and has alot of health, while the player is only armed with a lightsaber? mwahahahahahahahahahahahahahahahahah

 

I'm not talking about MOTS 15, i'm talking about GB. The rules of the game must be followed by both the computer and the player.

 

Even in MOTS, any situation is winnable by a good enough player. that robot, even if it could learn, would eventually fall to a good enough player, thats the nature of a game. Computers aaren't invincible, they will never be smarter than their creators.

[Artoo]

Computers aren't invincible I know but if a computer learned by watching then it learned by watching another person play then another, then another, then eventually it will be able to beat those players because it will absorb the strategies and then use the correct one in the correct situation it would win. But then you say a new person comes and defeats the computer using an unpreviously played strategy. The computer would learn and it would figure out the counter strategy therefore defeating the oppponent in the next game if the opponent uses the same strategy. It would learn until there is nothing left to learn and then it could defeat anybody until a new strategy was forged. Then it would learn again and so on. So the computer would eventually win.

[Guest GhostKeeper]

Computers AI will not be able to "think" forever (hopefully). AI will never be as smart as a human, humans can always do something the computer hasnt been programmed to do. Unless there was a huge test or something where millions played against an AI and the comp sucked up everything, even then you could still find a way to beat it!!!!! Plus would it be any fun to play a comp that would kick ur ass in 3 mins ?