Quote:
|
Originally Posted by sevenwarlocks
I say this because to have an effective plan, an AI need not be perfect, or even extremely complicated. The most important thing is that the AI have some plan. By which, I mean, a long-range plan that it will stick to - at least for a time. AI's that react on a turn by turn basis will never be effective. As a chess-master once said - it is better to have a bad plan than no plan at all.
|
Well, I guess all AIs have some sort of plan hardcoded into them.
However an AI that just gets some goals and formulates the plan themselves reacting to the situation is a whole different matter and far from easy imo.
It might be not even possible, take game theory, most of the algorithms designed to lead to the perfect strategical behaviour fail miserably when put against normal human behaviour. It's not even that humans play better, but they don't necesseraly take the determinable "best" step invalidating the underlying algorithm.