From newshub.ccs.yorku.ca!torn!cs.utexas.edu!convex!darwin.sura.net!spool.mu.edu!umn.edu!csus.edu!netcom.com!stas Wed Sep 23 16:54:28 EDT 1992
Article 6980 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!torn!cs.utexas.edu!convex!darwin.sura.net!spool.mu.edu!umn.edu!csus.edu!netcom.com!stas
>From: stas@netcom.com (Stanislav Malyshev)
Subject: Re: My definition of intelligence
Message-ID: <hb4n6km.stas@netcom.com>
Date: Sat, 19 Sep 92 07:46:40 GMT
Organization: Netcom - Online Communication Services (408 241-9760 guest)
References: <iordonez.715293767@academ01> <1992Sep9.025119.15500@uwm.edu> <1992Sep9.032813.19773@uwm.edu>
Lines: 56

In article <1992Sep9.032813.19773@uwm.edu> markh@csd4.csd.uwm.edu (Mark) writes:
>   Intelligence is a qualitative assessment of something regarding the
>complexity of its behavior.  Nothing more, nothing less.  No sudden emergence,
>no lights turning on, no POOF I'm alive!
>
>   It's an attribute that can apply to problem solving behavior.  You can say,
>for instance, this is an intelligent problem solver.  It could apply to
>navigation, then you'd have an intelligent navigator in direct proportion to
>the intricacy of its operation.  It could apply to language, you could then
>be referring to the intelligent use of language.
>

The best analysis should indicate intelligence then, not the intricacy
of the process, though the two often go together.

>
>   It also implies simplicity.  For instance, a naive or "simple" proof
>generator would appear to be textbookish and would be producing oversized
>output.  An intelligent proof generator operates in a far less naive and far
>more intricate manner than thus will almost always produce correspondingly
>more concise proofs.
>
>    So in cases like that conciseness is the hallmark of intelligence.  Thus,
>for example, the more intelligently written machine program is the smaller and
>simpler one.


Perhaps it's better to say that a system that achieves its goals in a prompt
and sensible manner is more intelligent than one that does not.  (In this
case we'd have to consider the 'goodness' of the goals too, for we don't
want them to be simplistic though easily achieved.)

Intelligence should be related to the goodness of the analysis you mentioned
in the beginning... I imagine concise proofs would follow as a result
of clearer understanding that better analysis would provide.

And speaking of program sizes and their complexity, a counter-example that
comes to mind is the bunch of matrix-multiplication algorithms that outperform
the algorithm that uses the "natural" and simple way of multiplying matrices
considerably.  The standard one is _theta(n^3)_, and the fastest one is
somewhere around _theta(n^2.3)_ or so.  Of course, the faster algorithms are
bigger and are more complex.  
This doesnt have anything to do with intelligence in general, only with that
of the authors.  But I thought I'd mention it in any case.

Cheers,

Stan
-- 

-------------------------------------------------------------------------------
Stan Malyshev		|    Open up the windows and let the fresh air out,
stas@soda.berkeley.edu	|    said the television to the shackled children..
stas@netcom.com		|		- King Missile
-------------------------------------------------------------------------------



