Air University Review, May-June 1976

Do computers facilitate, or do they control? How would you answer if you thought the computers were coming to life? Ponder the madness of computer intelligence in control of our vast network of communications, logistics, and weapon systems. Science fiction nonsense? But surely history is replete with science fiction "told-you-so's." Moreover, there have been hundreds of non-fiction works published over the past 20 years which seriously declare that computer consciousness is not only possible but likely.

With the possible exceptions of IBM and AT&T, the U.S. Air Force stands out like no other organization in its image of vast modern technology packed full of computers. Surely we should be the last to be unclear on this matter of machine intelligence--even for a discussion over a Friday beer at the club.

In hopes the reader will finish with at least a more comfortable answer to the first question, let us apply a cool but decidedly unmechanical approach to the issue.

One of the oldest computer jokes around tells of the latest IBM creation being fed the entire works of Saint Thomas Aquinas and then being asked, "Is there a God?" After a few tense seconds, a printout appears with the words, "There is now."

Despite its antiquity, this story serves well to keynote the eerie feelings that grip some of us as we learn of the amazing capabilities of today's computers. Moreover, the apprehension goes back further than the joke.

In the last century, English author Samuel Butler wrote "The Destruction of the Machines of Erewhon," foreshadowing serious consideration of machine life by many later experts. For most of us, this kind of speculation vaguely agitates our common sense, but few of us do more than shrug it off, and some of us believe it might actually be possible. As with many other subjects, we have become victims of our dependence on the "explanations" of the experts instead of masters of our own analyzing abilities.

This time, let's respond to our common sense, consult known scientific knowledge, examine the data, and get at the truth. We can begin by reviewing Butler's forebodings.

Butler feared that man would continue to evolve machines of increasing propensity for becoming living species. In fact, he wondered if they were already conscious, patiently plotting their evolutionary perfection and eventual domination of the world. If not, he was sure that one day man would produce a mutant having consciousness and independence. Moreover, he implied that the machine's consciousness would spring into being--like Pallas Athena, fully panoplied--with a value system underpinned by self-survival. You see, Butler's machines would just naturally view man as both enemy and servant: a threat to survival, yet essential to it. The machines would therefore ruthlessly conquer and mercilessly control mankind, that is, unless the hapless victims acted quickly to destroy all machines.

Butler quite obviously did not understand the concept consciousness. Just as obviously, he did not understand the conceptualization process required for extending the primitive survival instinct to the abstraction threat. Let us deal first with the latter.

To do this, let us imagine that, somehow, a machine has developed consciousness. Does this mean that it automatically has a tendency to survive? If it does not have this "value," it would certainly be indifferent to its own continued existence. It therefore could not regard anything as a threat. The term "threat" can have no meaning to an entity that does not value its own life. In fact, the behavior of such a conscious machine would be indistinguishable from a nonconscious machine. Without a survival or life value, no other value is possible. With no values, the machine would go on behaving according to the laws of physics, no more. So, we must conclude that, for the machine to be dangerous, it must awaken with a survival value.

Where does it get this value?

An updated Butler, though free of his Victorian assumptions, might still argue that a survival value is inherent in consciousness. If so, he must necessarily produce his proof from our knowledge of biology. He should not mind, therefore, if I too turn to that field. He might even be happy to let the argument stand or fall on my analysis, which purports to show that, by correcting some false notions traceable to the field of biology, the idea of machine consciousness becomes nonsense. If I am successful in presenting that analysis, the question of whether machine consciousness would include a survival value becomes moot.

On the other hand, if my intrepid opponent would bring the machine to life outside the laws of biology, that is, by an undiscovered process which contradicts those laws, he must compound the fiction by giving it a survival value from the same thin air from which he gave it consciousness. Keep in mind that we are dealing with reality, laws of nature, laws which have been discovered, not invented.

Finally, if the insistent mystic argues for miraculous machine consciousness without a survival value, he will have circled back to the here and now. He, like some modern scholars, would be indulging in the game of describing our current, nonconscious electronic automatons with undeserved and illogical definitions.

