COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	ideas[e84,jmc]	
C00012 ENDMK
C⊗;
ideas[e84,jmc]	

From some notes:
	"The is-a hierarchy requires that the primary attributes be
taken for granted".  What could I have meant by that?
Perhaps I meant that being a bird is a primary attribute, while
whether it can fly is secondary.  My axioms allow non-monotonic
reasoning about whether something can fly but not about whether
it's a bird.

	Heavy duty set theory has a recursion axiom built in.

1984 July 10

	Can a spacecraft pump itself (like a person on a swing)
from one orbit to another?  The question is rather what pumping
can be done.

1. The simplest configuration is two masses connected by a string.
Work is done by reeling in the string.  It would seem that if
the string is reeled in when the configuration is near periaspsis
and let out near apiapsis, energy would be added to the orbit.
The point is that this is should be very cheap momentum compared
to momentum gained by expelling mass, and it won't use up mass.

2. The rate at which momentum can be obtained will increase with
the length of the string and the eccentricity of the orbit.  It will
decrease as distance from the primary increases.

3. Rigid configurations that whose parts can be rotated may do some
things that balls on strings cannot, but they are much harder to
make big.

4. While it seems that the eccentricity of the orbit gives the
greatest leverage, it should be possible to work from a circular
orbit even without the leverage given by higher moments of the
mass distribution of the primary.

July 17 - It is now clear that the scheme won't change the
angular momentum about the primary.  It seems still possible
that the energy can change, but the effect may be too small
to be useful.

A ladder of bolos may work, but it seems likely that this has
already been treated.

Another scheme uses the idea of the dynamic tower in a new way.
Spacecraft propel themselves by throwing balls back and forth.
Overall changes in angular momentum are achieved by throwing
balls around the moon.

july 17 -
Bob Moore's "Theory of Knowledge and Action" looks quite good.
I think I'll try to push the common sense people into doing
things within its framework.  I still have doubts about
combining situations and possible worlds.  The formalism needs
some adequacy and soundness theorems, but what they should be
isn't clear.

It may be worthwhile to try to push Hobbs and Moore into some consolidation
of their approaches.  Perhaps via Nilsson or Stan R.

Some subjects are locally S5 in their knowledge, i.e. when someone
doesn't know something, he knows he doesn't.  Perhaps this should
be a non-monotonic conclusion.  Erroneous beliefs while common
are non-standard.

july 17 - We have value(x,context) and true(p,context).  However,
we often want to pick out subcontexts.  This may give us something
like
	value(x,context) = val1(x,value(subcontext1,context)).

july 28 - Diet and anger.  More people seem to be getting desperate,
going berserk, joining movements of the left, right or religion
or peace that provide an outlet for fanaticism.  It's hard to see
how this could be a result of objective conditions, because the
grievances are often second hand.  Consider the hypothesis that
the problem is in blood or brain chemistry and is affected by
diet.  This is probably testable.  A person of normal behavior
often has episodes of anger on which he can report and can often
notice that events that sometimes trigger anger sometimes don't.
Perhaps this can be correlated with blood chemistry.  Of course,
it is known to correlate with the use of alcohol in many people.
However, there must be other factors.

aug 11 - Consider actions.  When the action occurred in the past,
it has infinite detail, e.g. the trajectory of my arm as I moved
the block.  When considered hypothetically, the action has no
more detail than required.  When I say I will put the block on
the table, it doesn't make sense for you to ask what precise
trajectory I had in mind - I didn't have a precise trajectory
in mind, although I may have had some details in mind.  Clearly
an action formalism should take these differences into account.

aug 22 - commitment in programming and common sense.  Consider
a program statement committing the program to do something or
for something to be true in the future.  The program is correct
only if it fulfills its commitments.  It is possible also to
regard a commitment as an imbedded specification.  In this case
it is definitely non-executable, but again correctness requires
that the program fulfill its commitments.

In common sense, it seems simplest to regard commitment as an
internal act not necessarily accompanied by any external action.
Correctness requires that a person or robot fulfill commitments.
Presumably it also requires that the person or robot assert only
what it believes.  A promise is an assertion of commitment made
to a specific person.  Presumably the commitment made contains
a proviso that permits the promissee to release the promiser
from the commitment.  This seems to accomplish all that is
done by Kathleen Kells's axioms about commitment.

aug 22 - Another possibility for specification of interactive
programs is that the program gives the user the ability to
accomplish certain things.  We can omit stating what the user
has to do to accomplish the goals, and we will normally also
need to specify that the program gives him the information
that enables him to achieve the goals.

Sept 1

	Here's what's probably a bad idea.  Overload cons by
letting it take an arbitrary number of arguments.  The semantics
is given by (cons x y (3 4)) => (x y 3 4), which is the
same result as that given by (list x y 3 4).  The difference is
that  list  would do all the conses, using the extended cons, the
constant tail of the list is used.