Artificial Intelligence is a broad field. There are some untouched area in it. What makes an animal (say charector here) intelligent??? decision making ofcourse... But true intelligence is only possible if we immitate the real world. Traditional Procedural ways of infering a situation are dumb - at least in RTS games they fail to give challenge. I have been working on algorithms only and no real code. so can anyone tell me 1)if it will be wise to spend time on finding new algorithms for experience & inference based AI or it is available anywhere.

Ive seen discussions on computer analysis to build learning set to create behavioral logic, but it usually requires so mauch computing assets that the work is done off line (of the game) and new data logic is produced from experimental behavior sessions or watching a human play.

The Experience has to be recorded (not hard if you have alot of capacity to record/store every thing that happens).

Memories have to be extracted/built from the raw data -- classification/filtering of data and analysing a situation to determine what factors were significant (temporal aspects etc...)

Inference -- trying to decide what actions caused what results to happen -- allowing prediction if the same pattern happens again.


Computers still have a way to go powerwise before they can do such things realtime in a complex simulation (the learning operation).

But offline processing to build the behavioral logic (run as a static system) can be done now (though expect a much higher percentage and total game processing power to be dedicated to run it)