Aug 23, 2006
-------------

- Introuction to 4804

- What the course will be about

- What is AI?: many perspectives
	2X2 box comprising:
        - think like humans (cognitive science approach)
        - think rationally (laws-of-thought approach)
        - act like humans (turing test approach)
        - act rationally (rational decision making approach)

- Textbook
        - AIMA: Artificial Intelligence: A Modern Approach, by
          Stuart Russell and Peter Norvig

- AIMA primarily dwells on the last theme
        - make sure you get the green book, not the red book!

- AIMA's focus
        - design and implementation of intelligent agents

- Example agent tasks
        - spotting ads on televisions
        - playing chess
        - bagging groceries
        - driving or navigating a vehicle
        - getting an A in 4804

- Introduction to agents
        - rational agents do the right thing
        - map percepts from environment to actions taken
        - they have a notion of a performance measure
                - e.g., minimize #(squisshed veggies)
        - can even be sadistic (e.g., maximize pain!)

- Rationality and what it means
        - goes with autonomy
        - does not go with omniscience
	- does not go with blind faith

- Types of agents
        - rote agent (table-driven)
                - intelligence was primarily with
                  agent builder, not the agent
        - reflex agents (condition-action rules)
        - agents with internal state (a little bit of
          thinking goes on)
        - goal-based agents (have a sense of purpose)
        - utility-based agents (are calculative)
	- learning agents

- Reading assignments
        - finish Chapter 1 and Chapter 2 (on your own)

- Primary threads in 4804
        - search
                - AI "lifeline"
                - almost any AI problem can be formulated as
                  search
                - search compensates for lack of knowledge
        - knowledge representation and reasoning
                - knowledge compensates for lack of search
        - learning
		- given positive and negative examples
        - reinforcement learning
                - learning without a teacher
                - in some ways, envelops AI itself
                - can be viewed as a restatement of the AI
                  problem

- Example AI software systems in each of the above categories

- More on CS 4804
	- logistics
	- lots of thinking puzzles
	- lots of programming
		- pick your own language and wisely!
	- lots of math
		- I can simplify it but can't make it go away!