Physical (Analogue) computation and AI

[Sunny Bains]

The following paper, due for publication in the AISBJ later this year,
may be of interest to those concerned with analogue vs. digital
(vs. Turing machine) issues. Though written at a fairly abstract
physical/mathematical level, the new model's connection with real
machines, and particularly analogue neural networks, is discussed at
the end.

http://www.sunnybains.com/AISBJ2003.pdf

All comments gratefully received...

Sunny Bains
Imperial College London/The Open Universty
http://www.sunnybains.com

*****

Intelligence as Physical Computation

Abstract
In this paper, we describe a model of physical computation that allows
us to understand how embodied intelligent agents can simultaneously be
considered to be objects operating under the laws of nature (or
physics) and information processing devices. Using this type of
analysis has two major advantages. First, it allows us to make
concrete those issues relating to the relative merits of using
analogue values or symbolic representations such as numbers: our
analysis allows this longstanding point of contention in artificial
intelligence to be transformed from a question of philosophy to one of
physics. Second, it gives us a framework that should eventually allow
us to produce design rules that will enable our
artificially-intelligent (physical) agents to make better use of
influences from the physical environment. As should be expected, for
well-defined tasks, the results of this analysis are identical to
Shannon’s information theory.  However, for large, multi-functional
systems with ill-defined roles, the new model provides a novel way of
thinking about the complementarity of hardware and software.