COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	~In order to substantiate the idea that there is a science of  common
C00005 ENDMK
C⊗;
~In order to substantiate the idea that there is a science of  common
sense  to  be  discovered,  I  wish  to  recount some of AI's current
difficulties in the area.  We are  interested  in  computer  programs
that  will  solve  problems  that  involve  action in the world - for
example handling physical objects.   In  order  to  get  a  start  we
simplify problems.  Here are some of the simplifications.

	&.  In  the  real world many people (4 billion) are acting at
once and even more natural events are occurring.  We have  difficulty
with  even  two  people  acting  at  once, so we simplify to a single
actor. Everything that happens is what he does.

	&. In the real world our actor acts continuously, and this we
don't yet know how to express the facts about this so that a computer
can use them for  predicting  the  consequences  of  actions,  so  we
imagine  it  to act discretely, i.e. each action produces a change in
the situation, and then the next action changes that situation.

	&. In the real world some of what the actor does is  intended
to  get  information.   We  simplify this and deal only with the laws
that give the physical consequences of his actions, so that its plans
must be based on the information it has.

	&. In the real world the actor must choose plans that work in
any state of the world in which certain facts are true.  E.g. he must
plan  in a situation in which Joe has a pet that is either a cat or a
dog but it is unknown which.  We simplify to situations in which  the
actor knows all the relevant facts.

	Now  can  you  believe  that  there are still difficulties in
finding a scheme for representing what  the  actor  knows  about  the
world.  One set of difficulties - called the frame problem - concerns
how to specify what changes and what remains the same when an  action
is taken.