From: tim@iss.nus.sg (Tim Poston)
Subject: Re: ADMIN: What is sci.virtual-worlds all about?
Date: Wed, 2 Jun 1993 13:22:06 GMT
Organization: Institute of Systems Science, NUS, Singapore



ib@ivan.asd.sgi.com (Ivan Bach) writes:

: Does an expert system that consist of a few thousand inference rules
: written in, for example, Prolog have the same information capacity as
: a human expert whose body consists of trillions of complex cells?  I
: don't think so.  As a matter of fact I know so, because I calculated
: the maximum information capacity of a fertilized human egg cell.  The
: complexity of an information system can be measured by measuring the
: complexity of a blueprint necessary to transmit all the information
: needed to construct that system.

"All the information needed" includes the geometrical and physical
laws that govern the pressures, tensions, chemical gradients etc.
that control the actual physical form in which the process set in
motion with the DNA-specified initial conditions, in the maternal
milieu, will finally result.  Very different from a blueprint, which
explicitly lays down (for instance) doorknobs 80cm from the floor.
Nothing in the DNA matches that kind of simple correspondence between
blueprint and system.

Informative power depends enormously on context.  In London, I can
give a taxidriver a single instruction and set in motion a very
complicated sequence of events.  Here in Singapore, a far less
complicated city, I often have to keep up a running stream of
instructions.

I agree that a few thousand lines of `expert' code have little chance
of matching a human expert; but I doubt whether quantitative
information theory has much to do with that.

Tim Poston