From: Ben Goertzel (ben@webmind.com)
Date: Thu May 03 2001 - 06:53:38 MDT
> How
> difficult would
> it be for the AI (WM in this case) to read natural language
> descriptions of
> CS concepts instead of Mizar? In other words, what is the
> likelihood that
> WM will be able to read text books, web sites, etc. in order to learn?
Well, this is an unanswered question, as regards future AI's
I can tell you that, right now, the WM AI Engine could load in some Cs
theorems from Mizar and produce some interesting results, whereas it
couldn't read an academic CS paper
Perhaps I am overgeneralizing from the strengths and weaknesses of my own AI
system
However, I think there is some general sense to the conclusion that
understanding complex manifestations of human discourse is going to be
significantly harder for a computer (and thus come later in its development)
than understanding mathematical and algorithmic content (which, though
still, human-flavored to be sure, is more "objective" and less species-bound
in nature)
ben
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT