- Shopping Bag ( 0 items )
Around the world, scientists and engineers are participating in a high-stakes race to build the first intelligent robot. Many robots already exist -- automobile factories are full of them. But the new generation of robots will be something else: smart machines that act like living creatures. When they are brought into existence, science fiction will have become fact.What will happen then? With our prosthetic limbs, titanium hips, and artificial eyes, we are already beginning to resemble our machines. Equally important, our machines are beginning to resemble us. Robots already walk, talk, and dance; they can react to our facial expressions and obey verbal commands.
When they take the next step and become fully autonomous, what will they do? Will we be partners or rivals? Could we meld into a single species -- Robo sapiens?In Robo sapiens, Peter Menzel and Faith
D'Aluisio present the next generation of intelligent robots and their makers. Accompanying brilliant photographs of more than one hundred robots is an account of the little-known, yet vitally important scientific competition to build an autonomous robot. Containing extensive interviews with robotics pioneers, anecdotal "field notes" with behind-the-scenes information, and easy-to-understand technical data about the machines, Robo sapiens is a field guide to our mechanical future.
The MIT Press
But as they take on these new subjects, science writers must assume a certain responsibility. Though technology has always had its dark side, today the darkness threatens as never before; with the incalculable opportunities of the PC and the Internet come not only sore wrists and fat behinds, but also fraud and vandalism on a colossal scale; with emission-free nuclear power comes ageless nuclear waste; with industrial prosperity comes global warming; and on and on. It is no longer enough to be an advocate of a new technology, or even an explicator of it. Writers must think through the full spectrum of future possibilities, no matter how happy or dire.
Wherein lies another, almost opposite, hazard: The pace of technology is so fast that authors can be made the fool before their books reach store shelves. Think back to less than 20 years ago, when you unpacked your first XT computer. If you had been asked to guess how it would change the world, how preposterously wrong would you have been? Do you bother to guess about the future of science any longer?
In Robo Sapiens, an overview of the present and future of robotics, photojournalist Peter Menzel and documentarian Faith D'Aluisio face both challenges by letting the inventors of various robots speak of their ambitions, expectations and fears. The excellent photography and design - Menzel is the creator of the photo-essay Material World - brings the robot, and the dialogues, vividly to life.
Robots are already superior to humans in many ways. They are more accurate than people, require no lights, work 24 hours a day, and they never get sick, go to the bathroom or form unions. As one inventor says in the book, robots are saints.
Today's great spark-spewing industrial robots are only the beginning, though. The Grail is the intelligent, autonomous robot. All of the inventors here assume that robots eventually will be able to assess unforeseen situations and learn from their experience. They will be able to care for themselves on distant planets and assume tasks of profound difficulty on earth. Though roboticists differ on the way such machines will come into being - and on exactly what intelligence is - the hope is that this research will also provide new insights into the human mind (and what can go wrong with it).
But intelligence is also the rub, as the authors make clear. Many scientists see robots far surpassing human intelligence within the first half of this century. Some think robots will also develop consciousness - which will impel them to replicate in huge numbers, presenting a grave danger to our species.
This raises a basic question about the nature of human intelligence: How is it possible that these scientists devote their every living breath to the development of what could be the agents of our demise? We might applaud the scientists' ingenuity, their imagination - but their intelligence? Or perhaps we should take this as an important clue to what they will impart to robots.
In any case, the scientists in Robo Sapiens are divided over what kind of emotional states their creations might be capable of, if any. Some say that since robots will be born into a society created by people, that social consciousness will naturally integrate itself into the machines' make-up. Others say it is impossible for us to foresee the emotional composition of robots because of the abyss between their nature and ours. Still others say that the sheer intellectual capability of the machines will make their provenance almost irrelevant. Instead, the robots may well become indignant at their subservient roles and dispense with us altogether.
As robots evolve toward the human, humans, with our corneal implants, artificial limbs and, someday, brain chips, are also assuming robotic characteristics. The authors coined the phrase "Robo sapiens," meaning both "a hybrid species of human and robot with vastly superior intelligence to purely biological mankind," and "the dominant species in the solar system of Earth" by the end of the century. Their hope is that as robots become more socialized and people become more technologically enhanced, the two will peacefully meet.
It is an interesting dream, but one that is difficult to share. We humans will certainly have an increasing need for robots, just as we have rapidly developed an addiction to computer chips. But by most accounts in this book, robots will have a diminishing need for us. They have been brought into the world specifically for toil; when they free themselves of it, what would be the source of any altruistic impulse toward humans?
Some scientists answer - or avoid - the question by claiming that robots will be such fast studies they will not only develop their own culture, but in the process they will subsume ours as well; thus they will remain, so to speak, humane. But knowledge does not imply influence. Studies of prairie dog colonies, for example, have not much changed our approach to urban planning. The engines of human evolution have been reproductive and self-preservatory instincts and their concomitants, many of them unsavory and bloody; what could possibly compel intelligent robots to assume the complex of contradictory characteristics that have been the outcome (so far) of that long and painful process?
Our culture is a several-thousand-year-long attempt to come to terms with our human frailties. Robots will never write the equivalent of the Bible or a Shakespearean play; they would never feel the impulse. Robot culture will be based on robot history and robot struggles (if robots ever come to struggle) - which will be very different.
Consider the billions of people, the millions of years and the innumerable interactions and accidents that have made us what we are; there is no algorithm that could reproduce or simulate our species. There will always be a profound gap between our species and the robot, whatever our common features.
Even if we share the authors' hope that robots will retain warm memories of their early years in our labs, that is hardly a secure basis for an indefinite happy cohabitation. And what if by accident some of the nastier human frailties - a propensity for genocide, say - make their way into the robots' code? How much more difficult might they be, and how great their resentment?
The authors tell us that the word "robot" was coined in a 1920 play R.U.R, which ends in a robot uprising that wipes out the human race. In the April 2000 issue of Wired magazine, Bill Joy, cofounder and chief scientist of Sun Microsystems, wrote a much-discussed article titled "Why the Future Doesn't Need Us," which warned that commercial forces are compelling us to develop systems that will bring calamity. If we are not convinced by R.U.R's visionary playwright, Karel Capek, it is more difficult to ignore Joy, who not only has helped forge our technological lives, but whose warnings run contrary to his vast interests.
Menzel and D'Aluisio conclude that "with good luck, we might lose some of the poverty, fear and desperation that has always been the human lot. With bad luck, we may eventually destroy ourselves." In other words, we don't know what the hell is going to happen.
After reading some of the more frightening testimony in this book, one hopes that we are wise enough to leave as little to luck as possible. Still, the authors should be applauded for their refusal to be false and foolish prophets.
Robot: n (Czech, from robota, compulsory labor) 1. A machine that looks like a human being and performs various complex acts (as walking and talking) of a human being. 2. A mechanism guided by automatic controls. [Merriam-Webster Collegiate Dictionary, 1999]
Robo sapiens: n (English, from robot, a mechanism guided by automatic controls; and Latin, from Homo sapiens, mankind) 1. A hybrid species of human and robot with intelligence vastly superior to that of purely biological mankind; began to emerge in the twenty-first century. 2. The dominant species in the solar system of Earth. [Microsoft Universal Dictionary, 2099]
Before Faith and I began this book I would have attributed the term Robo sapiens to a science-fiction writer. I would have been amused, but would have scoffed especially hard at those modifiers "vastly superior" and "dominant" in its (admittedly. hypothetical) definition. I have been skeptical about technology's undelivered promises ever since I began a career as a photojournalist, more than two decades ago. More specifically, I am skeptical not about technology per se but about the way societies misuse and misunderstand it. Nuclear power, for example, is a wonderful technology that would have had a prominent place in any wise, far-seeing, and incorruptible society. In the future, I hope to encounter such a society.
Pessimism about society's potential to misuse technology is nothing new. Czech writer Karel Capek coined the term "robot" in a play he wrote in 1920 called R.U.R.[first performed in 1923]. The play's dark plot revolves around a factory—Rossum's Universal Robots, the R.U.R. of the title—that populates the world with artificial slaves, meant to relieve humans of the drudgery of work. Built in ever-increasing numbers and with expanding intelligence, they soon outnumber their human masters, and then they are used as soldiers. Eventually, a robot revolt wipes out the human race. It's interesting that the person who invented the modern concept of robots predicted that they would destroy us all.
