DESIGNING INTELLIGENCE: A FRAMEWORK FOR SMART
SYSTEMS
Steven H. Kim
Oxford University Press
New York, 1990.
Hardback, 273 pp.
(Reviewed, May 1998)
As occasionally mentioned in other sections of this database, it is possible
to observe that human knowledge, as wonderful as it has become, possesses
some Knowledge Gaps, Holes, Vacuums and Darkened Abysses. And in order
to establish relevance to the central contests of this database, this phenomenon
needs slightly to be entered into before going into the review proper.
The knowledge-gap observation doesn't refer to what has not yet been discovered
on the frontiers of the yet unknown. It refers to human factors that have
been around for a very long time -- such as, but not only, the factors
regarding:
(1) the wondrous spectrum of human abilities;
(2) the qualities and mechanisms that result in power and power structures;
and
(3) the nature and functions of intelligence.
These familiar factors have also been acknowledged as having crucial importance
and meaning -- but for which very little in the way of exact, organized
knowledge has accumulated. Thus, information about these factors can exist
in disorganized and random ways -- and possibly include various amounts
of cow pucky.
One somewhat easy way of identifying knowledge abysses has to do with becoming
aware of encyclopedias that are missing within the overall knowledge pools.
Ever since the Enlightenment, the educationally-minded among our species
have devoted considerable efforts and enthusiasm to producing encyclopedias
and compendiums that bring together, organize, list, describe, and detail
everything that is identifiable and known about categories of phenomena.
Encyclopedias in fact are database storehouses of cumulative information
-- and they also represent the fuller extent in various categories of what
is fondly referred to as "knowledge."
Thus, we possess encyclopedias or compendiums regarding just everything
imaginable -- except, for example, a detailed encyclopedia dealing with
power and its mechanisms is missing. Speaking only hypothetically, of course,
this missing informative run-down regarding power might be explained (even
justified in the minds of some) as a way of preventing any real understanding
and learning about the "secrets" of power -- thus keeping the
secrets out of the hands and minds of the many in favor of isolating them
in the hands of the few.
In any event, a broadly accessible encyclopedia of power and its elements
would certainly more widely increase understanding of power, its nature
and mechanisms. But this would also increase the number of contenders for
power -- and thus enlarge complications regarding power games and whom
is to have power. After all, power has traditionally been thought of as
something the few should possess, in order to have power over the many
more powerless. Hypothetically, then, any in-depth encyclopedia of power
might be considered a "threat" of some kind -- and so the best
preventive against this is not to encourage the availability an encyclopedia
of power.
Theoretically speaking, the spectrums of human abilities and intelligence
can be thought of as having something to do with power. In actual practice,
however, one might come to question this theory -- at least in the cases
of those who achieve power but somehow fail to exhibit too much in the
way of abilities or intelligence.
But it can be observed that if access to ability-enhancement is made available
to the depowered masses, then those few in power would probably have novel
problems to deal with. Along the same lines, encyclopedias that might result
in enhancements of intelligence hypothetically might increase more wise
observation and criticism of the powerful -- perhaps even help illuminate
the nature of stupidity (for which, by the way, an encyclopedia is likewise
missing.)
As it is (probably), ability and intelligence enhancement could be seen
as "approaches" to power, and such approaches might be viewed
as threats of no little consequence. If this would be the case, then approaches
to power must be equally guarded as is power itself -- with the result
that no competent encyclopedias regarding the approaches would appear either.
Needless to say, the superpowers of the human mind -- hypothetically also
having something to do with abilities, intelligence and power -- don't
enjoy an exhaustive encyclopedia. At least some of the reasons for this
missing encyclopedia can be obvious.
