COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	life[w83,jmc]		Gedanken Experiments with Life World Physicists
C00008 ENDMK
C⊗;
life[w83,jmc]		Gedanken Experiments with Life World Physicists

Abstract: J. H. Conway's Life cellular automaton system admits
large configurations of cells that function as
self-reproducing universal computers.  We discuss programming
such computers to be Life World "physicists".  Such physicist
systems will vary in their ability to interact with the world,
in their programs for doing physics, and (if there are many
in one Life World) in their
ability to interact with each other.  We discuss Life World
physicists in order to model philosophy of science and
epistemology issues.  We are interested in what structures
and programs will enable them to discover that the fundamental
physics of their world is the Life cellular automaton.
We hope to clarify issues in the theory of meaning and the
theory of truth by modelling them in a world simpler than our
own.

1. Introduction

	E. F. Moore (1956) studied what could be found out about
a finite automaton by a someone who must treat it as a black box,
i.e. give it sequences of inputs and see what outputs come out.
The typology of such simple systems shed light on questions of
epistemology and artificial intelligence.  Conway's Life World
permits carrying this investigation further, because here we
can model "physicists" studying the world from within and ask
what they can discover about it.

The Life World

Interesting configurations

Computing elements

Self-reproducing computers

Life world physicists

	#. There are many ways of constructing Life World computers.
If the program does not have any interaction with the world outside
the computer and it isn't given a priori information about the world,
it can never discover anything about it.  A Life World computer can
be simulating a quite different world, since it is universal.

	#. It is rather easy to provide Life World computers with
effectors.  The most obvious is a controllable glider gun.  Sensors
are more difficult.  It is rather like a world in which the only
means of transmitting information is with machine guns.
Interaction with random configurations is likely to destroy the
physicist.

	One possible solution is based on a shield.  This would
be an active configuration that provides information continually
on its state, so the physicist can tell if it is being damaged.
It would be best if the shield would have a high probability of
being able to advance into random configurations.  Of course,
we can't suppose it capable of advancing against arbitrary
configurations, because it might be pitted against another shield.

	If shields that can advance or at least hold their own
are not possible, we can imagine a shield that can keep penetration
slow enough so that the physicist can retreat and build more
shield faster than the original shield can be penetrated by
the randomness outside.  This assumes that there is open country
in which to retreat.  I don't know if the open country can consist
of all blanks or whether it has to be a periodic configuration
suitably prepared.

	The third possibility is that there is no protection against
random life world configurations, and that the physicists can survive
only in Life Worlds have certain regular characters.  This is not
so bad; humans can't survive in the center of a star.  However, if
the requirements for survival are too strict, the physics may be hidden.