Quote (Minkomonster @ 26 Oct 2015 21:41)
Artificial Intelligence is so loosely defined nowadays, that a case can be made for any application that applies a procedural based ruleset to solve a problem. What truly established an application as deploying "artificial intelligence" is it's ability to reason beyond simple deductions. This is a breakthrough we have yet to achieve.
Most video games have a finite ruleset it abides by and based on the current "state" of the game, it will deduce the best action to take based on its known rules. Old fighting games and RTS computer players follow this pattern. However, these systems are not intelligent in the sense that they cannot reason with inference. For this you need to be able to learn. A computer AI which recognizes patterns of attack its opponent typically makes, and then alters its strategy to best combat it using inference would be an intelligent system. This would be an application that learns to fight its opponent based on its observed weaknesses. Like a human would.
Ideally, if a problem is present, the system should be able to recognize it and apply the best known solution. If a solution is not known, it should then make decisions based on learned heuristics. And from the data it learns from those results, it can refine its approach until an approximate solution is found. This is very difficult to achieve, especially in a video game.
Chess AI comes to mind as an answer to your question though. Take a look at Deep Blue.
Also, atleast in my class we define a software/game that try to emulate a human/animal comportement as artificial intelligence. Like "flocking" and swarm intelligence in Starcraft2.
But ye, i agree the line between scripting and AI is so thin. If one thing, at the end of this class I might be even more confuse about it then I was before.
Btw, I choose Starcraft2 because I will have enough content with the pathfinding to do my homework.
This post was edited by LiFeR on Oct 27 2015 01:02pm