Computation: Finite and Infinite Machines

334 Pages·1967·7.093 MB·English

by  Marvin Lee Minsky

Prentice-Hall
Series in Automatic Computation
George Forsythe, editor
BATES AND DOUGLAS, Programming Language/ One
BALJMANN, FELICIANO, BAUER, AND SAMELSON, Introduction to ALGOL
BOWLES (editor), Computers t`n Humanistic Research
CESCHINO AND KUNTZMAN, Numerical Solution oflnitial Value Problems
DESMONDE, Computers and Their Uses
DESMONDE, Real- Time Data Processing S ystems: Introductory Concepts
EVANS, WALLACE, AND SUTHERLAND, Simulation Us1`ng Digital Computers
FORSYTHE AND MOLER, Computer Solution of Linear Algebraic Systems
GOLDEN, Fortran IV: Programming and Computing
GOLDEN AND LEICHUS, IBM 360; Programming and Computing
HARTMANIS AND STEARNS, Algebraic Structure Theory ofSequential Machines
HULL, Introduction to Computing
MARTIN, Design of Real-time Computer Systems
MARTIN, Programming Real- Time Computer Systems
MINSKY, Computation: Finite and Infinite Machines
MOORE, Interval Analysis
SCHULTZ, Digital Processing: A System Orientation
SNYDER, Chebyshev Methods in Numerical Approximation
STROUD AND SECREST, Gaussian Quadrature Formulas
TRAUB, Iterative Methods for the Solution of Equations
VARGA, Matrix Iterative Analysis
WILKINSON, Rounding Errors in Algebraic Processes
ZIEGLE R, Shared- Tinte Data Processing
PRENTICE HALL INTERNATIONAL, INC., London
PRENTICE HALL OF AUSTRALIA, PTY. LTD.,Sydney
PRENTICE HALL or CANADA, LTD., Toronto
PRENTICE HALL OF INDIA PRIVATE LTD., New Delhi
PRENTKI HALL ot JAPAN, INC., Tokyo
/COMPUTATION:
FINITE AND
INFINITE MACH|NE§
MARVIN L. |y\|Ns|<Y
Professor of Electrical Engineering
Massachusetts Institute of Technology
UN|VERS|TY
OF \¢‘v'.A.
ILRARY
PRENTICE-HALL, INC.
ENGLEWOOD CLIFFS. N. J.
Q 975d2
© 1967 by
Prentice-Hall, lnc.
Englewood Cliffs, N. J.
All rights reserved.
No part of this book may
be reproduced in
any form or by any means
without permission in writing
from the publisher.
Current printing (last digit):
I0 9 8 7 6 5 4 3 2
Library ol`Congress Catalog Card No. 67-|2342
Printed in the United States of`America
In memory of HENRY MINSKY

PREFACE
INTRODUCTION
Man has within a single generation found himself sharing the world
with a strange new species: the computers and computer-like machines.
Neither history, nor philosophy, nor common sense will tell us how these
machines will affect us, for they do not do “work” as did machines of the
Industrial Revolution. Instead ofdealing with materials or energy, we are
told that they handle “control” and “information” and even “intellectual
processes.” There are very few individuals today who doubt that the
computer and its relatives are developing rapidly in capability and com­
plexity, and that these machines are destined to play important (though
not as yet fully understood) roles in society’s future. Though only some
of us deal directly with computers, all of us are falling under the shadow
of their ever-growing sphere of infiuence, and thus we all need to under­
stand their capabilities and their limitations.
It would indeed be reassuring to have a book that categorically and
systematically described what all these machines can do and what they
cannot do, giving sound theoretical or practical grounds for each judg­
ment. However, although some books have purported to do this, it
cannot be done for the following reasons: a) Computer-like devices are
utterly unlike anything which science has ever considered-we still lack
the tools necessary to fully analyze, synthesize, or even think about them;
and b) The methods discovered so far are effective in certain areas, but
are developing much too rapidly to allow a useful interpretation and
interpolation of results. The abstract theory-as described in this book­
tells us in no uncertain terms that the machines’ potential range is enor­
mous, and that its theoretical limitations are of the subtlest and most
elusive sort. There is no reason to suppose machines have any limitations
not shared by man.
We nevertheless expect our findings to crystallize in the years ahead­
and we are confident that the ideas and the formal methods that fill this
book will remain in the mainstream of that theory which is yet to come.
vii
viii PREFACE
We know this is so for several reasons. a) The central concept of effective
computability appears necessary, even inevitable, once one has grasped it.
