<bbrown@transcient.com>On Friday, July 14, 2000 Wrote:
> Acheiving animal-level intelligence seems to be a much more tractible problem,
> and it would enable us to build robots good enough to automate almost all
> physical labor.
I'm not at all certain of that. 40 years ago everybody thought we'd have a robot janitor
long before we had a robot chess grand master who can beat any human alive.
They were wrong. That tells me that our naive notions about what is easy and what is
difficult are wrong. Today's computers don't have much trouble with higher math but with
more mundane things like pattern recognition, manual dexterity and that grab bag of skills
and knowledge we call common sense.
One surprising thing computer science has taught us is that an enormous about of
brain power is needed to perform an apparently simple task like moping a floor.
How do I pick up the mop? Is that dirt on the floor or is it a shadow? Should I move that
obstacle or go around it? How long should I try to get rid of that stain before I give up?
In retrospect we shouldn't have been so surprised, mundane stuff like that is essential
for survival and evolution wouldn't give us more intelligence than we needed to survive
in the jungle.
These problems don't seem hard to us because we're very good at them. Natural
selection gave a big advantage to those who were good at spotting the pattern
of a saber toothed tiger hiding in the bushes, but no advantage to those who
were good at solving partial differential equations. It is not mere rhetoric to say
that in an absolute sense a janitor has a more intellectually challenging job than
a professor of mathematics.
John K Clark jonkc@att.net
This archive was generated by hypermail 2b29 : Mon Oct 02 2000 - 17:34:37 MDT