Information Technology Reference
In-Depth Information
2.6
Lookup or Generate?
In Plato's Meno (c.427 to c.327 BC), there is a description of Socrates illustrating
to Meno his belief that all knowledge is somehow already within us all. He takes a
slave boy who has been brought up in Meno's household and proceeds to question
him in such a way that the boy 'discovers' for himself the relationship between the
area and the sides of a rectangle. Socrates comments:
You see, Meno, that I teach him none of these things which he (
the boy
) asserts; I only ask
him questions. And now this boy imagines that he knows of what length the lines are which
contain a space of eight square feet.
Meno was not convinced at this point but after a further set of intensive questions
from Socrates to the boy Meno agreed with Socrates' final statement:
If the truth of things therefore is always in the soul, the soul should be immortal. So that
whatever you happen now not to know, that is not to remember, you ought to undertake with
confidence to seek within yourself, and recall it to your mind.
The reason Socrates presumed that we have within us all knowledge is because it
is not easy to detect the difference between mechanisms that 'look up the correct
responses' from mechanisms that 'generate a correct response'. I would go further
and state that in principle it is
impossible to tell the difference
. However, in practice
any finite mechanism will have limitations that make storage of predefined knowledge
very unlikely. From a modern point of view, where a finite brain bound us, we do
not have access to the virtual storage of an infinite soul.
On the other hand, an infinite amount of a certain kind of knowledge can be
'stored' within a generator. For example the two times tables can be extended indef-
initely through the mechanism of multiplication. This suggests that the generators,
or at least the components of the generators, must be predefined. In this sense the
notion of 'recall' as a substitute for intelligent behaviour being born within us from
the beginning is correct. The specialisations of comprehension, word, number and
space associated with intelligence (Thurston) is supported from this analysis. The
underlying intelligence mechanisms of abstraction and abduction are only workable
if there is some predefined set of generators that invoke perceptions and concepts.
One of my research students, Mohamad Zakaria (
1994
), created a model, based
upon the above view of intelligence. This model is limited to a world of numerical
series. The abstraction of the basic features of the numbers was fixed to be the
number value. The range of concepts involved different kinds of curve fitting and
series. The criterion for success was based upon a notion of simplicity, and each
hypothesis generator was associated with its own notion of what that means. The
first hypothesis to satisfy its criteria for success was offered as a solution.
Thus, if you take the series:
181064_
The first successful hypothesis might be:
32
x
3
x
3
3
3
x
2
y
=
18
−
+
−
