Transcript Are Expert Systems Really Experts?

Introduction to Expert Systems
Università di Salerno: April 2004
Are Expert Systems Really Experts?
Dr David Anderson
University of Portsmouth
Different Approaches
Cognitive Simulation
The A.I. Approach
Strong versus Weak A.IT
The Chinese Room Example
Are Expert Systems Really Experts? Slide 1
Introduction to Expert Systems
Università di Salerno: April 2004
The Cognitive Simulation Approach
Point is to throw light on the (human) mind.
Programs serve to test theories .
Models are judged by the degree to
which programs can reproduce
human levels of performance
N.B. The machine must use the same
methods we use
Are Expert Systems Really Experts? Slide 2
Introduction to Expert Systems
Università di Salerno: April 2004
The Artificial Intelligence Approach
Aims to build machines that can do things
which, if done by us, would require intelligence.
Lack of concern with how computers perform
their actions
What computers can do is all important.
Are Expert Systems Really Experts? Slide 3
Introduction to Expert Systems
Università di Salerno: April 2004
Searle’s Distinction
Weak A.I.
The principal value of computers in the study
of mind is for the formulation and testing
of hypotheses.
Strong A.I.
Programs are not tools to test
psychological explanations
Program are explanations.
“Strong A.I. is playing at science.....
...(I seek) the relentless exposure of its
preposterousness.”
Are Expert Systems Really Experts? Slide 4
Introduction to Expert Systems
Università di Salerno: April 2004
The Chinese Room
The Elements
A room closed off from the outside world
except for ......
A communications hatch
A dumb waiter
Bushel Baskets (2 Sets)
Full of Chinese Symbols
A Rule Book (in English)
Rules for correlating the two sets
of symbols
You!
Are Expert Systems Really Experts? Slide 5
Introduction to Expert Systems
Università di Salerno: April 2004
The Test Itself
A message (in Chinese) is passed into the room
You compare the symbols which make up the
message with the symbols in the first set of
bushel baskets
Using the rule book, you select a set of symbols
from the second basket and arrange them as
directed.
The newly arranged symbols are passed
out of the room.
Repeat as necessary
Are Expert Systems Really Experts? Slide 6
Introduction to Expert Systems
Università di Salerno: April 2004
The Test from the outside
A question is asked in Chinese.
An appropriate Chinese answer is given.
Another question is asked.
Another answer is given.
Conclusion:- There is someone in the room who
understands Chinese
Are Expert Systems Really Experts? Slide 7
Introduction to Expert Systems
Università di Salerno: April 2004
The Test from the inside
Gibberish is sent into the room.
The rules help you select and arrange other gibberish
and pass it out of the room.
Conclusion:There is no-one in the room who
understands Chinese.
Are Expert Systems Really Experts? Slide 8
Introduction to Expert Systems
Università di Salerno: April 2004
Variations on a theme (1)
The Systems Reply
Understanding is a property of systems as a
whole.
The Robot Reply
House the program in a robot
so that it can do more of the
things which humans can do.
The Brain Simulator Reply
Simulate the actual sequence of
neuron firings of a native
Chinese speaker.
Are Expert Systems Really Experts? Slide 9
Introduction to Expert Systems
Università di Salerno: April 2004
Variations on a theme (2)
The Other Minds Reply
If you are right how can we ever say anyone is
thinking, feeling, etc.
The ‘Hunk of Junk’ Objection
Only something that has the
same causal powers as brains
can think.
Simulation
Why should anyone ever think
that a simulation of intelligence
should actually be intelligent?
Are Expert Systems Really Experts? Slide 10