If we could suppose, hypothetically, that development and enhancement of
the biomind superpowers would introduce variables into the usual Earthside
power structures, then those same power structures would view the superpowers
with something less than enthusiasm. As it could turn out, the new superpower-endowed
contenders would have "advantages" the unendowed did not have
-- advantages such as mind-reading, paranormal spying, direct mental infiltration,
enhanced foresight, clairvoyance; perhaps even psychokinesis that might
be utilized to melt not only brains but also the usual armaments of physical
weapons and learned stupidity.
However, the worst scenario might consist of the possibility that there
could not be any SECRETS or SECRECY -- or behind-the-scenes dealing to
the advantage of the powerful few and to the disadvantage of the powerless
many. In any event, it is generally understood that secrecy has some importance
to power, and so anything that might unmask secrecy is not viewed with
enthusiasm.
One of the possible results of this hypothetical situation is somewhat
amusing. As its history demonstrates, a species exists Earthside that possesses
biomind superpower faculties. But that same species subtly (if not overtly)
discourages any detailed knowledge of the faculties -- because any broad
understanding and development of them might unsettle the status quo of
the species vested power games.
If, then, any encyclopedic clarification and detailing of the spectrums
of human abilities and intelligence were to take place, well, the reasons
would have to be quite compelling. One such compelling reason, an easily
recognized historical one, is economic profit via the establishment of
new markets.
And such a compelling reason has recently taken place regarding INTELLIGENCE.
And thus, with novel and significant markets in view, it has become imperative
to organize, make sense of, and understand everything that is known, or
can be known, about it. And hereby, an amusing tale commences -- and which
roughly began some time ago with the concept of installing "intelligence"
into machines. This is to say that intelligence has become a commodity
-- at least as regards intelligent machines.
In its first conceptual incarnation, such an intelligent machine was dubbed
as a "robot" -- defined as "an automatic apparatus or device
that performs functions ordinarily ascribed to human beings or operates
with what appears to be almost human intelligence." As of the 1990s,
however, robots are no longer romantically termed as such -- that concept
having been replaced by the concepts of artificial intelligence and smart
(mechanical) systems.
The evolution from robots to smart systems has had its wondrous and uplifting
techno-achievement aspects that have excited and challenged many fine research
intellects. But it also has had a quite humorous aspect. This consists
of the problem of installing intelligence into a mechanical system that
is supposed efficiently to end up FUNCTIONING as a smart system all on
its own.
If this problem is to be solved (and of course ultimately capitalized upon
by marketing of smart systems) then it is necessary to understand and comprehend
all possible details about what intelligence is -- this in order to more
effectively and intelligently install it into smart systems, so that they
can be smart in the first place. After all, it would be somewhat counterproductive
to want to install intelligence into machines if one doesn't know the ins
and outs of what is to be installed. There is really no point in installing
dumb intelligence into hopefully smart ones.
Thus arose the problem of finding out what intelligence really is, what
its mechanisms really are -- not so much as to enlighten and enhance humans
about human intelligence, but in order to seize upon intelligence so as
to install it into smart mechanical systems.
Now, this problem has two major aspects -- the first being that in order
to entertain achieving the goal of what intelligence really consists of
fundamentally, functionally and mechanistically, one needs somewhat to
transcend theoretical approaches to it -- such theorizing equating to the
scientific approach. Theories about intelligence didn't work all that well
with robots.
The second aspect consists of a straight-forward technological approach
-- i.e., in this case the study of natural, already-existing intelligence
systems found in nature. If the intelligence mechanisms already existing
could be understood and mechanically replicated, then the naturally-existing
systems could be used as models regarding "how-to" duplicate
and install similar intelligence systems into machines.
As it turned out by the early 1990s, one of the better models for naturally
existing intelligence systems is to be found in the human being. The outcome
of this discovery is that human intelligence, for the first time, is being
intricately researched and studied right down to its quantum attributes.
The emphasis, perhaps, is not on what intelligence is, but on how it functions
-- all this not particularly on behalf of enhancing intelligence in humans
per se, but on behalf of translating human intelligence into smart mechanical
systems with obvious, and enormous, economic prospects.