A year ago, when our friend Thomas Borchert, technology and science editor at Stern magazine in Germany, asked me to document advances in robotics and artificial intelligence around the world, I was eager to seek out the planet's most advanced machines and their makers (he didn't mention any threat to the human race). My interest—and my skepticism—increased when I learned that roboticists were predicting that human intelligence would soon be surpassed by machine intelligence—if not in the next decade, then in the decade after.
How much was hype? Was there really a revolution in robotics, as the roboticists claimed? I knew that robotics had not lived up to the dreams of its pioneers in the 1960s and 1970s. Today, years after robots were supposed to be on the streets and in our homes, the machines are largely unseen. The only robots widely in use are computer-driven slaves bolted to factory floors and laboratory benches. Knowing this, I did not expect to be intellectually kidnapped by a robot. But then it happened, in Japan.
We were shooting a private demo of the Honda [P3]—a robot that looks like an astronaut in a white spacesuit—at the car company's top-secret development center outside Tokyo. (Yes, the Japanese automaker had quietly spent millions to build a bipedal robot.) Watched proudly by a host of attendant technicians, the robot walked down a path on its own two feet, opened a door, stepped through it, closed the door behind itself, then gracefully climbed a flight of stairs. Was there a little man inside? No. Just electric motors and electronics. We knew that the Honda humanoid was not autonomous—that every step was programmed—but we were still amazed that a machine could maneuver like a human.
At that time, we were nearly halfway through a month-long assignment. We had seen a number of very advanced robots, but most worked haltingly, and some only after interminable delays to load software or troubleshoot a fragile connection. No single machine had yet captured our imagination enough to make us believe the glowing predictions of robotic evolution—in fact, many machines were quite primitive beasts. But when we saw the Honda humanoid's jaunty demo, we realized that with enough time and money, layers of complexity could be added to any of the other nonhuman-looking robots as well—upgrading them from clumsy class into the realm of smooth operators.
The Honda robot is not alone. As we learned more and saw more, we realized that twenty years after their initial predictions, the futurists might at last be proven right. Machine intelligence is here, in infant form. And now it's learning to walk and talk.
After completing an additional ten-day robot shoot in Europe in the spring, we decided to continue the grand tour of robotics labs around the world, visiting and revisiting labs and talking to the people who want to build our mechanical future. We wanted to know where the robots were going. More than that, we wanted to know where we—that is, humankind—were going. We were amazed by what we found.
We discovered that the dawn of our postbioiogical future may arrive sooner than we imagine. Consider technological progress to date. It took billions of years for primitive life forms to evolve into mammals, millions of years for the descendents of those first mammals to lumber into the Stone Age, thousands of years for humans to advance from stone to steel—and only sixty-six years for people to soar from Kitty Hawk to a stroll on the moon. We accelerated from the Wright stuff to the right stuff so quickly that the question inevitably arises: how long—or how short—a time will it be to the next step, Robo sapiens?
One part of the answer is that we are already part way there—indeed, we began evolving in that direction in the last century. With our artificial hips, prosthetic knees, false teeth, heating aids, pacemakers, breast and penile implants, we are part cyborg today. (Well, not me personally, you understand.) That list of mechanical parts doesn't include transplanted organs, skin grafts, and plastic surgery. Or the new mechanical hearts that are under development today. Or the cochlear and retinal implants that will be here tomorrow. There may be general resistance to implanting chips in people's brains, but when a bio-chip is developed to easily enhance memory or linguistic skills or mathematical abilities—how long will people just say no? Kevin Warwick (see page 29) plans to address this question by implanting a chip in his own body next year. It could be true: the next step in human evolution could indeed be from man to machine.
Another part of this puzzle is provided by the silicon chip, which has rammed our technological future into overdrive. Computers, nearly ubiquitous now, soon will be. Every year, they get smaller, faster, cheaper, more powerful. According to what is known as Moore's Law, computer chips will become faster and more powerful by a factor of two every year or so. This exponential increase shows no signs of abating, which means that chips will get faster and faster even as they get smaller and smaller. Contrast this with the evolution of the human brain and many experts conclude that machine intelligence will inevitably surpass human intelligence—the only question, is what will happen when it does.
