perm filename REMARK[E79,JMC]1 blob
sn#475106 filedate 1979-09-15 generic text, type C, neo UTF8
COMMENT ā VALID 00002 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 Some remarks not suggested by work on circumscriptin but not about it
C00006 ENDMK
Cā;
Some remarks not suggested by work on circumscriptin but not about it
There is another formulation of the missionaries and cannibals
in which perhaps circumscription plays a lesser role or at least a
different role. Suppose we have an AI program that whenever it
gets a goal (either a main goal or a subgoal) first decides whether
to just do it or to plan first.
In the missionaries and cannibals, we may take the problem
at to go to the village of M. It is decided (how?) to plan, and
an obstacle appears, namely the river. The boat now comes in and
it is decided to go across using the boat. In the planning space,
the action of going by boat has an absolute character. If you have
a boat, you may plan to go by boat. It's only when you try to realize
the plan, or detail it, that obstacles arise. The first obstacle is
that the boat only holds two, and its solution is to relay.
Perhaps that the boat holds two is an obstacle to planning not to action.
That
encounters the obstacle that the cannibals may eat the missionaries,
and the solution to it is the well known
(331,220,321,300,311,110,221,020,031,010,021,000).
That the missionaries being eaten is incompatible with the goal
must be deducible.
It is only after this solution is found that considerations
of leaks appear. All we need do is to justify stopping at this
level of plan. That would appear to be the lack of further information.
This lack is often normal in real life when one is not on the scene.
The plan is to fill in the details when the boat can be seen.
It is also not obvious when one decides to give up planning
and go to the boat.
Of course, circumscription can help us get rid of the bridge
and fording the river. How does it get rid of fording?
MAYBE THESE REMARKS SHOULD BE IN THE CIRCUMSCRIPTION PAPER
We have the (hopefuly finite) regress:
Do X. How?
By doing Y. How?
...
By doing Z. How?
Here I'll show you.