🚀 go-pugleaf

In article <44...@bu-cs.BU.EDU>, be...@bu-pub.bu.edu (Blair M. Burtan) writes:
> A few years ago, there was a program floating around our IBM
> mainframe system, then a 3081, that seemed to be an unbelievable
> example of artificial intelligence.  When you ran the program, it
> would look and act something like Eliza.  But the difference was that
> it seemed to know a great deal about the person running it.
> [...]
> A few weeks later, the cat was out of the bag.  The program was actually
> a chat/talk style program that would communicate between the author's
> terminal and the user.  Or users.  Each user would open up a window
> on the author's screen.  He had to take the program down because so
> many people were "talking" to him and he couldn't keep everyone's
> information straight much less get any work done.

In the book Metamagical Themas, Douglas Hofstadter confesses being fooled
similarly: He had had lectures on the Turing test, and professed his belief
that no truly intelligent program would ever be made. One day his students
asked him to test one of their programs, to see if it would "reveal its
true nature". Mr. H. expected to do this very quickly, and was amazed at
the program's knowledge of semantics. Just when he was about to ask whether
the whole thing was a hoax, the program apparently hiccuped on his last
input line, and printed something like:

!#$&![] SYNTAX ERROR 3465 IN LINE 654

Whereupon he was instantly assured, and even encouraged his "disappointed"
students, saying that the program was actually very impressive, it just
needed some testing and debugging.

Kai Quale