Loading ...
Sorry, an error occurred while loading the content.

596[aima-talk] Re: The future of AI

Expand Messages
  • Robert Futrelle
    Dec 16, 2005
    • 0 Attachment
      [aima-talk] Re: The future of AI
      I guess what bothers me the most about Murray's work described below is that when a researcher launches out in an unusual new direction, with lots of specialized terminology and concepts, there are certain obligations to the community that can get ignored.  What can happen is that the researcher fails to explain the links between what he or she is doing and the large body of work already done and being done.  That allows the researcher to avoid any criticisms by comparisons.  Much of the work that has been and is being done in AI and Cog Sci is excellent and well-documented and carefully related to extant work.

      All (conventional) scholars are careful to describe how they are building on the work of others ( "If I have seen further it is by standing on ye shoulders of Giants." - Newton).  In fact, any paper for a journal or conference is reviewed with an eye to its discussions of the literature - do the authors fully understand what has gone before and how what they are presenting builds on that?

      When a researcher avoids the traditional, they can also isolate themselves from criticism.  I would be delighted to see a scholarly discussion of the Mind4th system that relates it to other work and relates its terminology and concepts to already existing ones.  If the Mind4th system is better, let's see a detailed comparison of its strengths versus existing work. If the terminology and concepts and relations among them are better, let's see a defense that describes current work and explains why the new work is better, why neologisms are really needed.  As a starter, how does it relate to the view of AI presented in AIMA and PAIP?

      All students and researchers should pay attention to the description I've just given of the way research and scholarship works.  AI invites people to wax rhapsodic about all sorts of vague and marvelous ideas.  But unless the ideas can be brought down to earth and tied into what has come before, they may not be that useful.  Einstein and Schrodinger and other scientists with radical new ideas understood this and related what they did to the existing knowledge of Galilean relativity and Newtonian mechanics.

      Here's a quote from one of the ACM URLs that was included in Murray's post. I include it for what it's worth, neither supporting or refuting it:

      "Six years ago in this journal we discussed the work of Arthur T.
      Murray, who endeavored to explore artificial intelligence using
      the Forth programming language [1]. His creation, which he called
      MIND.FORTH, was interesting in its ability to understand English
      sentences in the form: subject-verb-object. It also had the
      capacity to learn new things and to form mental associations
      between recent experiences and older memories. In the intervening
      years, Mr. Murray has continued to develop his MIND.FORTH: he has
      translated it into Visual BASIC, PERL and Javascript, he has
      written a book [2] on the subject, and he maintains a wiki web
      site where anyone may suggest changes or extensions to his design
      [3]. MIND.FORTH is necessarily complex and opaque by virtue of
      its functionality; therefore it may be challenging for a newcomer
      to grasp. However, the more dedicated student will find much of
      value in this code. Murray himself has become quite a
      controversial figure."

      Disclaimer:  I have not taken the time to study Murray's work in detail, but what little I've seen of it immediately triggered the concerns I expressed above.  After 30+ years of working in AI areas, I've seen lots of marvelous work and lots of off-the-wall things too. I can usually tell the difference.

          -- Bob Futrelle

      http://mind.sourceforge.net/mind4th.html mow invites Netizens
      to install a JavaScript Seed AI as "Mind" on their computers.

      http://mind.sourceforge.net/Mind.html now includes a recursive
      link back to itself so that users may update their local copy.

      http://mind.sourceforge.net/userman.html ordains AI diaspora.
      > First you start the JavaScript AI Mind by clicking on it.
      > You may also host the AI Mind on your own hard disk
      > or your own Web page and then bring the Mind to life by
      > running MSIE, then using "File/ Open" with the AI filename,
      > or by clicking on the Seed AI link that you have transplanted.
      > Notice at the end of this User Manual that there is a link to
      > C:\Windows\Desktop\Mind.html that will work only when you
      > have made your own local copy of the JavaScript Seed AI. You
      > have the opportunity to spread your artificial intelligence
      > by planting the Seed AI on as many computers as possible.
      > If the computer that you use is shared by a group of users,
      > other people may stumble upon the ghost in the machine and
      > may accept the challenge of programming a more powerful AI.
      > If you are a manufacturer or distributor of computers, you
      > are free to install the JavaScript Seed AI on every desktop.

      A. T. Murray
      --
      http://doi.acm.org/10.1145/307824.307853
      http://doi.acm.org/10.1145/1052883.1052885
      http://www.scn.org/~mentifex/agiradar.html
      http://www.sl4.org/archive/0205/3836.html



      YAHOO! GROUPS LINKS


    • Show all 11 messages in this topic