Re: Thinking about the future...

Dan Clemmensen (
Thu, 05 Sep 1996 22:50:55 -0400 wrote:
> In a message dated 96-09-05 15:27:46 EDT, Dan writes:
> << . (My actual model is that some
> >computer
> >nerd at MIT will do this while drinking Jolt cola at 2 AM when he should
> >be studying
> >for an english exam.) As a joke, the nerd will type in the command "make
> >yourself smarter."
> >The then-current rule set will be smart enough to act on the command but
> >too stupid
> >to get the joke.
> > >>
> In my limited understanding of how NN works, this command is neither a joke ,
> nor a human try to make AI "unbenign"- but the current method of programming
> - error correction for, example: improved pattern recognition. In essence it
> is a neccessary part of building up sufficient intelligence to make an
> "artificial" being or intelligence. This 'learn all you can command ' I
> understood as a given, but was interested in exploring how *after* it gained
> "consciousness - or replicated it's own AI's- which motivation AI would
> use, that which *we* programmed, or it's *own* directives - as it "wakes up"
> and perhaps questions its own "meaning of life" or existence and purpose (
> assuming it could ever get that kind of consciousness).
> Nadia

My understanding of the nominal "expert system" is not good enough to
describe as "limited". There is a slight difference, however, between
"learn all you can" and "make yourself smarter." "Learn" might be
interpreted as "add to your database". "get smarter", might cause
the system to begin distributing itself to other processors, or
optiimize its own code, or interface to another inference engine. I
was thinking in these terms. After such self-augmentation, the resulting
SI might indeed develop the philosophical questions you raise. I might
be smart enough to answer them, or it might decide it needs to be
smarter in order to answer them, and embark on additional