The Cabinet of Wisdom I — Metaphysics and Mechanical Men
The Cabinet of Wisdom
Part I Metaphysics and Mechanical Men
I trust the patient reader will allow a lengthy discourse on the topic of the possibility or otherwise of artificial intelligence and mechanical men, for the discourse involves some of the most contentious and confusing concepts in philosophy: physics and metaphysics, mind and body, man and nature, necessity and volition, truth, validity, virtue. None of these concepts has failed to generate volumes of dispute across generations of history.
This essay will be published over days to come in parts, of which this first part deals only with preliminary matters, clarifying what the topic statement means, what metaphysical arguments are, and what it might mean to say a machine is a man.
Alas, such a discourse necessarily involves concepts which the student of modern philosophy is spectacularly ill-equipped to investigate, if not crippled.
I submit that Artificial Intelligence is absolutely impossible.
This seems an overly bold declaration, considering the current state of cybernetic technology and medical neurology.
Currently, medical neurology seems on the brink of unlocking the deterministic brain-mechanisms of intelligence thought processes. Currently, cybernetic machines are currently able to react meaningfully to spoken conversation, and respond in what seems an intelligent fashion, hence cybernetic technology seems on the brink of passing the so-called Turing Test, which seems to be a test for machine intelligence.
All this that seems so is not so. In truth, the statement is not bold, but mere common sense; neural science will never have anything meaningful to say about intelligence; cybernetic machines programmed to mimic intelligence are not intelligent, and the Turing Test is not a test.
Philosophically, modern, secularism defines ideas as manmade hence subjective, recognizes none but empirical proofs, and acknowledges no form of causation aside from mechanical causes. Modern philosophy likewise admits of no reality aside from the material, that is, namely, aggregation of matter and energy reducible to the fundamental seven measurable magnitudes of mass, length, duration, temperature, current, candlepower, and moles of substance.
Given the axioms of secularism, the conclusion that artificial intelligence is possible is inevitable. Since human brain matter exists, and since nothing but brain matter can determine the thought-processes we subjectively call intelligent, and since the determination must be mechanical, as there is no other type of causation, a manufactured artifact which can mimic the mechanical processes of a living brain within sufficient tolerance to determine intelligent thought processes would and must produce intelligent thought processes.
Moreover, being artificial, the specific content of the thought would likewise be determined by the engineer. The intelligence, if truly artificial, could be engineered never to think certain forbidden thoughts, or programmed never to escape certain prime directives which must be obeyed.
Hence artificial intelligence is theoretically possible. Moreover, in the absence of empirical proof that it cannot ever be done (and there is no other kind of proof), no conviction that it is absolutely impossible is sound.
Were the secular axioms true, or even coherent, the conclusion would be valid. If the brain really were a natural machine that produces intelligence as mechanically as oxygen and iron produce rust then an artificial machine could do likewise by copying the production process.
This essay has two tasks: the first is to show that the secular axioms are incoherent; the second is to prove that artificial intelligence, properly so called, is unconditionally impossible.
We begin by clarifying impossibility.
Impossibility What
Let none misunderstand: the statement that artificial intelligence is impossible is not merely a lament that something theoretically possible, such as a manned mission to Pluto, is not feasible given our current level of technology and available resources; this statement is not merely a comment that something hypothetically possible, such as a faster-than-light mission to Proxima, is not possible given our current understanding of physics.
A manned mission to Pluto is impossible only in the current year, given our current resources and political priorities, that is to say, it is contingently impossible. Other years may be different.
Hyperdrive flight to Proxima is impossible if our current understanding of the universe is basically correct, but an honest scientist will always confess that further refinement, even radical changes, in the standard model of physics are hypothetically possible. Other universes may be different.
This statement that Artificial Intelligence is absolutely impossible, contrariwise, is absolute. It avers that Artificial Intelligence is a contradiction in terms.
Artificial Intelligence cannot now and never will exist in this universe because it cannot exist, even hypothetically, in any universe, for the same reason that a four-sided triangle or a causeless effect cannot: namely, the two concepts “artificial” and “intelligence” upon examination, are found to be mutually contradictory.
The statement is more akin to saying Time Travel is impossible, which is not a question of technology nor of physics, but of metaphysics. Time Travel means changing the past. Time Travel is metaphysically impossible because, were it not impossible, a time traveler could establish an inescapable paradox by traveling backwards to before his departure to prevent it, in which case he never travels back, never prevents the travel, and so is free to travel back and prevent the travel.
Now, to be sure, there are speculations, both in science fiction and in science, that indeterminate forking paths in time or an iron fatalism might allow an unobservant time traveler, at least for a time, to believe he has the power to change the past. Neither time travel limited by fatalism, nor liberated by indeterminism, is time travel properly so called. In an indeterminate universe, past events lead not to present results, but to a fork, where multiple results obtain. In such a universe there is no past properly so called, or, at least, not a past that necessarily leads to the present. In a fatalistic universe, events are fixed. In such a universe there is no change. In either case, there is no changing the past, hence no time travel.