Aided by twisted concepts of biology, some experts blithely continue to resurrect the notion that somehow the complexity of construction in a computer, together with its memory and learning abilities, will produce synapsis and bring a living being into existence. What does science say about consciousness? What are the errors the experts make that permit serious consideration of computer consciousness?

In the first place, most writers on computer development make the mistake of using the term "memory" as a primary, with a definition equally applicable to both computer and man. The justification is seldom explicit, but the implicit "logic" is that both man and computer exhibit some of the same memory characteristics (input, storage, retrieval); therefore, they both possess a memory. Several writers have described the differences in human and computer memories as merely differences in speed and size.

John G. Kemeny, President of Dartmouth College, in his book Man and the Computer, points out that man's memory is slow but has an enormous capacity, while the computer's is fast but has a relatively limited storage volume. Together, argues Dr. Kemeny, they can exponentially compensate for each other's deficiencies and move mountains. He argues for symbiosis, after explaining that he already regards the computer as a species. He implies that, if a computer has a memory, it must be conscious, even if at a level not yet recognized by man. He is right, of course. If a computer tally has a memory, it is indeed conscious. But Dr. Kemeny gives no hint that he understands the reason for the correctness of that position. Where he is wrong is in thinking the computer has a memory.

"Memory" for a computer is a component of materials in a separate, uniquely locatable place with assigned functions. Man's memory is a mental process subsumed under the concept consciousness, has no specific location, and performs intricacies far beyond the assigned retrieval functions of the computer "memory." Observe Dr. Kemeny pondering the human "retrieval" system:

…we are remarkably efficient at retrieving items from our memory. We seem to be able to do it through mysterious processes of association that no one has duplicated on the computer. How does one associate a phrase in a book with a conversation held ten years ago? How does one associate a smell with a childhood memory? How, in trying to solve a problem, do we pull out three unrelated memories going back to different periods in our life to come up with a new approach to the solution? And how do we sift through hopelessly large amounts of information presented to us daily by our senses and retain just the most relevant facts?¹

The phrase "retrieving items from memory" is perfectly appropriate for describing the action of a computer but is grossly misleading when applied to man or to any other living, conscious entity. Man does not activate a probe to scan a storage unit located under the cortex in a corner of the hypothalamus. He does not use positive and negative charges in binary or octal code to match location core address to call address. (A process, incidentally, which is independent of address content!) He does not then send the content forward to a display or process register. This is the computer procedure for retrieving an item from "memory." Computers retrieve, man remembers, two immensely different processes.

What notions helped obliterate the distinction? Part of the problem is the naïve awe with which some observers regard the astonishing speed of the computer, as if this alone endows the machine with intelligence and therefore with life, or at least the name "memory" for its retrieval system. But the main culprit comes in a package of false biological ideas, which seem to have influenced Dr. Kemeny and others who write on computers. These ideas spring from the "theory of reduction," which rests on the premise that all the phenomena of life can be accounted for, described by, and deduced from the laws of physics and chemistry; thus, by "reduction," an inanimate object has consciousness. That theory has been thoroughly refuted by Robert Efron, M.D., Chief, Neurophysiology-Biophysics Research Unit, Veterans Administration Hospital, Boston, Massachusetts, in his paper "Biology without Consciousness--and Its Consequences," presented at the Center for Philosophy of Science, University of Pittsburgh, on February 27, 1967.²

Dr. Efron explains that learning and remembering are two related mental processes performed by the conscious mind of a living entity. Just as walking is a physical process performed by a living entity and cannot occur independent of an entity, neither can learning or remembering occur as primary and independent phenomena. The reductionists, however, gave new and unsupported meanings to the terms "learning" and "memory," disconnecting them from consciousness. By what convoluted reasoning have the reductionists foisted these impossibilities onto the field of biology? The answer will help us see how these ideas landed unchallenged in the field of computers.

Dr. Efron shows that the reductionists use the epistemological method of definition switching, which consists of 

arbitrarily defining the contradictions away. The reductionist attacks the definition and usage of every word which has historically referred to an action of a living entity: "memory," "reflex," "free will," "cognition," etc. He then redefines that same word so that it will be applicable to an action of an inanimate entity.³

In the process, the reductionist uses the "stolen concept" fallacy,⁴ whereby a concept is used while ignoring, contradicting, or denying the validity of the concept on which it logically and genetically depends. For example, the concept orphan cannot exist without the antecedent concept parent. Similarly, the internal contradiction in the assertion "There are no absolutes" is often overlooked even though that very assertion is an absolute and depends on the existence of the concept absolute to have meaning and is therefore, by ipsissima verba, false. Electron is subsumed under (not a component of) mass and charge; as an electron has mass and electrical charge, we cannot deny the concepts mass and charge and still expect "electron" to have meaning. Action presupposes entity (that which acts); we cannot have a disembodied action any more than we can have remembering without a consciousness to perform that process. To declare that all definitions are only approximate is to be guilty of the same error. The concept approximate is not primary; it has meaning only in contradistinction to exact. It is by such trickery that the reductionists use the concept memory while ignoring its antecedent concept consciousness. This is the same error committed by computer experts when they refer to a computer’s retrieval system as its "memory." They follow the same pattern when speaking of the computer's "learning" abilities.

Computer scholars have borrowed, perhaps unwittingly, the reductionist's redefinition of "learning" when describing the computer’s performance alteration process. When a computer system compares newly sensed data with previously accumulated data according to programmed evaluation instructions and thereby modifies its actions, it has not "learned"; it has altered its performance. But this is learning, according to the reductionists' distortion of the term. What did the reductionists do to the historical concept of learning? They broke it off from consciousness in order to make way for the degeneration of memory to a function independent of life. Dr. Efron reports:

The link between the concept of memory and the concept of consciousness was explicitly broken when "learning," like all other concepts of mental functions, was redefined in the twentieth century. Once the concept of consciousness was eliminated from the concept of "learning." it necessarily vanished from the concept of memory.⁵

The reductionists had narrowed down meaning of learning to cover "only . . . the alteration of behavior or of performance as a result of a previous experience." This is most adequate for describing the "learning" process of computers. It also fits perfectly for describing the decomposition (alteration) of the human body after a fatal heart attack (previous experience). A corpse advancing in putrefaction, a computer varying a design process to accommodate new data combinations and a scientist discovering the means thwart a disease are all engaged in "learning." This is no more absurd than some of the applications of the new definition of learning found by Dr. Efron in the field of biology and some I have found in the field of computers.

H. O. Simon, in his article "The Shape of Automation," states, "We can now write programs for electronic computers that enable these devices to think and learn."⁶ His "proof'" is similar to that presented by other writers; he shows that a computer can be programmed to run a factory entirely automatically, detecting and compensating for deficiencies, recognizing variations, etc. In "The Psychology of Robots" Henry Block and Herbert Ginsburg say the computer is learning when it uses sensory devices to seek data, store them, evaluate them, and react to them as programmed.⁷ (Unexpected results reinforce the notion that the machine "learned,") G. Rattray Taylor, in an article on androids, explains the game-learning process of a computer thus: The rules are programmed in and the computer discovers the best move "substantially in the same manner [as] you and I . . . by looking at a lot of results and noting which ones pay off." (Note the embarrassing spectacle of a scholar stating the case and pretending to have thereby explained it!) He admits to altering the definition of learning, or to sensing that he is using an altered definition, when he says, "Unless, therefore, we define the word 'learn' in a very narrow way, we are bound to speak of these machines learning."⁸ Taylor may be so bound, but, as we are interested in explication, not drama, we will use clearer thinking.

The distinction between human learning and computer learning is so profound that it is preposterous that the two processes share the same name. Dr. Efron tells us that, before the reductionists changed it, the term "learning" referred 

to the acquisition of perceptual discriminations, acquisition of motor skills or habits, problem-solving, adaptation, association, "insight" solution, secondhand concept acquisition, and … the discovery of new conceptual knowledge.⁹

Only part of this list can be made to fit the computer "learning" process, whereas all are needed in combination to describe the achievements of the human consciousness. Furthermore, the abilities of the computer are totally and inextricably tied to measurement--specific, carefully detailed quantification. All computer programming and evaluation are carried out in terms of mathematics, whereas the human mind can perform feats of recognition, evaluation, and general conceptualization without reference to specific measurement. Reflect on how, after practicing algebra, you are able to make exchanges between both sides of the equation, skipping whole steps and sometimes even "seeing" the solution without performing any of the steps. Consider the differentiation and integration process of human value formations. Such words as "intuition," "curiosity," and "sentiment" have implications in human learning completely unanchored to either measurement or logic--those twin essentials of the computer.