In a convergence of ideas rare in science and mathematics, all sorts of
independent attempts to form a “Theory of Machines” have been swept
in this direction; and b) The sheer simplicity of the theory’s foundation
and the extraordinarily short path from this foundation to its logical and
surprising conclusions give the theory a mathematical beauty that alone
guarantees it a permanent place in computer theory.
THE MAIN GOAL
The main goal of this book is to introduce the student to the concept
of ejeczive procedure-a technical idea that has crystallized only fairly
recently but already promises to be as important to the practical culture
of modern scientific life as were the ideas of geometry, calculus, or atoms.
It is a vital intellectual tool for working with or trying to build models of
intricate, complicated systems-be they minds or engineering systems.
Its most obvious application is to computation and computers, but I
believe it is equally valuable for clear thinking about biological, psycho­
logical, mathematical, and (especially) philosophical questions. This
claim must seen impossibly pretentious, and I had better explain it: The
theory of effectiveness is useful not only to prove things about complex
systems, but is also necessary to prove things about proof itself l
The overall strategy of the book is very simple. We will try to engage
the concept of effective procedure from such a variety of contexts, for­
malisms, and viewpoints, that its character will become thoroughly under­
stood-in the sense that one can recognize it and adapt it, almost without
thinking, wherever and whenever it is appropriate. This concept has so
many different aspects that one cannot hope to know them all equally
well. It appears unexpectedly in aspects of practical computation, lin­
guistic theories, and mathematical logic. We must know some of these
aspects very thoroughly, and others at least casually, if we are to be able
to sense its style of operation in new situations. Both the text and the
problems are designed with this polyvalence in mind: we pursue a few
lines much further than most, but all these approaches keep linking and
intersecting to form a web of weirdly different but connecting threads.
Most fortunately, the theory is almost entirely self-contained, so that it
has no formal mathematical “prerequisites” beyond a reasonable high­
school algebra course. A superficial glance through these pages might
give a bewildering impression of the use of many complex and technical
formalisms, but this is most definitely not the case. The book is not
written in the formalisms, it is merely about them, and one is neither
required nor expected to ever have seen them before!
PREFACE ix
The book has also a second, more playful theme. The many-headed
concept which we call effective compatability often appears in an especially
potent form-the Universal machine or system. This character has an
astonishing variety of secret hiding places which we keep trying to ferret
out. It is easily found in the ponderous operation of a modern digital
computer’s order-code, or the gross, redundant structure of a FORTRAN
or ALGOL-like programming language. (It is not important, at least in
this book, that you know what these are.) But it can also lurk within the
tiny structure of a “computer” with two memory registers or two
instruction-tapes (Chapter I4), or in the seemingly childish bead-Stringing
primitives of a Post normal system (Chapter 13), or even in a machine
that (almost) neither reads nor writes on its tapes. While there is not
much practical use in finding these “minimal Universal systems” they
support our main goal by making us aware of exactly what is essential
rather than merely practical or convenient in the concept of computation.
Perhaps even more important, we learn of many difTerent ways in
which great complexity of behavior can arise from the interactions of
simple devices, simple actions, simple descriptions, or simple concepts.
We see also many different ways in which the same complex behavior can
arise from what appear to be utterly different kinds of interactive systems.
The reader is prepared for both of these explorations in Part I, which
is a training expedition into a smaller but essential (and in many ways
analogous) territory. Within the finite-state machine area there exists as
wide a variety of appearances of its own indigenous “central concept”
(we explore fewer of these, though) and almost as surprising an assembly
of its own minimal hiding places.
I wish I could have captured in the different approaches to the theory a
sense of the different styles of thinking that lead the early explorersm
Post, Church, Turing, Kleene, McCulloch, and the others-to approach
the same problems with such difTerent (technically and esthetically) formal
methods. The personalities are obscured because in my attempt to
modernize, simplify, and unify the results I have removed the individual
details of the original developments. It would be well worth one’s time
perusing Davis’ (1965) collection just to gain additional personal con­
tact with the masters!
