Re: TIB: Artificial Intelligence
[Prev][Next][Index][Thread]
Re: TIB: Artificial Intelligence
In a message dated 2/1/00 9:30:59 PM Eastern Standard Time, ETTamme1@aol.com
writes:
> Well i didnt originally write the post, but ill throw in my two cents
> anywho.
> You can make computers "learn" basically by determining which moves
cause
> it to win aginst you and which cause it to loose. Now obviously if it did
> the same moves everytime you would beat the crap out of it after the second
> game. So, what you do is after the computer has lost or one you create a
list
> or string (or anything permanent) and add a plus value to the segment
which
> represents that move so you are actually weighting the moves. Then you
> combine this with the standard AI that runs, and your computer will
> eventually "learn" your playing style. And if youve done a good job
> developing your AI and weight system your computer should be kicking the
crap
> out of you. But there is a drawback to this type of intelligence. It is
> specifically tailored to one person, so if a person with a more aggressive
or
> just different style comes along and plays the game he/she will think the
> game is too easy. However you can create individual profiles for seperate
> users that way the co!
> mputer will grow with your ability/ changes in style so as you get better
> the computer will too. So you can sort of creat an infinite game.... It's
> cool, but it can get HUGELY complex before you know it so be prepared.
> ~Eric
Exactly, an AI tailored to one person isn't so useful. The idea behind the
way I do it is that it figures out why it lost the game (not as hard as you
might think) and makes a note.
Grant Elliott