Since it has come up, let's pause a moment and look at computer "logic." Logic, for a computer, refers not to the comprehensive art of noncontradictory identification but rather to a method of procedure. That is, the computer uses a method that we normally think of as logical, rigidly logical, untempered by reason. If the programmer is not careful, the "logical" results can be annoyingly contradictory or totally unintelligible, as in the case, for example, of language translation machines. Computer logic is purely mathematical. For example, truth tables used by computers employ the relational operator "or" to connect two statements (operands) that were previously identified by the logical value true or false. If either statement or both are true, the conclusion is computed true. In real logic, however, two mutually exclusive propositions cannot both be true. For these reasons, my reference to measurement and logic as "twin essentials of the computer" is inexact. It would be better to say, measurement (quantification, mathematics) is the absolute essential of the computer; and its processing method is rigidly "logic-like."

Now let's return to the point we were discussing before this digression on logic.

The human mind performs learning feats totally unconnected with specific measurements or any form of quantification. These are feats impossible for computers. In regard to "discovery of new conceptual knowledge," mentioned by Dr. Efron, think about the phenomenon of serendipity and note the learning experiences suggested by such names as Pythagoras and Newton. Man's distinguishing characteristic, his rational faculty, is the means by which these conceptualizations take place.

Human concept formulation is a progression of abstractions from concretes, abstractions from other abstractions, and abstractions of consciousness. I will quote from Ayn Rand's outstanding paper on the human learning process:

A concept is a mental integration of two or more units possessing the same distinguishing characteristics, with their particular measurements omitted.¹⁰

In fact, we formed many concepts, such as color, before we knew how to measure them.

To form concepts of the consciousness, one must isolate the action from the content of a given state of consciousness, by a process of abstraction. Just as, extrospectively, man can abstract attributes from entities--so, introspectively, he can abstract the actions of his consciousness from its contents, and observe the differences among these various actions.

For instance . . . , when a man sees a woman walking down the street, the action of his consciousness is perception; when he notes that she is beautiful, the action of his consciousness is evaluation; when he experiences an inner state of pleasure and approval, of admiration, the action of his consciousness is emotion; when he stops to watch her and draws conclusions, from the evidence, about her character, age, social position, etc., the action of his consciousness is thought; when, later, he recalls the incident, the action of his consciousness is reminiscence; when he projects that her appearance would be improved if her hair were blond rather than brown, and her dress were blue rather than red, the action of his consciousness is imagination.

He can also observe the similarities among the actions of his consciousness on various occasions, by observing the fact that these same actions--in different sequences, combinations and degrees--are, have been or can be applicable to other objects….¹¹

Computer programmers and designers who can grasp this will dismiss out of hand any suggestion that a computer could be programmed to perform these feats. A computer can be programmed, for example, to classify objects by color or shape but cannot be programmed to conceptualize "classification" (or anything else). Without the ability to abstract classifications, the computer is unable to recognize entities without reference to specific measurements. You will find that attempts to prove otherwise are merely more exercises in stretching definitions to fit pet "insights."

To summarize, when correctly used, the words "memory" and "learning" stand for concepts that presuppose consciousness and therefore apply only to living, actively aware entities. Memory and learning are subsumptions (not components) of consciousness. Among living entities, only man has the ability to conceptualize, an ability by which be can grasp the abstractions memory, learning, and concept, either approximately, by intuition, or fully with scientific knowledge and objective epistemology.