Even so, for the first time those humans interested in comprehending and
perhaps enhancing functional knowledge of THEIR naturally-indwelling intelligence
factors have a chance of benefiting from the encyclopedic itemization of
human intelligence being utilized as a chief model to make smarter and
smarter machines.
This possibility is, of course, both ironic and hilarious -- since the
understanding of intelligence and its many astonishing mechanisms does
have implications to enhancing the spectrum of human abilities, enhancing
approaches to and development of power, and, as well, to the superpowers
of the human biomind.
You see, it is possible to consider that knowledge only acts to the degree
that it constructively feeds-back into intelligent systems that can comprehend
and appreciate it. THIS feed-back loop would be a criterion in the case
of intelligent machines, but it is also a function of human intelligence.
And so discovered attributes and mechanisms of human intelligence being
utilized as an EXACTING model for creating smart machines can easily feed
back into the intelligence routes and systems of the model.
DESIGNING INTELLIGENCE: A FRAMEWORK FOR SMART SYSTEMS, by Steven H. Kim,
was published ten years ago (in 1990). Dr. Kim probably did not intend
his book to represent an encyclopedia regarding the nature and functions
of intelligence.
But otherwise it is so well organized and detailed that it can easily stand
in for such an encyclopedia -- not only for those focusing on the wondrous
technological and economic potentials of smart systems, but regarding those
smart systems that might, with attention, chance to read it. And in this
sense, the cumulative and organized knowledge sense, the book cannot be
dated or obsolete.
By way of review of this book, as stated in the blurb on the flap of DESIGNING
INTELLIGENCE:
"Intelligent systems assume a critical role in society as computers
continue to expand into scientific and technological fields as well as
management and everyday life.
"Intelligent structures are used by biologists in modeling adaptive
processes, by cognitive psychologists for exploring reasoning mechanisms,
by engineers designing microprocessor-based devices, by industrialists
in planning automated factors, and by strategists in developing organizational
networks.
"However, the field of intelligent structures has lacked a systematic
theory and even a coherent framework for conceptualization.
"In this book, Steven H. Kim [Prof. of Mechanical Engineering, and
Director of the Knowledge Systems Program at M.I.T.] presents a systematic
basis for analyzing and synthesizing natural as well as artificial intelligent
systems.
"After defining the nature of intelligence and its attributes, he
provides an overview of intelligent systems -- from biological entities
to robots, automated factories and organizations.
"The book will interest all those involved with the challenges of
today's technology and its future application, including students and researchers
in artificial intelligence, computer scientists, practicing engineers,
business managers, and research and development scientists in many fields."
This book should also conceptually interest, in self-feedback ways, anyone
interested in activating and potentializing not only their own intelligence
thresholds, but various biomind superpowers -- and which, as loosely considered,
must be conceptualized as proceeding from some kind of intelligence systems.
It might be taken for granted that INTELLIGENCE is a quite involved, even
awesome subject. This may be one of the several reasons its parameters
have eluded philosophical, religious, scientific, cultural and sociological
comprehension -- and which themselves are quite involved, and often principally
based on bias, preferences and personal or social convictions.
DESIGNING INTELLIGENCE, however, presents information about intelligence
from the viewpoint of inserting it into smart mechanical systems. This
cannot really be done if comprehension of intelligence and its mechanisms
remains cluttered with misinformation or liberal amounts of counterproductive
cow pucky.
In other words, workers designing intelligence to be installed in machines
need what amounts to encyclopedic knowledge of it, and in this sense must
organize that knowledge in detailed ways. Dr. Kim's book has the advantage
in presenting knowledge regarding intelligence in what amounts to organized
encyclopedia-making methods.
The book is divided into five Parts, the first of which presents an "Introduction"
and "Overview of Framework." Says the author: "A science
of intelligent systems should consist of a series of layers for describing
intelligence phenomena. These levels reflect a progression from the conceptual
and qualitative to the rigorous and quantifiable."