These are the unknowns in the equation that make the future exciting, and possibly a bit scary—they blur the lines between science and philosophy. Since we don't understand the basis for human consciousness yet, how can we create it in machines? Or will mechanical minds start up by themselves when they reach a certain level of complexity? (Is there critical thought, like critical mass?) And if machine intelligence does jump-start itself, will we be able to turn it off if we don't like the results? If an artificial intelligence is composed of human-compiled facts, does this mean it will be kindly disposed to humans? Or will it be ethnocentric (or should I say mechanocentric)? Could it threaten us?
But this assumes that all the evolution will be on the mechanical side. What if we could start from early childhood with the complete storehouse of knowledge from our predecessors rather than having to reinstall it each time in our organic-hard-drive brains? Herein lies the lure of robotic silicon intelligence—bio-chips (brain chips, one might call them). Electronic memory can be accessed a million times faster than human synapses—it does not have to sleep, and can be downloaded to others (similar machines, or people with compatible chips). It can be stored, compressed, sorted—even spindled and mutilated. (No need to get angry, greedy, or jealous.) Of course this downloaded information would only be data, and data is not knowledge, let alone wisdom. But in a postbioiogical world, how long would it be before we learned how to take the next step and download wisdom? (Could it be a program?)
Now factor in the realm of the very small, where nanotech and bio-chips promise things unheard of (in fact, undreamed of). Micro Air Vehicles, on page 156, are already here—robotic flies are on the way. Kris Pister's smart dust (see page 26) may seem like science fiction, but for how long? Sir Arthur C. Clarke wrote, "Any sufficiently advanced technology is indistinguishable from magic." Today, the technological magic is more than accepted, it is expected. We believe that our dreams are not just dreams, they are sneak previews.
In one such preview of primordial mechano-motion, we watched Ariel, an eight-legged crab robot, sidle along the shore of a wooded pond in Massachusetts (see page 100). Probably the most complex self-contained machine capable of underwater ambling, Ariel is still very limited in what it can do. When I mentioned this to Ed Williams, the robot wrangler technician who was demonstrating the machine, he Wrightly countered, "What could an airplane do in 1920? Get you killed in it, that's about all."
The most exciting part of our journey was bearing witness to the first halting steps—spasmodic evolutionary spurts—of robotics at the beginning of the new millennium. To be sure. there were glitches along the way, and there will be more. At the MIT Leg Lab, we saw a robotic Troodon dinosaur (see page 114) twitch its way up to a standing position, ready to walk forth, only to be crippled by a software glitch before it took a single step. Mimicking the incredibly complex biological systems that took millions of years to evolve—and to survive and function in a narrow niche—is a daunting challenge. Shortcutting this process with high-tech tools to analyze animal locomotion, Robert J. Full's biology lab at UC Berkeley (see page 90) provides data that roboticists at Stanford, Michigan, and McGill use to build robots that translate these movements into mechanical systems guided by the genius of biological evolution.
The ultimate quest, the Grail of many roboticists today, is to build a humanoid robot. The Honda P3 that so astounded me is but one of many attempts to reach this goal. The quest is moving along at a steady pace that will accelerate with advances in materials, computing power, and electromechanical interfaces. But why build humanoids? Walking upright on two legs is as difficult as flying. Is there any other living creature more unstable than a human in an ambulatory position? Many roboticists say the reason to build such an unwieldy being is psychological—they believe that robots will be more easily accepted by humans if they are built in the humans' own image. Others think it folly; they admit the motive for making a machine in one's own image may be psychological, but see it more as a fascination with playing God. The well-known MIT roboticist Rod Brooks (see page 58 frankly explained that he jumped from building insect robots to Cog, a humanoid, because he didn't want to spend a lifetime crawling along the robotic evolutionary path when he sensed he might be able to have a run at something approaching an android.
Shigeo Hirose, one of Japan's most respected roboticists, thinks the humanoid shape may not be the best idea, in engineering terms, but he argues that any robot engineered to be intelligent could be engineered to be moral. Robots could be saints, he told us. We could build them to be unselfish, because they don't have to fight for their biological existence (see page 89). They can download their artificial intelligence to another machine, thereby continuing their "life." They don't have to be like the rampaging machines in R.U.R. Nice robot, smart robot, saintly robot.
In spirit, it seems we are more ready for robots than they are for us. Despite our fears of Frankenstein's monster or Hollywood's Terminator, if something robotic can make our life better, we embrace it. We already accept and expect robotic systems to help us: power steering, cruise control, and GPS systems in our cars; autopilot and IFR landing systems in airplanes. We trust these machines with our lives—we have built them with multiple levels of checks and safety features. Our comfort level with robots is rising, too. Many swimming-pool owners have them cleaning the bottom of the pool every day, and several companies are developing robot vacuum cleaners for the home. Robots already work in many factories—there is no resistance, really. In the General Motors plant that we visited (see page 190), robots do all the dirty work, and workers welcome the relief. They say they don't fear job loss from robots—if a robot takes over their position on the factory floor, they get to be transferred to less physically demanding work within a more productive factory.
Robotic progress today is poised to take off like the personal computer did a decade ago. The shortcomings that have kept robot research away from millions of small inventors and researchers—no universally accepted operating control system and a lack of standardized parts—should soon be complaints of the past. A standard operating control system and new lightweight, compliant materials (see page 98) on shape-deposition manufacturing), coupled with more powerful, energy-efficient actuators, could bring robot design into the wide, swift mainstream of everyday science.
Today's robots are more than factory workers; they are explorers, space laborers, surgeons, maids, actors, pets—the list gets longer every day. We should expect to be surprised, because our imagination will create many, many more roles. Our mechanical destiny is not to be denied, and the questions arising from the creation of these creatures are ones that will shape the future of humanity, in whatever form it eventually assumes.
This book is not meant to be Genesis-like, detailing each mechanical iteration. Instead, I hope that my camera and Faith's tape recorder have produced a field guide to our mechanical future that will make the passage less frightening. Should we assume the worst: that Robo sapiens will eventually run amok like the machines in R.U.R.? In Act III, during the robots' siege of the factory, the optimistically naïve Helena, one of the last surviving humans, tragically laments to Radius, a robot leader, that his intelligence should engender understanding, not conflict. In the 1920s play, it didn't. In tomorrow's real world, we hope it will. Radius can be a saint—if we understand the process that created him and the things in ourselves that drove us to do it.
HELENA. Doctor Gall gave you a larger brain than the rest, larger than ours, the largest in the world. You are not like the other Robots, Radius. You understand me perfectly.
RADIUS. I don't want any master. I know everything for myself.
— R.U.R. by Karel Capek, 1920
Electric Dreams: What the future may hold
There are great and wondrous robots in our future, say Those Who Know. Robots will assist the elderly and infirm into and out of wheelchairs and beds, be conversant in several languages, intuit despair, watch over babies, and provide a sympathetic ear to the lonely. Smartly appointed robotic vacuum cleaners, robotic cars, robotic maids, robotic cleanup squads, and robotic personal assistants will lead to greater efficiency and safety in the world, working where humans can't or won't and providing more free time for their masters. All but human, ubiquitous, they will be woven invisibly into the fabric of our lives.
There are terrible times in store for the human race, say Those Who Know. Robots will begin as elder-care assistants and vacuum cleaners, but they won't stay there. They will take what we teach them and learn to want more. They will make themselves ever smarter and stronger, until finally, discovering that they are better than we are at everything we do, they will refuse to take our orders. Far from becoming indispensable components of our lives, they will find our lives ever more unnecessary to them. If they don't end up ignoring us, they'll eliminate us.
Who are Those Who Know? The robot pundits. The prognosticators of our mechanical future. The digital soothsayers. Academic and corporate researchers, for the most part, they study such exotic domains as artificial intelligence, cyberneurology, and biomimetics. Often at odds with one another but never unsure of their auguries, they claim to know the future of the human race, and to know that it will involve robots. Robots, they all agree, will transform the future. The problem is: they differ on the details. Like whether robots will serve us—or we will serve them.
For more than a year, Peter Menzel and I explored robotics laboratories in Europe, Asia, and America, looking at projects in development and speaking with researchers. Over time, we concluded that these pundits are at least partly right. Clearly, the robots are coming. Although the machines we saw were often barely functional, they were gaining in capacity. The discipline of robotics—a quirky union involving the fields of artificial intelligence, computer science, mechanical engineering, psychology, anatomy, and half a dozen others—is perhaps moving faster than even the researchers know.
The discipline is advancing so rapidly, in fact, that some roboticists have begun questioning the direction in which their work is heading. Not every roboticist we encountered felt inclined to speculate on the future, of course. Like all branches of science and engineering, robotics is full of researchers who try to focus entirely on their work in the present. For the most part, these people take pains to distinguish themselves from the robot pundits. But, there is something so magical about the creation of artificial living creatures—mechanical entities with lifelike behavior—that even the soberest of the researchers find themselves wondering what lies ahead for their creations, and for humankind. The robot revolution will happen, whether we like it or not. From now on, we and the robots are in this together. All the more reason, thought Peter and I to try to figure out what's coming down the pike.
Even before 1920, when the Czech writer Karel Capek invented the word "robot" in his play R.U.R., the world had begun to embrace the concept of artificial workers with humanlike capacities. Japanese inventors and artisans had created tea-serving automata, or karakuri, as early as the seventeenth century. Automata—mechanical contrivances designed to act as if they were under their own power—were familiar diversions in eighteenth-century European courts. And by the nineteenth century, automatons were creeping into science-based fiction and folklore in the form of golems, clockwork men, and Frankenstein's monster.
One of the first attempts to match reality to fantasy occurred after the Second World War, when aerospace engineer Joseph Engelberger (see page 186) conceived of machines that could perform repetitive tasks tirelessly and more accurately than their human counterparts, and brought robots to the factory floor. What began primarily as a field for mechanical engineers grew to include engineers of all stripes. Their machines were in essence puppets—expensive, beautifully designed devices that were controlled completely by the strings of their instructions. They couldn't think, create, or react; they simply performed their tasks, moving with the reflexive precision of a pendulum.
Robotics did not acquire its present scope until the arrival of modern computers, which inevitably brought with it the idea of stuffing some sort of brain into the robot. In the 1940s, English mathematician Alan Turing laid the groundwork for artificial intelligence, famously theorizing that a machine is intelligent when there is no discernible difference between its conversation and that of an intelligent person. At the Massachusetts Institute of Technology, trail-blazing computer scientists John McCarthy and Marvin Minsky founded what in the late 1950s became the world's first laboratory devoted to artificial intelligence. One goal was to use artificial intelligence to advance the study of human intelligence. Another was, of course, to build robots.
In those optimistic days, computer power was growing so fast that true artificial intelligence seemed to be just around the corner. It wasn't. Some AI researchers were able to program computers to behave intelligently in certain narrow functions, but they were never able to create a machine that could speak or read or solve unexpected puzzles. Dumping a dictionary into a computer—their approach, roughly speaking—didn't produce a book. Minsky tried to build an "intelligent" arm that could stack blocks atop one another. It never worked. Beset by difficulty, artificial intelligence as an active field of research declined in the 1980s.
Partly to blame is the inherent difficulty in creating a simulacrum of a phenomenon that nobody understands. If the nature of intelligence and consciousness remains a subject for speculation to this day, how can scientists manufacture it in artificial form? If we needed to know how our brain works—where thoughts come from and how memory works—in order to use it, all of us would be in a heap of trouble. Even the scientists who have charged themselves with the task of discovering the secrets of the brain, and are shrinking the pile of unanswered questions and conundrums at a faster and faster pace, are still working at the level of the educated guess. And if they get to the bottom of the pile of riddles, will they have the answers they seek? If the magic of a single thought is made not of illusion but allusion, will its genesis be any more possible to discern?
Many roboticists today avoid the quagmires of AI by building what are in essence dumb machines, without a hint of consciousness but programmed cleverly enough to perform complex tasks—searching for breaks in a municipal sewer system or pumping gas at a service station (see page 195). Using cheap, scavenged electronic equipment, Mark Tilden, a researcher at Los Alamos National Laboratory (see page 117), can make small, insectlike machines that walk over irregular terrain with as much aplomb as if they had eyes to see where they were going and minds to adjust their step.