To imagine that scientists would attempt to upgrade the computer's data retrieval system to a full memory and its performance alteration system to full learning ability, with the aim of bringing about consciousness, is ludicrous for two reasons. First, such an effort holds little promise of improving computer efficiency, and therefore motivation goes begging. Second, and more to the point, the attempt would be much like endeavoring to sculpt clay to the shape of the molecules of gold, hoping for an exactness that would bring about the alchemist's dream. We do not fully understand the force that binds the molecules, but we can observe and understand its effects. Likewise, we do not fully understand the neurophysiological "force" that provides the synaptic confluence called consciousness, but we can observe and understand its processes. There is little to justify the hope that we could stumble onto it by building improved computer components, expecting them to behave as subsumptions.

There is even less justification for speculating about a computer consciousness spontaneously springing into existence during the process of evolving better machines. Such a miracle would require repeal of the laws of biology and physics, i.e., a consciousness effect without a biological cause, something created from nothing. By definition, such speculation takes us out of the world of reality and into that of mysticism. Let's leave that for the entertainment provided by fantasy writers.

In short, what I am saying is this: if computers "learn," they are able to remember; the one goes with the other. If a computer can remember, in the accurate sense of the word, it is indeed conscious, for remembering is something only a conscious entity can do. After examining the real meaning of learning and remembering it is clear the computer can do neither. It is therefore, no more than another in the long line of magnificent nonconscious achievements of the mind of man.

I hope my point is now clear that we must stop and analyze when our common sense warns of questionable theories and semiclear pronouncements by the "experts." It is not an easy task, but we must take the responsibility or we will find ourselves very much like the computer in respect to "garbage in-garbage out." When we see computer garbage output, we correctly question the input. No less is required on the human level if we want to break the grip of the experts. This is particularly important when the experts for contradictory definitional traps or when they forget the courtesy of labeling their stories "fiction." Samuel Butler, at least, did not fail us in that respect.

Except that it gave his compelling story a good, old-fashioned midnight eeriness, the machines of Erewhon need not have been destroyed, after all.

1. John G. Kemeny, Man and the Computer (New York: Scribner, 1972), pp. 17 and 18.

2. Robert Efron, "Biology without Consciousness--Its Consequences," Perspectives in Biology and Medicine, vol. II, University of Chicago Press, 1968, pp. 9-35.

3. Ibid., p. 18.

4. See "The Stolen Concept" by Nathaniel Brandon in The Objectivist Newsletter, January 1963.

5. Efron, p.26.

6. H.O. Simon, "The Shape of Automation," Perspectives on the Computer Revolution (Englewood Cliffs, New Jersey: Prentice-Hall, 1970, pp. 161-67; repr. from Management and Corporations, 1985, Ashen and Bach, eds. (New York: McGraw-Hill Book Co., 1960).

7. Henry Block and Herbert Ginsberg, "The Psychology of Robots," Perspectives on the Computer Revolution (Englewood Cliffs, New Jersey: Prentice-Hall, 1970, pp. 246-55; repr. from Psychology Today, April 1968, pp. 50-55.

8. G. Rattray Taylor, "The Age of the Androids," Perspectives on the Computer Revolution (Englewood Cliff, New Jersey:. Prentice-Hall 1970), pp. 168-83; repr. from Encounter, London, November 1963; pp. 36-46.

9. Efron, p. 25.

10. "Introduction to Objectivist Epistemology," The Objectivist, vol. 5, no. 7, July 1966, pp. 5 and 6.

11. "Introduction to Objectivist Epistemology," The Objectivist, vol. 5, no.9, September 1966, p. 2.

Additional sources

Amosov, N. M. Modeling of Thinking and the Mind. London: Macmillan & Co., Ltd., 1967.

Anderson, Alan Ross, ed. Minds and Machines. Englewood Cliffs; New Jersey: Prentice-Hall, 1964.

Butler, Samuel. "The Destruction of the Machines of Erewhon," (adapted from Erewhon), Perspectives on the Computer Revolution. Englewood Cliffs, New Jersey: Prentice-Hall, 1970, pp. 161-67.

Fogel, Lawrence J., Alvin J. Owens, and Michael J. Walsh. Artificial Intelligence through Simulated Evolution. New York: John Wiley & Sons, 1966.

Jaki, Stanley L. Brain, Mind and Computers. New York: Herder & Herder, 1969.