And likewise for the various other imaginative attempts to evade the paradox: as a matter of logic, if time travel is possible, time is not time.
But this is a discussion for another day: my point is that there are arguments claiming certain concepts are paradoxes. Paradoxes are impossible is not because of limits imposed by current technology nor by the current understanding of physics. Such arguments do not deal with the practical, the theoretical nor the hypothetical. Such arguments deal with the rational or metaphysical, and is found convincing or unconvincing according to the standards applied to purely logically or purely metaphysical argument.
The paradox of the absolute impossibility of Artificial Intellects is just such a metaphysical argument. As with other arguments of this kind, it is proposed as a matter of logic, that if artificial intelligence is possible, intelligence is not intelligence.
Disqualifications
But wait! Surely the candid reader, if he is familiar with this author’s body of work, knows that many story starring artificial intelligences, mechanical men, and time travelers have flowed abundantly from his pen, not to mention tales involving faster than light drive, and a manned mission to Pluto. How can John C. Wright, self-anointed grandmaster of science fiction, who makes his living by spinning malarky to trusting, wide-eyed, myriads of legions of loyal fans, fans who are ear-starved for high-quality yet well crafted masterworks of speculative fiction, cast aspersions on the several core concepts of the genre, especially artificial intelligence? He has earned filthy lucre telling trusting, wide-eyed fans stories assuming such things as artificial intelligence and time travel and faster-than-light drive are possible! Dare he now scoff?
My only answer is that I have also written ghost stories, and everyone knows ghosts are real. Other than that, everything in fiction is fiction. The purpose of science fiction is to explore the realistic implications of such unreal conceits. The art of science fiction is to lull the disbelief of the reader into somnolence; this is done by disguising the unrealistic or unrealized elements of the make-believe to make those elements seem believable.
The purpose of philosophy is to find the truth through reason, which is the opposite mental effort. A healthy mind can do both, each in its proper role.
Indeed, many of the ills afflicting the modern intellectual landscape spring from an inability of modern intellectuals to distinguish between science fiction speculations, whether uttered by accredited scientists of otherwise, and philosophical reasoning, where accredited scientists have no particular accreditation nor basic competence.
Metaphysical Argument What
A metaphysical argument concerns fundamental first concepts or axioms, that is, non-conditional hence necessary statements. If a metaphysical argument is true, it is true necessarily, true as an a priori statement, that is, it is true regardless of what our senses tell us about the behavior of matter and energy in the natural world.
Metaphysics is a non-empirical science, and, despite what you may have heard, its conclusions are truth, and constitute real knowledge, not merely word games nor tautologies.
Make no mistake: metaphysical arguments are about axioms, but such arguments are not themselves axiomatic.
Axiomatic arguments are “take it or leave it” as when Euclid asks us to accept, without proof, Playfair’s axiom that parallel lines neither converge nor diverge. This axiom is foundational for Euclidian geometry, but, due to Goedelian limits on formal systems such as geometry, it cannot be proven with a proof taken from Euclid. So, the reader must accept Playfair’s Axiom before he accepts any conclusions of Euclidian geometry about parallelograms, congruence, or trigonometry. But he need not: there are logically coherent non-Euclidian formal systems.
But a properly constructed metaphysical argument is not axiomatic. We humans find ourselves in a physical realm of time and space, matter and energy, where the five senses inform us of the various motions and changes of sensible objects through time and space. And we find ourselves in a mental and moral realm of reason, a realm of concepts, and of symbols to represent them. In this mental realm, our inner faculties can and do distinguish true from false, logical from illogical, concrete from abstract, right from wrong, fair from foul.
The first realm is nature and the second is reason.
By the nature of these realms, we humans could neither discover, discuss, nor examine their secrets unless certain axioms are always and eternally true and valid and right. Without a law of cause and effect, for example, no sense impression could convey any meaningful information about matter and energy in time and space to our minds. Likewise, without a moral law that applies to all conditions, a golden rule, there could be no meaningful thought hence no discussion either of blameworthiness nor praiseworthiness of human action, nor of duty, nor any concept of law, authority, dignity, liberty.
For this reason, a metaphysical argument is like opening a wrapped gift. You already own whatever is in the box beneath the bright paper and gay ribbons, but you may or may not know it. Likewise, here, you use the necessarily metaphysical axioms needed before thought and discussion about nature and reason are possible. And, as a part of the human condition, thought on nature and on reason cannot be avoided, even by the most diligently unphilosophical of minds.
At best, one can adopt without inspection the assumptions and conclusions of other men concerning the topics of nature and reason, and mimic the opinions of your parents and teachers, and they from theirs, but, in that case, you are adopting their metaphysical axioms also. There was a first teacher or first parent, a sage or philosopher, prophet or nomothete, who first put a name to the hitherto unnamed axiom on which his thinking (and yours) was necessarily based.
A metaphysical argument is successful when it makes explicit an axiom that inescapably must be granted as logically prior to the acceptance of conditional conclusions you have, by the mere act of thinking about nature or reason, already granted. A successful metaphysical argument shows that an axiom thought to be conditional is actually inescapable.
A metaphysical argument fails when it proposes as inescapable an axiom to which a logical alternative exists. As when an axiom thought to be inescapable is actually conditional. For most human history, Playfair’s axiom was regarded as inescapable, but with the explicit discovery of coherent geometric systems by Reimann or Lobachevsky that eschew this axiom, there is no sound metaphysical argument for the exclusive truth of Euclid, only for Euclid’s validity.
Here, the argument against Artificial Intelligence is metaphysical. The argument must prove that no artifact can have intelligence, even if, to Alan Turing, or some other unobservant outside observer, it may appear to have.
This argument is not empirical. Whatever neurologists discover about the human brain, or engineers contrive in electronic brains, makes no difference one way or the other.
Terminology
For the sake of clarity, and because the term “Strong A.I.” may be misleading, and because the phrase “Artificial Intelligence” is currently used for a technology whereby a machine is designed to have its reactions adjusted over repeated trials (sometimes incorrectly called “learning”), I will use the less ambiguous term “Mechanical Man” or “Mechanical Mind” and pursue the argument that any object that is mechanical cannot by definition be a man.
If he has a mind, he cannot be a machine; if it is a machine, it cannot have a mind. This is the proposition to be proved.
As with all arguments, we begin with definitions. Sadly, nearly all argumentation is people talking past each other, using the same words to refer to concepts that are not the same. Alas, the beginnings of any such discussion are the most tedious, and where the most mischief is done. Let us take care.
Mannequin and Man
A mechanical man, if by this we meet a wax dummy with motorized joints who can be made to gesture and blink and move wax lips and tongue while a gramophone hidden in its chest makes soundwaves deceptively similar to words radiate from its wax mouth, to have a man make an mechanical man is certainly possible.
I have consulted with Walt Disney’s Meet Mr. Lincoln at the 1964 World’s Fair, and seen with my own eyes that my own eyes can be easily deceived by an imaginative copy.
Many a radical secularist would play along with the deception, and allow himself to be convinced that this is the real Mr. Lincoln, provided only that a sufficiently unobservant observer has reason to be convinced.
But were this the real Mr. Lincoln, he has recovered from his assassination, and has been raised by necromancy. For this reason, I do not confuse the animatronic puppet with the man it represents. In 1964, when I was three years old, I just so happened to fail to see the exit wound left by the bullet of the assassin. Also, in the years since, for theological reasons, I have come not to believe necromancy can truly raise the dead to anything but a hideous and unholy mockery of life. For these reasons, I did not then and do not now mistake the moving mannequin for the real Mr. Lincoln, who sleeps with his ancestors.
Be that as it may, unlike a secularist, we can and must distinguish copy from reality, image from object, word from concept.
Specifically, we can see a distinction between a man and mannequin.
Men have each a mind or spirit, an animating principle, a personality, an internal reality, a self awareness, a point of view.
Mannequins, such as Mr. Disney’s Imagineers can contrive, do not. They are artifacts made by intelligent beings deliberately made to symbolize, represent, or mimic the outward words and actions of intelligent beings.
The basic distinction between mannequin and man is the difference between map and territory, symbol and object, significator and signified, and, in sum, the difference between the word and that which the word represents, between the thought itself and object of thought.
A singer can sing a song. A phonograph can record in wax a soundtrack or groove-pattern representing the pressure waves produced by the original the song, which grooves a needle can pick up and play into a speaker, which can produce pressure waves mimicking the same song. With a sufficiently faithful speaker system, the human ear can detect no difference between live and recording. But a phonograph cannot sing a song, only replay what a living singer recorded.
Singing is a volitional act, and phonographs, no matter how sophisticated they may one day become, are not now nor ever can be endowed with volition.
It is a machine. When correctly engineered and undamaged, a machine does what it is designed to do, no more, no less. A machine malfunction may be an example of Murphy’s Law, but it is not an example of free will.
Likewise, a looking glass, when undamaged, undistorted, unstained, can only reflect the image of an object placed before it. It cannot produce new images voluntarily, as an act of imagination, because the faculty of imagination is not one that exists in looking glasses, only the power to reflect images.
To be continued in Part II
About John C Wright
John C. Wright is a practicing philosopher, a retired attorney, newspaperman, and newspaper editor, and a published author of science fiction. Once a Houyhnhnm, he was expelled from the august ranks of purely rational beings when he fell in love; but retains an honorary title.
