Insight and Reason - Natural and Artificial Reasoning - page 5

The implication is that if we cannot devise a test that will distinguish between a
thinking person and a machine that imitates thinking then for all practical purposes
the machine can be said to think.
Rather than deal with the general notion of thinking, let us limit ourselves to
considering an important by-product of thinking: the display of intelligence.
Intelligence has practical consequences that can be observed and tested, for ex-
ample, through the ability to solve problems. So we will change the imitation game
so that the objective is to determine the question, “Is the machine intelligent?”
If we can identify the practical effects of intelligence we are then in a position
to test if X or Y displays these effects or not. But if we can identify features of
intelligence where the answer will be “X is intelligent,” or “Y is intelligent,” then
we no longer need to distinguish between them, since the answers are not exclusive
anymore (i.e. machine and not machine). We have already crossed the intellectual
boundary that rejects out of hand that anything other than a human can be intelligent.
The only reason for such a comparison between man and machine is to act as a
standard so that we might calibrate an answer to the enquires: “Is this intelligence
like that displayed by a person?” or “How much intelligence is shown?” or “What
are the limitations of (this) particular intelligence?”
We now no longer need play the game; we only need tests for intelligence; tests
that depend upon well-identified features. Further, we no longer have to show that
machines are of equal intelligence or have the same breadth of intelligence as a
person. We merely have to find a distinction in observable behavior that separates
intelligent behavior from non-intelligent behavior. We don't even need to detect every
kind of intelligent behavior, only to recognize a form of intelligent behavior. In this
case we might expect to provide some scheme that will give a measure, a grade or a
limitation to the observed intelligence.
Of course, once given such tests, the scheme can be applied to a collection of any
interacting objects other than a recognized machine.
So what are the practical effects of intelligence?
Rather than pursue the unusable definitions of intelligence that take the tortuous
route to avoid the notions of animal behavior and unconscious actions but include
some unobservable aspect of people, let us ask the pragmatic question, “What do we
want intelligence to achieve?”
Let us consider one possible answer through the measurement of intelligence as
carried out by psychologists on people where this decision has already been made
and agreed upon. This consideration may provide a possible game plan.

1.3 Intelligence Tests

We should note that according to one of the early intelligent tests leading advocates
(Eysenck 1962 , 1966a , b ) intelligence testing:
•

has no firm scientific basis.

• is, however, successful in its application.