Wednesday, January 9, 2008

DPS906 Lecture 1: January 9th 2008

Professor: Catherine Leung
Class: S215249
  • Artificial Intelligence A Modern Approach (second edition) Russell, Stuart J. and Norvig, Peter
Going to be looking at two specific topics in this class. Physics and AI.

AI (see notes)

CS majors dont like the term AI... they are trying to push the term Autonomous agents. Many different fields of AI...
  • Natural Language Processing
  • Machine Learning
  • Knowledge Representation
  • Expert Systems
When applied to games, AI represents (and means) the way in which the computer controls NPC (non-player characters). The agent acts giving information about the "world" and makes decisions on its own.

First thing we will look at are Classical Game Trees. (we use tic-tac-toe to explain)
Trees are used to organize data, we want to use this organized data

In a game tree, we want to get to a "terminating state" (the game is over). To do this we build the tree from the root notes and make "what if" child notes. (Root: I put X in the middle, I put O in the corner...I put X "there" and the game is a tie). For each terminating state, we give a "score"

The agent wants to get the highest score possible, but it must assume that its opponent wants the agent to get the lowest score possible... This is called mini-maxing.

Cathy then showed us an example of tic-tac-toe

Tic-tac-toe is a perfect knowledge game... you have the ability to draw the game tree

At the end of class Cathy and I had a bit of an argument about whether Chess is a "perfect knowledge" game. Cathy says chess is not a perfect game because a computer cannot fully draw a game tree of it (not enough resources on current computers). I say Chess IS a perfect game because the game itself is not based upon computational "realities." Since you can see every piece in the game, given enough time and resources every "terminating state" can theoretically be determined.

No comments: