from pages 84 to 88 of

    The New Yorker
    January 9, 1989

    by Brad Leithauser




No Loyalty to DNA

Qeer happenings are afoot in the tree of life. That's the message one gets, in any case, from Hans Moravec's "Mind Children: The Future of Robot and Human Intelligence" (Harvard $18.95). At a time when many books for the layman document the steady pruning of life's tree -- the dwindling of plant and animal species through environmental spoliation -- Moravec focusses on a prospective branching out: the emergence of new life-forms that soon will, he feels quite certain, "mature into entities as complex as ourselves." He is asking us, in effect, to transform our taxonomy at the roots. No longer would the complementary kingdoms of plants and animals represent the primary division; he foresees a still more fundamental bifurcation -- that between what might be called the empire of organic life and the empire of inorganic life. We are about to enter a "postbiological" world. Imminently -- perhaps within the lifetimes of our children -- robots of such advanced capabilities will emerge that "our DNA will find itself out of a job, having lost the evolutionary race to a new kind of competition."

As a genre, the popular science book boasts a number of distinctive traits, several of which are present in "Mind Children." Books about rapidly developing technology can often seem undeservedly cheerful, and at various points Moravec might be faulted for scanting the darker psychological and moral implications of the vision he conjures up. He brings to his subject both the natural optimism of the technician whose field is mushrooming before his eyes and the belief -- which seems to keep creeping in, despite his disavowals of it -- that technology will necessarily save us from technology. "Mind Children" also illustrates, as books of the genre commonly do, that the modern proliferation of research effectively makes each of us -- scientist and nonscientist alike -- a layman; as knowledge expands breathtakingly, in every direction, even the scientist must greet most discoveries with head-shaking incomprehension. Although one's primary reason for picking up a book like "Mind Children" may be to get a grip on a burgeoning new field, one is probably also hoping to throw off some of the dazzled numbness that comes of living in a technologically explosive age. Doubtless many readers who turn regularly to the best science popularizers -- to authors like Loren Eiseley and Stephen Jay Gould and Paul Colinvaux and Douglas Hofstadter -- do so to combat the numbness; the reader seeks the haphazard saving moment when an impossibly distant object or an unthinkably complex equation comes alive. Such sightings are likely to be unpredictable and personally idiosyncratic. I remember once discovering, in a mathematician's biography, that a hypothesis whose implications I did not understand had been modified for all integers less than 10101034, a figure beside which the number of elementary particles in the observable universe looks infinitesimally small. That so mind-boggling a figure had been of use and application in somebody's line of work seemed in itself an affirmation of the wonders of the universe.

Another characteristic of the genre is that the guilt one feels toward unread books diminishes on a steady, almost graphable basis. The reader who buys but fails to open, say, "The Tale of Genji" or "Njal's Saga" or Hobbes' "Leviathan" or Boswell's "Life of Johnson" introduces into his home a durable source of guilt; such books are classics, and they reproach us unremittingly as long as they remain unread. But the reader who buys the latest volume on artificial intelligence or quantum theory or paleontology or black holes knows that with each passing month its urgency will fade; within a few years, it will make no claim whatsoever, since by then it can safely be deemed out of date. One would be making a mistake, though, to let "Mind Children" recede unopened into a guiltless oblivion. It's a tonic book, thought-provoking on every page. And it reminds us that, in our accelerating, headlong era, the future presses so close upon us that those who ignore it inhabit not the present but the past.

Hans Moravec, who is director of the Mobile Robot Laboratory of Carnegie Mellon University, possesses a lucid, reassuringly commonsensical style and a flair for analogical simplification which together make the recondite seem approachable and the revolutionary plausible. Ever since completing his graduate training at the Stanford Artificial Intelligence Laboratory almost a decade ago, he has concentrated on robot locomotion and vision, and he devotes part of the initial chapter of "Mind Children" to a detailing of the difficulties inherent in any attempt to endow a robot with sight. The theoretically simple business of "hooking up" a computer to television equipment proves fiendishly complex in practice. The field of robotics is, in fact, full of unexpected reversals. Tasks that look elementary often prove formidable. In general, scientists have had a much easier time teaching a robot to perform the "higher" functions that formerly belonged solely to human beings (reading, proving theorems, diagnosing diseases) than the "lower" functions that animals have mastered (hearing, seeing, grasping objects). Improbably enough, a robot is more easily taught to play expert chess than to move the pieces.

As Moravec points out, there are evolutionary reasons for the higher being more accessible than the lower: "Encoded in the large, highly evolved sensory and motor portions of the human brain is a billion years of experience about the nature of the world and how to survive in it. The deliberate process we call reasoning is, I believe, the thinnest veneer of human thought, effective only because it is supported by this much older and much more powerful, though usually unconscious, sensorimotor knowledge." Given the irregularities and uncertainties of terrain outside the laboratory, freewheeling movement is not merely tricky but often hazardous for a robot. The monumentality of the programmer's task becomes evident when one considers that a truly flexible and autonomous robot would have to have enough of what researchers call "world knowledge" to translate all relevant physical conditions -- every shifting object, every stray obstacle -- into the strings of binary numbers which are its language of operation.

So thoroughgoing and convincing is Moravec on the subject of the complexities of robot movement that one comes away from his early pages feeling that the goal of autonomy is almost insurmountable -- and also feeling that in the face of such a sober-minded assessment one must treat even the most outlandish of his subsequent predictions with respect. In its less than two hundred pages of text, the book undertakes quite a journey. By its close, the reader has met robots that can go on risky vacations for vicariously adventurous human beings, "protein robots" so miniscule that they can assemble machinery molecule by molecule, even robots that can construct other robots in factories out in the asteroid belt.

Near the start, while discussing some of the ways in which a mechanical object could be programmed to behave like a human being, Moravec takes an intellectual sidestep. "The conditioning software I have in mind would receive two kinds of messages from anywhere within the robot, one telling of success, the other of trouble," he begins. "I'm going to call the success messages 'pleasure' and the danger messages 'pain.' Pain would tend to interrupt the activity in progress, while pleasure would increase its probability of continuing." But once he has dropped the quotation marks around "pleasure" and "pain" he treats the terms as though each was genuinely synonymous in its robotics and its human applications. He acknowledges no potential confusion when describing machines in emotive language: "Modules that recognize other conditions and sense pain or pleasure messages of appropriate strength would endow a robot with a unique character. A large, dangerous robot with a human-presence detector sending a pain signal would become shy of human beings and thus be less likely to cause injury." In short, he appears to accept as a given that the hypothesis that mind is merely a kind of machine -- one whose meditations and commands are ultimately duplicable by other, inorganic machinery -- and therefore finesses a question that lies at the core of current debate in the field of artificial intelligence: Is there any area of human activity which is obdurately, permanently inaccessible to machines? Moravec, who once remarked that he has "no loyalty to DNA," may be sound in assuming an underlying identicalness between the human mind and the machine (most experts in the field would probably agree with him), but he was unwise in choosing to pass over so fertile and significant a controversy. Readers who look elsewhere for a discussion of the subject -- perhaps to the contentious essays assembled in "The Artificial Intelligence Debate," a new collection edited by Stephen R, Graubard -- will likely encounter a tangle of human emotions, including skepticism, anger, foreboding, and indignation, that are all but missing in Moravec's expansive, self-assured projections.

But readers who are willing to go along with him, at least temporarily, on the issue of duplicability -- something which the patent joy he derives from speculation invites one to do -- will find that his arguments proceed with a sureness that verges on the inexorable. Attempting to place the modern computer in historical perspective, he ventures back a hundred years to examine Herman Hollerith' punchcard tabulator (a device that eventually became a sort of founding father of I.B.M.), and he concludes that since the beginning of the century "there has been a trillionfold increase in the amount of computation a dollar will buy." He estimates that in terms of computational power the largest of the present-day supercomputers "are a match for the 1-gram brain of a mouse," but that in time we may be able to build machines that operate at a million million million million million (1030) times the power of a human mind. What can one say in response to such a number? If duplicability is possible, is it not inevitable? And even if we assume that it is not possible how can we deny that machines of such unreckonable energies would not be capable of a rich and ranging inner life of their own?

Isn't it only a matter of time, Moravec asks, before we can transfer, or "download," our minds into computers? Copies could then be made of copies and stored in separate, secure places, not all of them on the earth -- a procedure that would virtually insure our immortality. He foresees a number of ways in which downloading might take place. A person could wear each day a miniaturized observational device, whose data, compiled over years and years, would serve as the memory bank of a new intellect. Or you might enter the hospital for brain surgery to be performed by a robot whose hands are microscopically precise and whose command of speech allows the two of you to proceed collaboratively. Since the brain registers no pain when it is subjected to incision, you could be fully conscious during the entire operation. Equipped with an encyclopedic understanding of human neural architecture, and proceeding millimetre by millimetre, the robot surgeon would develop a program that would model the behavior of a discrete layer of brain tissue. This program would produce signals equivalent to those flashing among the neurons in the area under scrutiny, and a series of cables would allow the robot to create "simulations," in which the program is substituted for the layer of brain tissue. The simulation process would be analogous to what's now available in sophisticated audio shops, where a customer can test and compare components at the push of a button and without breaking the flow of the music:

To further assure you of the simulation's correctness, you are given a pushbutton that allows you to momentarily "test drive" the simulation, to compare it with the functioning of the original tissue. When you press it, arrays of electrodes in the surgeon's hand are activated. By precise injections of current and electromagnetic pulses, the electrodes can override the normal signaling activity of nearby neurons. . . . As long as you press the button, a small part of your nervous system is being replaced by a computer simulation of itself. You press the button, release it, and press it again. You should experience no difference. As soon as you are satisfied, the simulation connection is established permanently. The brain tissue is now impotent -- it receives inputs and reacts as before but its output is ignored. Microscopic manipulators on the hand's surface excise the cells in this superfluous tissue and pass them to an aspirator, where they are drawn away. . . . Eventually your skull is empty, and the surgeon's hand rests deep in your brainstem. Though you have not lost consciousness, or even your train of thought, your mind has been removed from the brain and transferred to a machine.

A slower and seemingly less traumatic transfer might be achieved by installing in the corpus callosum -- the main cable that unites the brain hemispheres -- a microscopic monitor linked to a computer that would "eavesdrop" in order to make a model of your mental activities:

After a while it begins to insert its own messages into the flow, gradually insinuating itself into your thinking, endowing you with new knowledge and new skills. In time, as your original brain faded away with age, the computer would smoothly assume the lost functions. Ultimately your brain would die, and your mind would find itself entirely in the computer.

Any such event would compel a further modification in our taxonomy. The distinction between organic and inorganic life -- and, indeed, all the subdistinctions by which a species is fitted into a unique biological niche -- would dissolve. Although Moravec has disappointingly little to say about religion, his ultimate vision incarnates widespread theological convictions about the "oneness" of all life:

Mind transferral need not be limited to human beings. Earth has other species with large brains, from dolphins, whose nervous systems are as large and complex as our own, to elephants . . . and perhaps giant squid, whose brains may range up to twenty times as big as ours. Just what kind of minds and cultures these animals possess is still a matter of controversy, but their evolutionary history is as long as ours, and there is surely much unique and hard-won information encoded genetically in their brain structures and their memories. The brain-to-computer transferral methods that work for humans should work as well for these large-brained animals, allowing their thoughts, skills, and motivations to be woven into our cultural tapestry. Slightly different methods, that focus more on genetics and physical makeup than on mental life, should allow the information contained in other living things with small or no nervous systems to be popped into the data banks. The simplest organisms might contribute little more than the information in their DNA. In this way our future selves will be able to benefit from and build on what the earth's biosphere has learned during its multibillion-year history. And this knowledge may be more secure if it is preserved in databanks spreading through the universe. In the present scheme of things, on our small and fragile earth, genes and ideas are often lost when the conditions that gave rise to them change.
Our speculation ends in a supercivilization, the synthesis of all solar-system life, constantly improving and extending itself, spreading outward from the sun, converting nonlife into mind.

Actually, Moravec might plausibly contend that conventional theological debate is hardly germane to his argument. If he is mistaken about human duplicability, most of his projections at once reveal themselves as pipe dreams that connect only remotely and hypothetically with religious issues. And if, on the other hand, he is correct in supposing that human minds will be transferred into or otherwise fused with machines, it seems likely that traditional religious questions -- and traditional religions themselves -- will either melt away or suffer wholesale metamorphosis. Debates about Heaven or Hell -- to take but one example -- would hold little relevance for an immortal creature. One wishes, however, that he had accorded greater space to psychological considerations. Many people experience an instinctive unease at the incursions of the mechanical -- a feeling concisely summed up by Emerson a century ago: "Machinery is aggressive." And although such people might reconcile themselves in time to the notion of a man/machine cohabitation -- most of us, in the course of modern life, have already grown used to hearing computers speak to us -- the conviction that there is something innately "special" about human beings would surely die hard, and at great cost.

The modern scientist and his offering are often likened to Mary Shelley's Victor Frankenstein and his monster, but the nineteenth-century novel that Moravec most vividly evokes is Stevenson's "Dr. Jekyll and Mr. Hyde." In a document found at his death, Dr. Henry Jekyll explained how, troubled by the "polar twins" that dwelt in his "agonised womb of consciousness," he conceived the prospect of a sweet divorce" "If each, I told myself, could be housed in separate identities, life would be relieved of all that was unbearable." These words are echoed in Moravec's prologue: "In the present condition we are uncomfortable halfbreeds, part biology, part culture, with many of our biological traits out of step with the invention of our minds. . . . It is easy to imagine human thought freed from bondage to a mortal body." "Mind Children" makes light of the possibility that a deathless human being is not a human being at all -- that the condition of mortality so informs our lives as to render them unrecognizable without it. It may be (to pose a paradox of a sort that Moravec himself might relish) that on the day when man makes himself immortal he makes himself extinct. The future that Moravec sees is certainly one in which "timeless" truths -- the eternal verities of the poet -- are set on their ear. Algernon Swinburne observed that "all men born are mortal but not man." Moravec everts this dictum: in his world, the individual would become deathless, but man in the aggregate -- that species whose hopes and expectations have been framed in the phrase "threescore years and ten" -- would vanish.

One has to wonder how the art that we have safeguarded throughout the centuries would survive the transformation. Whether one is listening to Hamlet speculate on the bourn from which none return or contemplating a ukiyo-e print of rice harvesters or reading "Gilgamesh," the appreciation of any work of art generally requires us to cross a gulf -- both geographical and temporal -- on the bridge of our kindred uncertainty and helplessness in the face of death. Any art that might be fabricated in Moravec's new world would be composed, in effect, in a new language. Without question, it would be extraordinary. But surely much of what we now revere would suffer in translation.

Readers who are curious about Hans Moravec, and long for greater personal detail than is provided in "Mind Children," will find him in fine form, witty and engaging and professorially eccentric (a favorite snack is Cheerios topped with bananas and chocolate milk), in Grant Fjermedal's "The Tomorrow Makers," whose subtitle is "A Brave New World of Living-Brain Machines." In pursuit of his book, Fjermedal spent a number of clearly quite exhilarating months in this country and Japan, drifting from one artificial-intelligence center to another and meeting a wide range of individuals who, for all their singularities, seem to share a penchant for working all night and sleeping catch as catch can on the morrow. And who share, as well, a daily, ingrained perception that the intertwined evolution of man and machine -- of which downloading might be regarded as the apotheosis -- is steadily speeding us toward an alien world.

Even readers who view the prospect of downloading with confident disbelief or squeamish distaste will appreciate the poignance that at present suffuses the field. A number of researchers have come to believe that they were born just a little too early -- that the immortality toward which their collective efforts are reaching will not be attained soon enough. For them, death is an ailment that will not be cured in their lifetimes. Fjermedal quotes one researcher as saying, "Everyone would like to be immortal. I don't think the time is quite right. But it's close. It isn't very long from now. I'm afraid, unfortunately, that I'm the last generation to die.

According to Moravec, however, such pronouncements are unduly pessimistic and final. He speculates that by dint of a gathering mastery of a range of disciplines, including history, genetics, anthropology, and computer simulation, much that has disappeared may prove retrievable. Having effaced so many familiar categories, the future will eventually soften even the distinction between itself and the past. Time will turn ductile:

Now, imagine an immense simulator (I imagine it made out of a superdense neutron star) that can model the whole surface of the earth on an atomic scale and can run time forward and back and produce different plausible outcomes by making different random choices at key points in its calculation. Because of the great detail, this simulator models living things, including humans, in their full complexity. According to the pattern-identity position, such simulated people would be as real as you or me, though imprisoned in the simulator.
We could join them through a magic-glasses interface, which connects to a "puppet" deep inside the simulation and allows us to experience the puppet's sensory environment and to naturally control its actions. More radically, we could "download" our minds directly into a body in the simulation and "upload" back into the real world when our mission is accomplished. Alternatively, we could bring people out of the simulation by reversing the process -- linking their minds to an outside robot body, or uploading them directly into it. In all cases we would have the opportunity to recreate the past and to interact with it in a real and direct fashion.

When that day comes, we will have a choice about which pasts we want to consign to the past and which we will summon to accompany us into the future. In the meantime, though, the reader is left to wonder what the human cost would be of never losing anything.

-- Brad Leithauser