X-Newsreader: Geminisoft Pimmy 3.2 Eng - www.geminisoft.com
From: "John Colagioia" <JColagioia@csi.com>
Newsgroups: rec.arts.int-fiction
Subject: Re: Non-Imperative Parsing (was Re: Another worm for the NLP can...)
Date: Thu, 18 Jul 2002 07:27:12 -0400
References: <Xns924645FEF5002joaomendesnetcabopt@194.65.14.150> <3d30301f$1@excalibur.gbmtech.net> <dy1Z8.20238$eH2.7849894@ruti.visi.com> <3d356ca2$1@excalibur.gbmtech.net> <JYfZ8.20297$eH2.8081219@ruti.visi.com>
MIME-Version: 1.0
Content-Type: text/plain
Content-Transfer-Encoding: 8bit
NNTP-Posting-Host: ool-182f30fa.dyn.optonline.net
X-Original-NNTP-Posting-Host: ool-182f30fa.dyn.optonline.net
Message-ID: <3d36a624$1@excalibur.gbmtech.net>
X-Trace: excalibur.gbmtech.net 1026991652 ool-182f30fa.dyn.optonline.net (18 Jul 2002 07:27:32 -0400)
Organization: ProNet USA Inc.
Lines: 64
X-Authenticated-User: jnc
Path: news.duke.edu!newsgate.duke.edu!news-hog.berkeley.edu!ucberkeley!news.maxwell.syr.edu!nntp.abs.net!uunet!dca.uu.net!excalibur.gbmtech.net
Xref: news.duke.edu rec.arts.int-fiction:106303

thornley@visi.com (David Thornley) wrote:
>In article <3d356ca2$1@excalibur.gbmtech.net>,
>John Colagioia <JColagioia@csi.com> wrote:
[...]
>>Same reason a lot of small-corporate code is in BASIC
>>and Pascal; the implementation was either free or
>>cheap when they started work, and the reference
>>material was easily available.
>For some time, there have been good Common Lisp implementations
>freely for the more popular versions of Unix, and reference
>material has generally been easily available.  It may be that
>the C and Perl was already installed, but a decent AI program
>is usually sufficiently much work that it's worthwhile to put
>a serious effort into tools.

I know that, and you know that, but you're overlooking the
driving forces of (particularly pre-everyone-on-the-
Internet) industry.

First, finding LISP is hard if you don't know LISP.
Related, wander down to your local book shop, and see if
there's still a copy of Steele's book; I haven't seen one
in many years.  When the AI book has examples in C, guess
what the implementation language is going to be...

Second, and most importantly, no matter how bright the
person, people (especially research-y industry types)
absolutely hate scrapping code.  Again, I'm guessing
that you're like me:  time is budgetted into projects
for the explicit purpose of rewriting just about the
entire thing.  Other people hate doing that, for some
reason, though.

Believe me, I've tried helping.

[...]
>>One thing I've learned in my time working with
>>computers:  Implementation languages are rarely
>>chosen with the application domain in mind.
>Yes, but if this is real AI it seems like an unusually large
>mismatch.  (Then again, I had a classmate in a pattern recognition
>class that turned her homework in in Fortran.)

I once had a student hand in a distributed database in
APL; *That* was fun to grade, and very, very
enlightening...

>For real, serious AI it will be less work to learn and implement
>Common Lisp than to do the work in C and Perl.

I'm going out on a limb, here, but I'm guessing you've
never had to make this argument at work.  I ended up
leaving my last job because my coworkers wouldn't
accept just this concept.  Code was written by
consultants, and we *certainly* weren't going to go
rewriting those couple-hundred lines of buggy crap...

...and besides "everyone" uses Java for low-level
database access and high-speed network
communications...

Oh, well...such is life, I suppose...

[...]
