COMMENT ⊗   VALID 00006 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	     Many thanks for your time and the various memos you have sent me
C00011 00003	HOW FAR ARE WE FROM PRACTICAL PROGRAM PROVING
C00015 00004	To: Ralph Gomory, Joel Birnbaum, Forest Baskett
C00026 00005	notes
C00030 00006	Notes on Kahn's three books
C00039 ENDMK
C⊗;
     Many thanks for your time and the various memos you have sent me

on energy problems.
In general I agree with the points you make, but I think that some

important considerations have been missed.
I hope you will find this reaction interesting
and useful, but partly I am writing it with the intention of adapting

it for publication elsewhere.  First I should explain the nature of

my interest in energy problems.
 
     If all were well I would have only the interest in energy posessed

by an
average reader of Scientific American.  I have a semi-professional

interest in new technology, but I am much more interested in new
services than in new ways of providing customary products and
services.  In fact, I have a half-written book on what should be
invented.
 
%My specific interest in energy is almost entirely based on the
fear that it won't be available to realize my other schemes or
even to continue our present way of life which I find valuable.
I was much happier with an energy system in which energy companies

were eager to meet whatever demand would develop than one in which

a public policy is developed that plans how much energy the country

will need over an extended period.  Having to convince social planners

that a new facility that will initially be expensive is socially
desirable and should be allocated energy is much worse than having to

convince a manufacturer that people will want a new product.
 
     My fear is that an ideological transient will bring about
wrong decisions over a prolonged period, causing missed opportunities,

suffering, and maybe even disaster.  In my opinion your papers
and reports do not take ideology into account.  Let me try to put

the point in a way that doesn't presuppose that my views are
rationally determined while others are motivated by irrational
ideology.
 
     There are two extreme models of the political response to a
situation.  One model is that the facts of the matter become known,

people react independently to these facts in accordance with their

interests and values, and the action taken is a compromise among
the preferences of the population.  On many issues this is substantially

what happens.  While there are exceptions, I would say that local

governments determine spending on libraries and fixing potholes
approximately according to this model.
 
     The second model is that policy is substantially affected
by ideological and purely political events.  Ideologically, a movement

develops, attracts recruits, propagandizes and strives for power.

As it develops it learns what slogans energize its members,
attract recruits and optimize its relations with other groups.
There are two problems.  First the evolution of slogans for
their mobilizing power may have no relation to their truth -
for those slogans that are assertions about the social or physical

world.  Second there seems to be a Gresham's law of politics.
A threat of disaster has more mobilizing power than a promise of
benefits, and such a threat is enhanced in its power if it can
be ascribed to an enemy.  I remember an engineer attending a lecture

assert that if only society would state its values, engineers would

be happy to work in accordance with them.  Unfortunately, it seemed

to me that the value being advanced was that engineers are an enemy.

 
     The points you make about the limitations of government second-gues
sing
may be supplemented by considering the amount of effort that
goes into decision making.  Suppose that it is necessary to
prepare estimates of electricity demand in a region and compare
the technologies and schedules for meeting the demand.  Suppose
that the utility employs 50 people over three years in preparing
an estimate.  Propose further that the 50 man estimate is incomplete

from the external point of view, because it takes for granted
certain ideas common and accepted in the industry and in the
particular company.  Now suppose that the results of this study
are to be second-guessed by and environmentally oriented "public
interest" organization and by a government agency.  Unless this
reaction is to contain substantial elements of blind approval
or blind opposition, each of these outside bodies will have to
put in an equal amount of work.  Moreover, the work will have to
be consecutive rather than parallel to the utility study.  Besides

that each of the studies will have to be more explicit, because
it must convince an outside authority not necessarily sharing the

culture of the study groups.  Consequently, the costs will be high,

the finally implemented plan will be more obsolete, and the quality

will be lower, because the people capable of doing this kind of
study well are divided among three groups rather than one.  One
can imagine situations in which such costs and delays are warranted

by very high costs of a mistake, but I don't think that utility
planning is usually one of them.
HOW FAR ARE WE FROM PRACTICAL PROGRAM PROVING
 
	The applied goal program proving is to replace debugging programs

by computer checked proofs that programs meet their specifications.

When suitable tools are available, programmers will write their
programs in languages whose designers had program proving in mind.

As the programmer developed a program, he would develop proofs
that the parts of the program met their specifications and that
meeting hei
their specifications would ensure that the program as a whole
met its specifications.  The development process might involve
writing the program or parts of it in non-computable form that
are most amenable to proof of specification and transforming it
to runnable form by transformation processes that preserve
extensional equivalence and give guaranteed speed and storage
performance.
 
	At present we are far short of full program proving capability.

