Class 2:
Some Lisp; The Nature of Agents; Search & Game Playing

Selmer Bringsjord

Misc Stuff

• Reminder: Class pictorial roster, w/ email addresses & web sites$\dots$
• We are learning to assemble toolkit $\ldots$ for building things
  • anchor learning with constant eye at (micro) construction
  • and games provide opportunities, e.g., GUESS WHO?
  • but also ponder your own domain, constantly

Programming Paradigms

• Procedural
  • e.g., Turing machines, Register machines
  • and familiar languages (e.g., Pascal)
• Functional
  • ``Pure" Lisp
• Declarative
  • only $\approx$ Prolog
  • Theorem Provers (for us: OTTER)

First Look @ Code

• NASA and immobots and Lisp_j
• Response_r
• Some Lisp
  • Ralph's interrogation re. net worth
  • ``virtual fences"
• On the delarative front: SECRETS in OTTER

The Agent Approach

• What is agent, put simply?
  • Skinner would be proud
  • See the picture: 2.1
• Rationality vs. Omniscience: the case of the 747 door
• Definition of Rationality_h

Rationality

That which is rational at a given time depends on four things:

1.

The performance measure that defines degrees of success.

2.

The percept sequence.

3.

What the agent knows about the environment

4.

The actions the agent can perform.

Definition of ideal rational agent:

Ideal Rational Agent

For each possible percept sequence, such an agent does whatever action is expected to maximize its performance measure, on the basis of the evidenced provided by the percept sequence and whatever built-in knowledge the agent has.

• What do you think? Is this acceptable?

We're concentrating on the program:

• agent = architecture + program

Intelligent Agents; Some Questions

• Why would an ideal rational agent need to have what R&N call `autonomy'?
• What would a PAGE description for
  • BRUTUS
  • SHERLOCK
    • SECRETS
    • GUESS WHO?
  • S³G BOT
  look like?
• WILLARD: A simple reflex agent

Environments $\ldots$

Properties of Environments

Accessible

The agent's sensory apparatus gives it access to the complete state of the environment.

Deterministic

The next state of the environment is completely determined by the current state and the actions selected by the agent.

Episodic

The agent's experience is divided into ``episodes," each episode conisting of the agent perceiving and then acting.

Static

If the environment can change while the agent is deliberating, then the environment is dynamic; otherwise it's static.

Discrete

There are a limited number of distinct, clearly defined percepts and actions we say that the environment is discrete.

Env.	Acc	Det	Ep	Static	Dis
Chess	Yes	Yes	No	Semi	Yes
Hyperbot	Yes	Yes	Yes	Yes	Yes
BRUTUS	?	?	?	?	?
SHERLOCK	?	?	?	?	?
S3G	?	?	?	?	?

• GUESS WHO? and the environment simulator$\ldots$
• GUESS WHO? and problem-solving agents (search: Figure 3.1)$\ldots$

Search & Game Playing

• Problem-Solving Agent (pseudocode)
• Example: Problem in 8-queens problem
  • Goal Test, Path Cost, States, Operators
  • explore, build on your own here; we'll return to it

Minimax, Chess, Go$\ldots$

• The minimax algorithm
  • Tic-Tac-Toe (figure)
  • Two-Ply Example (figure)
• Is there a perfect winning strategy for Chess? Yes!
  • but too computationally demanding, so evaluation function
  • and $\alpha \beta$ pruning (figures)
• ``Chess is Too Easy"
• What about Go and Poker, though?

  

Figure 1: Agent Overview.

  

Figure 2: Sample PAGE Descriptions.

  

Figure 3: Simple Reflex Agent.

  

Figure 4: Simple Reflex Agent With Internal State.

  

Figure 5: Examples of Environments and Their Clocks.

  

Figure 6: Basic Environment Simulator.

  

Figure 7: Simple Problem-Solving Agent

  

Figure 8: Search Tree for T-T-T.

  

Figure 9: Two-Ply Game.

  

Figure 10: Two-Ply Game Generated by Alpha-Beta.

  

Figure 11: Generalized Alpha-Beta.