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

667Re: [aima-talk] lisp and free book

Expand Messages
  • Robert Futrelle
    Apr 6, 2006
    • 0 Attachment
      Re: [aima-talk] lisp and free book
      I have to agree that symbolic computation has definite limits.  You could even go so far as to consider it a human conceit to try to reduce rich and complex phenomena to a set of discrete symbols that we create.  Corpus-based NLP, HMMs, situated robotics systems, neural nets, etc., demonstrate the power that's accessible beyond human-designated symbols.  The argument that we have symbols in our heads is a hard one to support, and it's only getting harder.

       -- Bob Futrelle

      _______________________________________________________________
      > Robert P. Futrelle      | Biological Knowledge Laboratory
         Associate Professor  | College of Computer  and Information
                  >              |     Science MS WVH202
      Office: (617)-373-4239  | Northeastern University
      Fax:    (617)-373-5121  | 360 Huntington Ave.
      futrelle@...    | Boston, MA 02115
               http://www.ccs.neu.edu/home/futrelle
        http://www.bionlp.org   http://www.diagrams.org

                http://biologicalknowledge.com
             mailto:biologicalknowledge@...
      _______________________________________________________________


      On Wed, Apr 05, 2006 05:20:29PM -0400, Robert Futrelle wrote:
      > If you have a chance to look at Norvig's book,  Paradigms of Artificial
      > Intelligence Programming: Case Studies in Common Lisp, it can teach you a
      > lot.  It will show you how compactly Lisp can represent and manipulate
      > knowledge-related structures and algorithms.

      Thanks. I will.

      > C is a poor language for AI because its syntax is so distant from the concepts
      > you are working with.

      If you mean GOFAI, I consider the Python sources of AIMA much easier to
      understand than the lisp sources. For embodied AI, like Robot programming, C is
      the one. The best robot simulations I know, Player/Stage Project and Webots, are
      in C/C++

      > When trying to do symbolic computations in C, most people end
      > up unwittingly re-implementing many of the basics of Lisp

      "Symbolic Computations" is exactly the problem. "Symbolic Computations" are the
      reason why AI has not yet managed to create robots as intelligent as, say, a
      fish. A good book about it is: "Understanding Intelligence" by Pfeifer and
      Scheier.

      > BTW: What free AI book is being referred to?

      http://www.markwatson.com/opencontent/
      and some books in wikibooks.


      I think symbolic algorithms are an interesting field in AI, but not the only one,
      not even the most important one, and I think that fact is the first thing any AI
      student have to heard.

      Coming back to the OP's question, I would say the best languages to learn at
      the moment for AI are python because it is easy; and C/C++ for performance.

      --
      Ivan F. Villanueva B.
      artificialidea.com
      <<<                   European Community Patent will bring            >>>
      <<<                     Software patents by the backdoor              >>>
      <<<                      
      http://wiki.ffii.org/ComPatEn                 >>>

      YAHOO! GROUPS LINKS


    • Show all 7 messages in this topic