Stephen Goulter
School of Mathematics and Statistics, University of Sydney
The Turing definition of machine intelligence
Wednesday 28th, July 14:05-14:55pm,
Carslaw Lecture Theatre 273.
In this talk, the Turing definition of machine intelligence is discussed.
Issues of context, interpretation, duality and symbiosis arise.
The possibility that machines can produce a form of creative thought
which may well be beyond human beings does not seem to have been
considered seriously. Yet such a result may be based only on a prior
set of assumptions about the primacy of human thinking. This requires
closer examination.
In this sense, the role of prior knowledge, context and uniqueness
seem essential. The role of perception and of analysis are examined in
relation to creative synthesis. The connection with elementary
catastrophe theory as given by Thom is noted.
It is argued that these human notions of primacy in thinking are
really merely conventions, which do not seem to be justified on any
abstract grounds. Several arguments are advanced to support this view,
some quite abstract. Other arguments are based on historical or
statistical reasoning.
A schematic program is suggested by which a machine might be
considered to produce a form of creative thought. The approach is
simple, based on set of introduced symbols, some form of grammar or
syntax, and an alphabet which is capable of extension, combination and
nesting.
If test shows this point of view has some merit, then the original
definition by Turing can be seen as very provocative and
fruitful. Indeed this is the point of view which lead to this work
being done.