He then proceeds to organize both the information and the reader's in-take
of it by indicating (*) a framework for discussion, including a systematic
set of concepts; (*) a model describing the relationship among the objects
of the framework; (*) a theory consisting of a set of principles describing
the fundamental behavior of such systems, as well as implication for their
synthesis.
Part Two then breaks down, or apart, Factors of intelligent systems as
to Purpose, Space, Structure, Time, Process and Efficiency. Part Three
then considers Interfactor Trade-offs of Space versus Time, of Mechanism
versus Process.
Part Four considers Application, mostly in the direction of creating smart
mechanical systems -- but the considerations also can feed back into readers'
smart system.
The book has six Appendices: A General Model of Design (which is pertinent
to intelligent systems, even in the human models, need to undergo and be
shaped by some kind of design; Predeterminism in Reasoning and Creativity;
Axiomatic Approach to Design; A Formal Framework for Learning Systems;
A General Model for Information; and, Levels of Intelligent Design.
On page 237 appears the following statements under the heading of Purpose:
"The design of an expert system, as with other engineered products,
begins with the perception of a need and its translation into a set of
specifications. The purpose of a knowledge-based system is to liberate
or assist a user in making decisions within a domain of applications. By
incorporating knowledge from diverse sources, the performance of an expert
program can match or even exceed that of humans. . . .
". . . enhanced capabilities are possible through the integration
of knowledge from a spectrum of sources, including multiple human experts
and heuristics from design manuals. Another mechanism for quality enhancement
is through a systematic exploration of alternative designs rather than
a straightforward dash for the first [apparently] satisfactory solution."
An essential, and very important by-product of this book is that it can
be seen that human intelligence, existing naturally (or in a "raw"
state) within our species and specimens of it, undergoes various "designing"
in order to translate "raw" or potential intelligence into working
within "a set of specifications."
Whatever else "a set of specifications" might imply (cultural,
philosophical or sociological), the "set" implies limitations
and boundaries, because without these the "specifications" cannot
serve as specifications.
Thus, in the individual sense one's intelligence in its "raw"
state may be very large, so to speak, but that it is likely to be "designed"
into "specifications" that are much smaller. The likely outcome
might then be that the intelligence of the individual will "work"
or "perform" only within the constraints of the "designed
set of specifications."
On page 201, the author points up: "An intelligent agent should be
able to learn from its experiences rather than remain forever captive to
its initial set of facts and inference procedures [i.e. captive to its
initial intelligence or mind-programming.] In this way, the agent may break
free of its original limitations, improve in performance over time, and
perhaps even generate new results that might be labeled creative."
On page 180, Dr. Kim indicates: ". . . an intelligent framework should
also facilitate its own development. This may be achieved by exploring
tools for employing the framework against its own attributes [or designer
set specifications] as illustrated in Appendix F.
"The six factors of intelligence -- purpose, space, structure, time,
process, and efficiency -- satisfy these objectives. The appendices explore
a number of these critical issues, and provide a glimpse of the road ahead
in the systematic exploration of intelligence and its implementation in
engineered systems."
One of the major "messages" of this book, one somewhat hidden,
is that by utilizing their intelligence as a model to build smart mechanical
systems, humans can test the workability of designed intelligence. And
if it works in smart systems, then the workability can, so to speak, be
reverse engineered back into human knowledge in order to efficiently redesign
and enhance human intelligence systems.
This book is somewhat "technological" in places. But much of
the information in it is accompanied and clarified by illuminative graphs
and charts, and it can be read easy enough without becoming seriously defeated
by its technological passages. It may be out of print by now, but it is
well worth an effort to ask Amazon, etc. to find it. Various aspects of
this book will be discussed in forthcoming essays to be placed in this
database.
The overall "message" of this book is that former concepts about
intelligence can be hauled to the cultural landfills -- that new age of
understanding and enhancing intelligence has begun. In short, this book
is TERRIFIC.