MATHEMATICAL TECHNIQUES
As we stated above, the book is largely self-contained, given enough
high-school mathematics to know about functional notation (even this is
reviewed in Chapter 7) and mathematical induction (reviewed in Chap­
ter 4). MIT freshmen and sophomores have encountered no difficulty
with the material, and though they may be more sophisticated mathe­
X PREFACE
matically than most college students, I cannot imagine any good science
student or computer programmer having serious trouble. I can, however,
imagine educators (even mathematicians who have not encountered ‘“re­
cursive function theory”) becoming suspicious and wondering whether I
have somehow constructed a simplitied or diluted version of the theory by
“leaving out the calculus.” Quite the contrary: the calculus was never
in it! This really is a branch of mathematics that stems directly from
non-numerical, logical foundations. Indeed, a number of research
workers are currently attempting to bring together this theory with that
of Newton and Leibniz; the resulting new subject (currentlycalled “re­
cursive analysis”) has barely gotten offto a start.
It is worth a few words to try to explain why classical mathematical
analysis plays such a small role here. In developing a Theory of Compu­
tation we are trying to deal with systems composed of a great many parts,
or very intricate structures. Classical mathematical methods can do this
only in very special situations, and their limitations are very serious.
Classically, one is unable to cope with even a few simultaneous non-linear
equations, to say nothing of a few dozen, or a few million. Now it is true
that under certain special conditions mathematical analysis can “revive,”
as it were, when the situation gets complex in such a way that the parts of
the system can be treated as individually and independently random­
this is what happens in Statistical Thermodynamic theories. But it must
be stated, explicitly and emphatically, that this is just what does not
happen when, as in a computation system, the structure has a more or­
ganized, purposeful structure. The statistical analysis works beautifully
for things like gases. It works for precious little else. There simply is no
reason to suppose that as computations grow large, one will discover any­
thing non-trivial by trying to “average-out” the effects of many events.
The effect of the “conditional” (see Chapter 10) is too strong to allow
anything like a “conservation” concept to have a place in the theory.
Fortunately, the systems of computation have other features that make
possible some analysis, though of a very different kind. Instead of the
statiszically defined events used in physics, we use logically-defined classes
of computations or expressions. They are tied together, not by geometric
or energetic properties, but by their relations to similar machines or simi­
lar definitions. We can use machine parts so simple and with such simple
interactions that we can apply the utterly transparent Logic of Proposi­
tions, where for an equivalent actual physical machine we would have to
solve hopelessly opaque analytic equations. Chapters l, 5, and I2 explain
in detail how the problem is reduced until this kind of analysis can be
made. One could even complain that the main results are obtained by
a most deceptive sort of parlor trick: instead of trying to solve equations,
we study their descriptions! It must be admitted that in the long run this
PREFACE xi
method avoids entirely too many problems, and we can be quite sure that
the development of the Theory ofComputation, in the years to come, will
be centered around the problem of how to back off a little from this sim­
plitication, without going all the way back. The trouble now is that we
do too much in one step.
AS A TEXT
This book has the form ofa textbook, and I have used earlier versions
of it at MIT for several years. Students specializing in computer-related
studies can absorb most of the material in a one-term course, since anyone
with computer-programming experience is already subconsciously primed
with the essential concepts, and needs only to see them precisely formu­
lated. For them, or for students with some higher mathematics courses, I
would be inclined to assign Part I as a reading exercise, have them work
the problems, and begin “teaching” with Part II. In any case, I feel that
the material of Part I ought to be integrated very early somewhere in
every scientific or engineering curriculum; the material of Parts II. and III
can come considerably later (with the exception of Computer Science
students). This edition is also designed for independent reading outside
the classroom, and that is why I have given substantial hints for solving
most of the harder problems.
THE PROBLEMS
The reader is asked to solve a number of problems scattered through­
out the book. Only a few ofthese are routine: most of them point toward
branches the theory could take, and a few of them are diflicult research
problems. The emphasis is on the application of new concepts to new
situations rather than on acquiring technical proficiency through drill
and repetition. My view is that the best way to understand a new idea is
to discover by oneself how to use it to solve one or two reasonably hard
problems; the point can easily escape after a hundred simple exercises.
Solving a hard problem once can make the ideas a part of one’s thinking
resources.
The reader is therefore enjoined not to turn too easily to the solutions;
I10t unless a needed idea has not come for a day or so. Every such con­
Cession has a price-loss of the experience obtained by solving a new
kind of problem. Besides, even if reading the solutions were enough to
HCquire the ability to solve such problems (which it is not), one rarely
finds a set of ideas which are at once so elegant and so accessible to