COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	farout[e87,jmc]		The far out future of AI
C00008 ENDMK
C⊗;
farout[e87,jmc]		The far out future of AI

	I find myself on the conservative side of predictions
about AI, and I emphasize the fundamental conceptual problems
that remain before AI reaches the human level of intelligence.
This gets dull after a while, so this article discusses what
we'll do when we get there, although I want to leave you the
reminder that this may be anywhere between five years from now
and five hundred years from now.  We haven't identified some of the major
conceptual problems, and mere money can't guarantee to solve
them until that happens.

	Science fiction usually depicts robots of about human
level.  This is convenient for the writers, because it permits
the adaptation of conventional plots to a world in which robots
are some of the inhabitants.  Robots can be evil conquerors,
victims of oppression or beset with psychological problems according
to the taste of the author and the conventions of the times.

	It seems much more likely that AI of human level will
last a very short time, e.g. a few weeks.  Once human level has
been reached, the application of money to put the AI in faster
computers will cause human level to be far surpassed.

	What shall we do with it?

	The first question is what do you mean by ``we''?  Some
people think the question is ``What will society do?''.  However,
there are many societies in the world, and different societies
will have different preferences.  Also there are individuals, and
they will have different preferences.  Moreover, these preferences
will develop with experience.
  We can even hope that some societies
will leave it up to individuals to decide how they want to use
AI --- presumably within limits prescribed by law.

	Firstly, we'll get used to having robot servants.  Everyone
will have servants of better quality than the richest people today
or past royalty.  If we're smart the servants will be suffiently
unlike people so that we will be tempted neither to ``punish''
them when they malfunction to our disappointment nor feel sorry
for them when they work 24 hours a day.

	The widespread availability of robot servants will raise
standards of sanitation, general cleanliness and neatness far
beyond the level of the present or any previous culture.  I
believe that it will also lead to cultures with levels of
decoration far beyond that reached by the Victorian rich.

	If it is genuinely smarter than humans we will ask it
to discover and explain the answers to scientific problems
that puzzle us including the consequences of different kinds
of individual and social choices.  One kind of question we will use it
to answer concerns the consequences of different policies
about the use of AI.

	A natural question is what will people do when we have
robots smarter than ourselves.  What will we do when we are all
rich?  The latter question is readily partially answered.  There
have been people rich enough not to have to work throughout history
and there are idle rich today.  They have their complaints and their
psychological problems, but they seem to prefer being rich to being
poor.

	As to having robots smarter than ourselves, another partial
answer is available.  Everyone but the one smartest person in the
world has this problem already.  There are people smarter than I
am, and I find their company interesting.  I think most people don't
avoid the company of people smarter than they are.  It's the other
way around.  The smart ones find the conversation of the dumb ones
uninteresting.  However, the robots will be subservient, and we'll
have built them incapable of boredom.