Hmmm when you talk about AI I think you need to distinguish between AI's for real time combat (the "Headshot" bunch) and strategic AI's. RT AI's defintely need to be dumbed down as they have total awareness of a battlefield and in a game where "making the shot" is part of gameplay, they are too perfect becuase it is all about calculation.
In a strat game it is totally different. I think once you get past tic tac toe, checkers and chess, there is no need to "dumb down" the AI. The eternal struggle is to get a Strat AI to be able to fuzzy logic its options. We all have played strat games where the AI does something well but then does it in a situation where that is totally stupid. This is the ugly edge between "routines" and "understanding". And while your routine can be very sophisticated in what parameters it takes into account, in a wide open game, it is still vulnerable to the occasional "not taking reality into account" move.
And as for pathfinding...well that is a different ball o wax again...but since hundreds of millions have gone into 2 cross country AI races over the past 2 years and only got winners this year. I think we can say that telling a computer player how and where to move a unit around a tree is trickier than most people think :help:
|