Re: Article - Why the Loebner Prize Will Never Be Won...

"Ray Dillinger" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]
> Tom Breton wrote:
> >
> > "Gorbag" <[EMAIL PROTECTED]> writes:
> >
> > > "Brownell" <[EMAIL PROTECTED]> wrote in message
> > > news:[EMAIL PROTECTED]
> > > > http://www.powerofq.com/Q/loebnerarticle.htm
> > >
> > > Interesting, but presupposes the Loebner challengers will all be Eliza
type
> > > chatterbots.
> >
> > Well, I'm not sure how you get that.  I didn't think the article
> > assumed that.  It assumed that even a perfect chatterbot has a limited
> > amount of world knowledge ("limited" as in "huge, but no match for a
> > 6-yo kid").  It argued that a judge could easily expose any practical
> > program's limited world knowledge.

What I meant was the chatterbot approach (matching patterns) is flawed.  Cyc
(if it works) would be an argument for a
different approach. There are others.

Also note that "never" is a long time. But if the author supposes that
Loebner will die before the contest is won, that is certainly a morbid
possibility.

> Well, I scoped out the rules for the Loebner prize last year, and
> they're not allowing anything that takes a while to respond or
> requires more than an average machine to run on.  These restrictions
> militate against any "serious" competition to eliza style chatterbots
> in the context of that contest.

Which just means you have to wait a bit longer for serious machines to be
available as a notebook.
I just got my dual processor G5... maybe not "average" but still a home
machine... and a nice workstation for Lisp, as it happens. Far more serious
a contender than the commercial workstations of even 2 years ago, let alone
when lisp machines roamed the wild.