COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc] source
C00004 00003	.bb Laws of Motion
C00008 ENDMK
C⊗;
.require "memo.pub[let,jmc] source;
.cb EPISTEMOLOGICAL PROBLEMS OF COGNOLOGY

	Epistemology is the study of knowledge - what knowledge there
is to be had and the ways of getting it.  Cognology is the study of
intelligent behaviour in man and in computer programs.  Since
intelligence requires knowledge, cognology depends on epistemology.
It also depends on ⊗heuristics - the study of search for solutions to
problems.

	We recognize that the above point of view may be
controversial among people who do artificial intelligence as well as
among philosophers.  Arguments in its favor will follow the
presentation of the epistemological problems of cognology that we
have so far been able to identify.

.bb Laws of Motion

	To any person with a decent scientific education, there is
one obvious approach to representing laws of motion.  Introduce
a real variable ⊗t representing time, introduce a vector ⊗x of
variables representing the state, and express the laws of motion
by some combination of differential equations, integral equations,
and difference equations.
Let us now explain why we can't do that.

	First, what we know about the common sense world's laws of
motion doesn't take that form, and we aren't likely to discover
a complete set of laws in that form anytime soon.  Second, even if
had laws in that form, the information a person or a computer
can observe about the world won't usually take the form of the
initial conditions for such a law.  Third, even if we had the
initial conditions and knew the law, we probably wouldn't be able
to do the computing.

	Consider such a favorable example as the law giving the
consequence of spilling a cup of coffee on the desk.  The laws of
motion of a viscous but incompressible fluid might well serve for
predicting the immediate flow, but even they won't tell us that some
papers will absorb the coffee and others will be missed - let
alone which will meet which fate.
We can't observe the initial conditions well enough to initiate
the computation, and even this hydrodynamic calculation is probably
too difficult for most computers to carry the computation to the
point of determining which papers will be wetted.
Therefore, we and our programs must resort to qualitative laws
saying something like, %2"If a coffee cup is spilled on the table
in the general direction of some papers, some of them will
most likely get wet.