[comp.ai] How far can Expert Systems go?

edwin@fornax (06/01/91)

I am posing this news message for a Computer Science student at SFU,
if there is any reply to this topic, please forward the message to
the following e-mail address:
	internet :	c411953@csil.cs.sfu.ca  OR  _craig@csil.cs.sfu.ca
	bitnet :	c411953%csil.cs.sfu.ca@relay.cdnnet.ca
		   OR   _craig%csil.cs.sfu.ca@relay.cdnnet.ca
--Thanks.

==============================Message begins here=========================

     How far can expert systems go??
  or,
     What, potentially, can they accomplish??
     Will they ever replace human experts??
     Will they ever be trusted??

I would like anyone to e-mail me (or post) any references
to articles, books, journals, etc. that might try to answer
the above questions.

I would also welcome anyone e-mailing me with
their opinions....

Another student and I have to do a discussion, in class, on
this topic and we have found very little info so far.

Thanks,
Craig.

============================Message ends here=============================

kingsley@hpwrce.HP.COM (Kingsley Morse) (06/04/91)

IMHO expert systems share an Achilles heel with programming languages. They're
both limited by the number of lines of code that can be maintained. Once they
reach about a million lines of code, my understanding is that its too unwieldy 
to maintain or debug. Thus a million lines of code is the "glass ceiling" for
now.

On a more positive note, research into machine learning is addressing this 
limit.