Newsgroups: comp.ai
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel-eecis!news.mathworks.com!newsfeed.internetmci.com!in2.uu.net!fox.almaden.ibm.com!garlic.com!news.scruz.net!cruzio.com!news
From: ababian@cruzio.com (andrew babian)
Subject: Intelligent chat script?
Reply-To: ababian@cruzio.com
Organization: Cruzio Community Networking System, Santa Cruz, CA
Date: Tue, 2 Jul 1996 22:34:16 GMT
Message-ID: <DtxtD4.2oI@cruzio.com>
X-Newsreader: WinVN 0.92.6
X-Nntp-Posting-Host: cruzio71.cruzio.com
Sender: news@cruzio.com (The News User)
Lines: 29

Since the various chat systems on the internet are the actual format
for the Turing test,  its seems a likely medium to try some AI research.
Is anyone working on this seriously, and who are they?

Actual chat script processing is typically a simple eliza type text 
processing,  but it should have much greater potential than that,
certainly creating models the other people would be a great step forward.
I imagine that most of you out there doubt that there actually is 
any intelligent life out there in chat land, but there is a little.
I suppose the kind of programming i'm talking about here would be 
unrespectable hacking,  instead of serious ai,  but it might be
interesting.

The model that i'm considering for the experimental chatter would
just be an additional menu of possible statements to say.   This
would be a way to monitor the programs output without flooding,
and to monitor the quality of the program output while considering
changes.  The user could then select an appropriate response.  This 
has a possible benefit of allowing for supervised learning,  should
the programmer get that ambition.  Similarly,  the user will still
be able to chat,  and the program will be able to use the 'expert'
responses as desired.  Chatting is actually a fairly complicated
social behavior.  The decision to speak requires questions of
maintaining the mood and the topic,  and also requires an ability
to speak creatively but relevantly.  The actual problem of a good
chatting system is beyond our technology,  but i think this might
be an interesting research format.

Anyone interested, please contact me.