We have reasonable techniques for proving certain kinds of facts
about limited classes of programs.  However, these techniques are
not integrated into systems that make the combined task of writing
the specifications, the program and the proof of correctness as
easy as possible.  The purpose of this paper is to measure how
far we are from this goal, and to discuss whether we are within
development distance of the easier goal of making practical
systems that will be genuinely useful for some subclasses of programs
that people want to write.  Our conclusion is that we are within
development distance of useful systems, i.e. that applied projects
are worthwhile, and that the most immediate task is in the area of
improving the tools for writing the specifications, i.e. studying
how to write partial specifications that will convince potential
customers for a program that the validation of these specifications
will substantially increase their confidence in the program and
willingness to pay for it.
To: Ralph Gomory, Joel Birnbaum, Forest Baskett
From: John McCarthy
Subject: Possibility of including 801 computer in
		IBM-Stanford joint study
.fo
.sk 2
.ju
	This memo represents my personal views which have been partially

discussed with you and with John Cocke.
 
	I have been much impressed by the 801 computer project, and I think

it would make an ideal basis for improving the fruitfulness of the

IBM-Stanford joint study for the following reasons:
.sk
.in 5;.un 5
1.	The machine would be ideal for developing a new cost-effective

time-sharing system that would have a long life.  In my view, time-shari
ng
technology has reached a state where it is time for new projects
to develop systems.  I have the following features in mind:
.sk
	The system should be based on large main memory and could
emphasize swapping and paging less than present systems.
.sk
	Its cost-effectiveness requires that it be planned not to
require machine operators on any shift.
.sk
	The 32-bit address and the capability for full duplex
interaction with terminals gives a chance for man-machine
interaction much better than has been possible with the 370.
.sk
	The interrupt system code needs to be optimized for terminal
interaction.
.sk
	Individuals and groups at Stanford, M.I.T., and IBM have
independently developed the idea that the next generation
time-sharing system should be based on an efficient screen editor.

From the user point of view, it is desirable to be able to edit
output when it appears on the screen to remove redundancy
and often to put some of it back in as input.  There has been
an increasing tendency to make system commands available
directly from the editor.  The M.I.T. Lisp machine has realized
some of these advantages, but no-one has yet put them into
a time-sharing system.  From an efficiency point of view, most
present editors in time-sharing systems are a disgrace.  Editing
should take up only a tiny amount of computer time, but both
at IBM and on the time-sharing machines at Stanford, editing
is an activity of average compute-intensity.
.sk
.in 5;.un 5
2.	There is a potential basis for IBM-Stanford collaboration
in research on home computer services.  The Computer Science
Department is proposing a project using the 801, and McCarthy
at Stanford has been working at a low intensity for many years
in this direction (having developed a news service and other
facilities interesting in this direction), and the Stanford
Computer Science Department would like to intensify this
effort for which the 801 would make an ideal vehicle.
.sk
	Preliminary discussions with Forest Baskett, who is the
Stanford principal investigator of the joint study, indicate
that his planned research would benefit greatly from an 801
even if it were substituted for the planned 4331 and if there
were a considerable delay in acquiring it.
.sk
	The Stanford Computer Science Department is quite self-sufficient

in hardware and software maintenance and would not need a commitment

of continuing IBM support for the machine in order to keep it
operating and useful.
.sk 4
.in 0;.un 0
John McCarthy
.sk
/sh
notes
 
Todorovich has promised Irving Kristol that I would submit an outline
and summary of an article for Public Interest.  The theme would be
energy and ideology and its main point would be the complexity of
the forces that are currently affecting attitudes of individuals
and the policies of the U.S. and other countries.  We have
 
technological facts
scientific facts

holding energy policies hostages for other issues such as redistribution
of wealth

values - collective vs. individual, what is good life
primitive value systems, i.e. who are good guys and bad may precede policy
beliefs

ideology as a horse to be ridden to fame and power

     the contrast between lawyers and scientists.  If a lawyer wins his
case, he is permanently ahead, even if it is soon decided that his
client was actually in the wrong.  A scientist who convinces others
to a mistaken position can lose all he won and more when the truth
finally comes out.

sociological facts - Do we regard our descendants as children?

moral facts - What are legitimate risks?
      To what extent should we take seriously the moral pretensions
of the anti-technology people.  They apply double standards of risk
and of truthfulness.

the changing salience of issues
how people hill-climb in a space with so many dimensions that they
can't keep track of all the issues that have been raised.
"Yes, but ...".  Often arguers talk past one another.
 
reaction against previous generation - especially against a heroic generation
however, reason has force with some
experience teaches some

ideology always has excuses for failure, so they won't always change
voluntarily no matter how unsuccessful they are.  In particular, hardly
anyone gives up power voluntarily.
 
Foreign countries, even those that are currently passing the U.S. economically,
are still amazingly dependent on the U.S. for ideas about the future
of the world.  Also for technological initiative.
While Britain is economically much weaker than Japan and slower than
Japan to commercialize ideas, it generates many more ideas than
Japan does.
 
The current ideological paralysis of the U.S. requires us to hope
that other countries will begin to take technological initiatives, especially
in the nuclear field on which others depend much more than does the U.S.
Notes on Kahn's three books
 
World Economic Development - 1979 and beyond
 
starting p. 388
 
Kahn's theories about the Kondratieff cycle and its variants seem
dubious, because the time scale is too long.  Since the total capital of
the U.S. is less than four years GNP, and the differences in capital
between shortage and plenty are a fraction of that, a fifty year
cycle seems implausible.