XPost: alt.folklore.computers   
   From: antispam@fricas.org   
      
   In alt.folklore.computers rbowman  wrote:   
   > On Thu, 1 Jan 2026 10:30:54 -0000 (UTC), Waldek Hebisch wrote:   
   >   
   >> But are 'expert systems' really AI?  Theoretically so called expert   
   >> system shells could do smart things, but examples I saw were essentially   
   >> a bunch of "if ... then ..." which could be written in almost any   
   >> programming language.  One example of samewhat succesful 'expert system'   
   >> is supposed to guide a user trough installing Unix.  Description   
   >> suggests that is is not smarter than modern Debian installer.  And   
   >> nobody thinks that Debian installer is AI.   
   >   
   > I never thought so. Like you I've looked at Lisp and Prolog and came away   
   > with the thought 'you *could* use that approach but why would you? It adds   
   > nothing to C but obfuscation.'   
      
   You mean 'expert system' coded in Lisp or Prolog?  Or just general   
   coding in Lisp or Prolog?   Concerning general coding IMO Prolog   
   is great for backtracking search and a few similar problem, but   
   not good for most of programs.  On the other hand Lisp is quite   
   capable general purpose language.   
      
   > I don't think they call it an expert system but Arch Linux has a very   
   > detailed description of installing the system. There is also a sketchily   
   > maintained script that automates much of the process although the 'I use   
   > Arch btw' crowd considers that cheating.  Then there is EndeavourOS and a   
   > couple of others that act like Debian, Ubuntu, or other installers and   
   > install Arch, throwing in several useful tools.   
   >   
   > Then there was 'fuzzy logic' that had its day although you don't hear much   
   > about it lately. Perhaps it was overtaken by neural networks.   
      
   I looked a bit at 'fuzzy logic'.  But I did not see more in it than   
   principle "if you do not know better, then use crude approximation".   
   This principle is resonable, but I did not see any reason to prefer   
   specific crude approximations advocated in various texts (with   
   approximation varying depending on the text).   
      
   > During   
   > training of a NN in successive iterations you calculate the loss function   
   > until you reach a point where it's 'good enough'.  That technology is   
   > interesting that while you can define and explain each mathematical   
   > operation what's going on in the total sum is cloudy.   
      
      
   --   
                                 Waldek Hebisch   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)