I need to talk about types of AI used in video games in my research, so I would like to know which kind of AI could be found in a video game. I know there are : -Artificial Neural Networks -Finite State Machine -Fuzzy Logic -Genetic Algorithms -Ruled base AI -Path finding -Some probabilities -Scripted AI Did I have forgotten some ?

Here's a few more:

-Decision Trees
-Markov Chains
-Influence Maps
-Terrain Analysis
-Fuzzy State Machines
-Flocking Behaviors
-Steering Behaviors
-Obstacle Avoidance

Planners....

I know of Markov Chains from statistics. I do not think they are strictly an AI topic with it having existed since the turn of the century before last. errm 1900s b4 1910. They make use of them in video games?! I suppose it could be used to predict behaviour more simply than Neural networks. Hmm..

I do not think path finding or probabilities are a type of AI as they are tools stemming from graph and probability theory respectively, which have use in creating AI behaviour.

well...

Finite state machines are just as simple as a common switch statement. is that ai?
Genetic algorithms are just some random search crap that is as far from intelligent as you can get ...is that realy ai?
Decision trees are just a pile of if-then-else ...is that ai?
Influence maps just use simple rules from thermodynamics ...ai?
Obstacle avoidance can be as simple as "something is infront of me?!? lets take a step to the right" ..amazing isn't it?

You see.. what ever the method is stemming from, its all about HOW you use it. A simple if statment could be "AI" if it's used to make an agent smarter.

having a bad day here ..sorry

There's also the "Subsumption Architecture" ( http://ai.eecs.umich.edu/cogarch0/subsump/index.html ) which we apparently use in a sports/fighting game we're currently working on ( www.hoopworldgame.com ). I didn't implement the system, but it basically seems like a set of small modules ( similar to states, but without any high level logic ) in a hierarchy, and as the tree is walked it basically asks "could this child module possibly do something given the state of the game? Yes? Tick this module and have it access the situation, and possibly ask the same questions of its child modules if it has any, or possibly use the virtual control pad to do something.". It's a very reactive system, but higher level behaviour can be achieved by the overall arrangement of the hierarchy. I'm sure there's more to it than that, especially in how the modules interact with eachother, but I'm not the AI programmer so that's the best I can descibe, sorry. :)

@glSmurf

If I follow your last statement, this forum "Artificial Intelligence" has nothing to do in this web site...

Uhm? ..No, that was not my point at all.

Most of what I wrote was just some sarcastic crap I wrote as an answer to Daerex post. My actual point was that most methods (e.g. pathfinding) has their roots in something else than AI, but that doesn't change the fact that it is AI, despite that they also belong in graph theory, thermodynamics or simple boolean logic.

I had no intention to slight. See, I felt distinctions had be made and I will explain the logic behind their orderings.

First the original poster, jolyqr, asked specifically for types of AI used in video games. That at once reduces the scope of the set in question. With respect to A = {pathfinding, probabilities, Markov Chains} vs B = { Fuzzy logic, finite state machine, ANNs}, the elements of A do not on their own lead to nor wholly consist of an AI topic. All the elements of B originated or benefited from and developed extensively due specifically to AI. ANNs and FSMs as a whole lead to the control of agents which may have aspects of intelligent behaviour. The elements of A are on their own fully independent topics which require pairings with elements of B to evince AI. For example, probabilities or Markov Chains require or are utilized by decision trees or FSMs to genereate AI. Generally Elements of B are topics while those in A are methods. The two types are not equivalent.

Essentially my ordering criteria is how well a topic can stand on its own, how independent it is as a topic type in AI.

Quote: Original post by Daerax
I know of Markov Chains from statistics. I do not think they are strictly an AI topic with it having existed since the turn of the century before last. errm 1900s b4 1910. They make use of them in video games?! I suppose it could be used to predict behaviour more simply than Neural networks. Hmm..

I do not think path finding or probabilities are a type of AI as they are tools stemming from graph and probability theory respectively, which have use in creating AI behaviour.

You'll find info on most of these topics in the Russell and Norvig books, so you can be pretty much sure that they are considered part of AI. I don't think the question was ever supposed to rule out anything that has applications outside of AI.