Path: utzoo!attcan!uunet!mcsun!ukc!edcastle!aipna!cam
From: cam@aipna.ed.ac.uk (Chris Malcolm)
Newsgroups: comp.ai.philosophy
Subject: Re: Emergent Properties
Keywords: chaos, science, prediction
Message-ID: <3369@aipna.ed.ac.uk>
Date: 25 Oct 90 17:46:58 GMT
References: <1990Oct12.214636.7945@ncsuvx.ncsu.edu> <30@tdatirv.UUCP> <1990Oct19.201604.7280@ncsuvx.ncsu.edu>
Reply-To: cam@aipna.ed.ac.uk (Chris Malcolm)
Organization: Dept of AI, Edinburgh University, UK.
Lines: 45

In article <1990Oct19.201604.7280@ncsuvx.ncsu.edu> fostel@eos.ncsu.edu (Gary Fostel) writes:

>    Chris Malcom, in a different post, dismissed Michael Bender's 
>    query of "How can we build something we can't define" by claiming
>    this was done all the time in research prototypes.  This is also
>    sloppy thinking.  (It may well have been intended as a joke, but I  
>think the point is serious.) Prototypes are not objects which 
>    satisfy any definition; they are instances in search of a definition.
>    If one is unable to define a property a system is supposed to have,
>    and someone claims to have built a research prototype that has that
>    property, there is no way to check that the property is there. 
>
>    The purpose of science is (I thought) to search for descriptions
>    of the world, to be offered in terms of defined quantities, and
>    then to test those descriptions for adequacy.

That is a simplified rational reconstruction of science, and is a
suitable slogan for keeping teenage wannabee scientists in some kind of
order. Since it is also communicatively efficient, scientists, just as
do mathematicians, endeavour to present their results in accordance with
this paradigm; and, just as with mathematicians, it bears little
resemblance to the way they actually work.

For example, this news group has recently been full of complaints and
suggestions from people who -- motivated by precisely the view of
science you present here -- would like all AI research to stop until
someone has found a good definition of intelligence. When I said that in
constructing research prototypes we often found ourselves building
things we couldn't define I was perfectly serious. It is not uncommon
for a good research prototype to be the subject of heated debate over
precisely what it exemplifies for a number of years. The final
interpretation is sometimes -- even in the eyes of the inventor -- quite
different from the original intention. One of the distinguishing
features of a good research prototype is just this long-term
fruitfulness.

There is a great deal more to constructing a fruitful research prototype
than "to search for descriptions of the world, to be offered in terms of
defined quantities, and then to test those descriptions for adequacy."
I'm sorry that I can't specify exactly what it is, but I know it when I
see it!
-- 
Chris Malcolm    cam@uk.ac.ed.aipna   031 667 1011 x2550
Department of Artificial Intelligence, Edinburgh University
5 Forrest Hill, Edinburgh, EH1 2QL, UK