Does a robot need to have much of a brain? It depends on what you want it to do. Usually, machines we saw in the laboratories were bolted to a bench; some could maneuver around a finite pristine space under close supervision. But if robots are to inhabit the world's kitchens, as Tilden puts it, "You probably want the robot to know that it shouldn't suck up the cat kibble." (Let alone the cat.) When a robot operates in a human environment, the programming becomes more difficult. Safety concerns, mobility issues, and space requirements suddenly emerge. If more than one task is involved, the difficulty increases exponentially. Even a harmless, just-for-fun device like the Sony AIBO robot dog (see page 224) is subject to these constraints—that's why it moves slowly, is soft-edged, and costs twenty-five hundred dollars.
Sophisticated programming alone is not enough to make a machine seem lifelike. Curiously, what is required often is not great intelligence or startling skills, but randomness. Predictable behavior is computer-like; randomness is human. To accommodate this perception, Sony has added touches of spontaneity to the AIBO: in no particular or repeating order, at any given time, it might "play" with its ball or lie down and wave its legs in the air. But the notion of randomness in a machine dismays Engelberger, the robotics pioneer. "Maybe it is more fun and interesting if it screws up now and then and it does something a little different," he told me. "But I don't want that. I want the thing to be utterly reliable." ("Robots like the AIBO have a different purpose." I observe. "Yeah," he says. "To horse around.") It's understandable that Engelberger would feel that way—he designs industrial and health-service-oriented robots, which must be undeviatingly dependable. But even in the more relaxed atmosphere of the home, unexpected robotic behavior could be less than charming. A vacuum-cleaning robot that spontaneously broke out into a little dance might be funny the first few times, but a householder's patience might wear thin if that expensive robot vacuum cleaner were dancing instead of working, and wearing out its custom wheels and the Carpet in the process. Thus even the most well-programmed, occasionally random automata will not be fit companions for people. If robots are to fulfill their creators' dreams, they will have to be given truly intelligent brains, which means that even if they want to, researchers will no longer be able to avoid wrestling with the riddles of AI.
Today, there are two main approaches to creating a clever machine: weak and strong artificial intelligence. Weak AI is the argument that a machine can simulate the behavior of human cognition, but it can't actually experience mental states itself. Even though such machines would be able to pass Turing's test of intelligence, they would still be little more than extra-complex clock radios. Proponents of strong AI argue, by contrast, that machines are capable of cognitive mental states—that it is possible to build a self-aware machine with real emotions and consciousness.
Strong AI greatly distresses some philosophers, including John Searle of the University of California at Berkeley. If a computer can have cognitive mental states, he points out, then a human mind would have to be simply a computer program implemented in the brain. To Searle, this contention is absurd; consciousness is a first-person, subjective phenomenon that no mechanical computation, no matter how sophisticated, can produce.
Daniel C. Dennett, a philosopher at Tufts University, takes the opposite view. Consciousness, he says, is at its core algorithmic—that is, the brain has a series of rules for dealing with incoming sensory data, and the summed execution of these rules in the lower strata of the mind generates consciousness in the mind's upper strata. If Searle is right, robotics faces inherent limitations—we will never be able to build a truly intelligent machine. Dennett offers a more hopeful picture, at least for roboticists. But there is a chance that both might be wrong. Robots may need a brain to do everything their advocates imagine, but they may not need a brain that is humanlike. It is possible that circuitry utterly unlike the human brain could make robots behave in ways that seem indistinguishable from the workings of intelligent consciousness.
The construction of such machines—a race of intelligent aliens made right here on Earth—would be an ironic triumph for AI advocates. Proof that artificial intelligence is possible, these robots would still be incomprehensible; they would provide little or no insight into the human mind. Worse, they would plunge humankind into an immediate ethical quandary. If a robot has a brain equal to that of a human being, should it also have the legal and political rights of a human being?
