RE: Real AI

From: Ben Goertzel (ben@webmind.com)
Date: Wed Jun 20 2001 - 20:59:36 MDT


Hi Michael,

Jason can't get his chat bots to be very smart based on the technology
currently underlying them. But *he* may be smart enough to, given a few
years effort, build a real cognitive architecture underlying future versions
of his chat bots. Who knows. there's no real evidence that he's doing this
now, however.

more on Monday when I get back from the business trip I'm about to embark
upon,
ben

  -----Original Message-----
  From: owner-sl4@sysopmind.com [mailto:owner-sl4@sysopmind.com]On Behalf Of
Michael Korns
  Sent: Wednesday, June 20, 2001 4:38 PM
  To: Ben Goertzel; sl4@sysopmind.com; Peter Voss
  Cc: Gilda Cabral
  Subject: Re: Real AI

  Ben, Peter,

  I visited the site. The "Botson" device was certainly just a surface level
chat bot.

  How intelligent (i.e.responsive) could Jason get his chat bots? For
instance, how responsive, in a domain (say investing), could his chat bots
get without really being intelligent?

  Most chat bot technology is very brittle even when restricted to a domain
of expertise.

  ******************************
  Michael F. Korns
  ------
  9 Pembroke Lane
  Laguna Niguel, Ca. 92677
  (949) 443-4847
  ------
  201 Harrison, Suite #809
  San Francisco, Ca. 94105
  (415) 764-4847
  ------
  www.Korns.com
  www.InvestByAgent.com
  ******************************

    ----- Original Message -----
    From: Ben Goertzel
    To: sl4@sysopmind.com ; Michael Korns
    Sent: Wednesday, June 20, 2001 11:38 AM
    Subject: RE: Real AI

    I know that group well; it's a small team, housed in a mansion in
Israel,
    headed by my friend (sort of) Jason Hutchens. Jason was a grad student
in
    perth when I was a research fellow there. he's a world-class chat bot
    builder, but in my view is not a very serious thinker about cognition,
    pattern recognition, etc. his approach is very behavioral -- he wants
to
    make a series of chat bots that will sequentially emulate older and
older
    human children, until he gets one that passes the Turing test. He
relies
    heavily on statistical, Markov-chain-type learning methods. I like his
    stuff but unless he's undertaken a radical revision of the conceptual
    foundatiosn of his work, I think it's very unlikely he'll build a system
    with much real intelligence.

    Ben

> -----Original Message-----
> From: owner-sl4@sysopmind.com [mailto:owner-sl4@sysopmind.com]On
Behalf
> Of Peter Voss
> Sent: Wednesday, June 20, 2001 10:55 AM
> To: sl4@sysopmind.com; Michael Korns
> Subject: Real AI
>
>
> Somebody else is trying to do Real AI: http://www.a-i.com/
>
> It's worth looking at.
>
> www.optimal.org - Any and all feedback welcome: peter@optimal.org
>
>



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT