History: AI -- mid-50's , Expert Systems -- early 70's

Definitions:

1. The study of how to make computers do things that, at the moment, people do better.
2. The study of how to solve exponentially hard problems in polynomial time.

What does one expect from a machine to act "intelligently"? to be an expert? (See figure 1.1, p 5)

One wants a system to "do it all"...

early attempts (60's) :

find general methods to solve broad classes of problems

problems: the more classes of problems a single program could handle, the more poorly it handled individual cases

later methodology:

write individual modules

use union

problems:

"problems " are not separate and identifiable

Knowledge issues

1. AI "subgroups"
   1. to understand and simulate human intelligence (psychology ... cognitive science)

   2. to produce intelligence by whatever means (robotics)

2. Philosophy
   IS the human a machine?

   Can human intelligence be explained in mechanical terms?

3. How do you know if you have achieved AI?
   turing test ... ? Searle's Chinese Room...

   Understanding (complete empathy, cognitive understand, make sense)

• Turing Test - valid? ...Eliza (Weinzenbaum)

• Analog vs. Digital Systems

  (brain) (computer)

  (slide rule)

• Mathematical

  - probablistic scheme (Tversky and Kahneman)

  - logic

• Engineering

  computing elements (minimal basic computing elements?)

  *perceptrons

  *neurobiology (what is capturable)

  (too much gap between low & high level processes

• High level processes (Top down) (could start here)

  - statements of problem solving

  - symbolic structures

The Physical Symbol System Hypothesis

A physical symbol system has the necessary and sufficient means for general intelligent action.

___________(what is this saying?)__________________________

Principles of AI programming(like Software Engineering)

• explicit naming
• modularity
• graceful degradation
• least commitment
• exploit natural constraints

heuristics: rules of thumb (general short cuts)

simplifications that effectively limit search for solutions

Theory
Type 1	Type 2
look for underlying math of problem... then get computations	theory useless ... still need to make work!
"neat"	"scruffy"

Is there an underlying theory of elegant semantics?

(e.g., is there an "equation" for chess) (type 1)

Theory -- Marr's approach

1. define problem what

2. analyze problem how

3. choose method system and implementation

   (vs choose method first..."Procrustean Bed")

**** AI questions : how should knowledge be

represented so that it can be accessed

& &

organized (representation) used (control)

Point:

Knowledge and Proper Representation (e.g. choice of data structure) make a problem easier to solve.

Initial state: 8X8 board, 30 white 32 black left on checkerboard grid. (top left and bottom right are white)

We have tiles, half white/half black

Final state: Does there exist a covering of 1X2 tiles (search done through assertions not problem space)

Read Chapter 1...Different ways to represent a problem

When doing problem solving, always consider:

What am I addressing?

Why is it good?

How is it useful?