Three Threads Of Squeak

Justin Walsh jwalsh at bigpond.net.au
Wed Nov 7 05:05:01 UTC 2001


> I'd be interested in more intelligent objects, but it's still two year
> before I do AI.

Doing  AI is fine, it can't hurt. I was interested in it in the late 80s but
chose Expert Systems instead. Experts Systems is a kind trade off, owing to
the fact that nobody could agree on just what AI was. (sounds a bit like
what I'm being accused of).
Expert System do work very well.
The prefered language for the Expert System builder is Prolog (although any
language will do if you enjoy pain)
Basically all you describe is the Knowledge Base.
The Inference Engine will allow various Tree Navigation Methods ie
Depth First, Breadth First, Hill Climbing and Least Cost. There are actually
no instructions if you use the "grunt" Depth First approach.
If you decide to take control of the Navigation yourself then you have two
instructions you can use The "cut" and the "fail". It was in those days
notoriously lacking in file handling capabilities and OO was unheard of.
That is all changing now.
Ok now to the bad news. At the Enterprise level (zero) there were no proper
design methodologies for Prolog. The best that they could offer me was
Binary Relation and Semantic Modelling. Both are dismal failures. Because of
that Prolog was dumped from some of the larges projects in Australia:
Telecom and the AMP.
So Smalltalk is not alone in that regard.
There is only one Methodology that is suitable for designing Hierarchical
Knowledge Bases that I know of and that is   http://ripose.com/
It treats the Expert System as a Business Expert (iterative level 1,2,3).
Whilst it assumes a higher Corporate Expert position (level 0).
Smalltalk Play (experiment) at (level 4) is missing both these levels.
If Smalltalk is happy at this level then it does not really need any of the
levels above it.
It is already, in respect of its Purpose, successfull.
Some disatisfied person ported Mike Tengs Prolog to Digitalk/V (PCXT) and
the V286.
It's great!
K Bolot http://minnow.cc.gatech.edu/squeak/1000
Prolog enthusiast  http://minnow.cc.gatech.edu/squeak/574
Best I stop here


----- Original Message -----
From: "Gary McGovern" <garywork at lineone.net>
To: <squeak-dev at lists.squeakfoundation.org>
Sent: Wednesday, November 07, 2001 2:43 PM
Subject: Re: Three Threads Of Squeak


>
> ----- Original Message -----
> From: "Doug Way" <dway at riskmetrics.com>
> To: <squeak-dev at lists.squeakfoundation.org>
> Sent: Wednesday, November 07, 2001 2:25 AM
> Subject: Re: Three Threads Of Squeak
>
>
> > Anyway, let's consider the three levels.  In what way does Squeak not
> support the Conceptual and Logistical levels?  Are you implying that
another
> language/environment/tool does support all three levels?
> >
> > Squeak/Smalltalk is a high-level language relative to most commonly used
> programming languages (e.g. C++, Java, Basic, Perl, etc.).  And it is
> reflective and includes an IDE, written in itself, which lets you analyze
> code at a higher level than a text editor, and also lets you build tools
to
> potentially analyze at a higher level still.  Thus it should be able to
> support the sort of higher-level thinking you're talking about
(conceptual,
> logistical) better than most others.
> >
> > Granted, it might be nice to have more tool support built-in for this
sort
> of higher-level thinking.  But existing Squeak tools like ThingLab and
Ned's
> ConnectorMorphs sort of point in this direction.  (I'm not sure exactly
how
> "high-level" we're talking.)
>
> Hello Doug,
> I'd like to mention a couple of points here. Justin mentioned Prolog to me
> offline and following Alan's citation url a few messages back, Man -
> Computer symbiosis is mentioned, and a key inference of that idea would be
> greater intelligence on the part of the computer.
>
> I'd be interested in more intelligent objects, but it's still two year
> before I do AI.
>
> Perhaps that's what Justin is getting at.
>
> Regards,
> Gary
>
>
>
>





More information about the Squeak-dev mailing list