The question of robotic rights could be less of a concern for the robots Kris Pister envisions. An electrical engineer at the University of California at Berkeley, Pister has the courage to think small. He and his colleagues are working toward a robotic future that is dominated by what he calls "Smart Dust"—autonomous robots no more than the size of a gnat (one cubic millimeter). Although equipped with sensors and ways to move around, each tiny machine will be relatively simple. But when thousands of them are combined into a network that permeates the environment, the totality will be capable of extraordinary things. Sprinkled on a child's clothing, the minute devices could monitor his or her location and state, sounding the alarm if the child climbed out of the crib or onto the back of a chair, or seemed to be choking. Instead of using a keyboard, people could stick Smart Dust on their fingers and run their computers by making finger motions. By scattering Smart Dust equipped with moisture and acidity sensors on food, homeowners could monitor the freshness of the items in their refrigerators. Mechanical motes on workers' clothing could signal office heating and cooling systems to raise or lower the temperature. In effect, the dust would turn the entire environment into an invisible robot, constantly on the alert to do the human's bidding.
The first exemplars of Pister's version of a microelectrical and mechanical system are expected in 2001, but they will only be able to perform simple tasks like monitoring the temperature around them. Pister's students are developing minuscule legs and wings that will make the dust more mobile. Greater communication power and more complex sensors will come later. But the goal is always the same: individual components that can work together in clouds, literal clouds. Pister believes that in the future millions of these tiny robots will be constantly floating through the air. "When you fly across the country," he told me, "pilots are always saving, `The last guy who passed through here 20 minutes ago told us that we are going to hit some turbulence.' If these guys were passing out Smart Dust from the back of the plane, they could query it as they went through to see exactly where the turbulence was."
Because inventions that don't yet exist can't be photographed, Peter set up a photo illustration to conceptualize Pister's ideas. The researcher stood before a projection of a pop-up micro Fresnel lens—the kind of lens used in lighthouses—and blew glitter into the light. I asked what he saw when he looked at the finished picture (above). "When I see the Fresnel lens," he said, "mentally I see a beam of light come out that's modulated on and off and transmitting information. Or maybe it's drawing a picture on a screen or on the back of my retina. Then, the sparkling glitter coming out of my hands—when I see that, I think it's the little lasers that the Smart Dust particles use to communicate with each other. On our time scale, we see these things rushing out like a mad jumble. On their time scale, it's a very slow, beautiful, underwater dance. They're talking to one another as they're flying through the air, telling each other what they are seeing—`Who's where?' and, `Let's set up our network.'"
Smart Dust has obvious military, applications and in fact the research is backed by the US military's Defense Advanced Research Projects Agency (DARPA). Pister admits that his ideas could have a downside: permeating the world with invisibly small robots will make it possible to watch anyone, anywhere, at any time, but thinks the benefits of this technology will far outweigh the negative.
Without the Pentagon, US robotics in its present stage of evolution could very well collapse; in our research for this book, it was rare to find a top American researcher who was not sponsored at least in part by DARPA or the US Navy's Office of Naval Research (ONR), though other entities such as the National Institute of Health also provide some funding. Even foreign nationals like German researcher Frank Kirchner of the prestigious German National Research Center (see page 113) has US military funding for a robot project with Northeastern University's Joe Ayers (see page 110).
Is military backing for robotics a touchy subject with researchers? Not particularly. As aerospace pioneer Paul MacCready (see page 156) dryly observes, "It's been my experience that whatever we've done for the military has done more good for the nonmilitary." Generally, researchers go where the money is. DARPA's program managers solicit, vet, and fund proposals submitted by universities, private research laboratories, and corporations; they run conferences and meetings for prospective and current awardees just as professional organizations do for their members—an especially valuable resource in robotics, say researchers, because of the field's multidisciplinary nature. And the military supports nonsecret projects with humanitarian uses, including robots that search for and disarm land mines (see pages 101 and 111). "We'll play in anyone's back-yard," says Georgia Institute of Technology's Ron Arkin of his lab's Pentagon support (see page 152). "But we'll only do open, unclassified research. That's where I draw the line, personally. Everything we do we can publish."
There is an unavoidable bottom line. "You have to have somebody who's a receiver for the technology," says Arkin, lamenting the lack of support in long-term robotics research by US corporations, which generally want a more immediate return on investment. "The question is," he says, "what industries are going to be ready to nurture that technology for the five or ten years it might take to build that product?"
Because its US-designed constitution is intended to forestall military adventurism, Japan does little defense research. Instead, its private corporations and the government are heavily involved in robotics research for commercial applications. Japan's equivalent of DARPA's robot program is its nationwide initiative to create a humanoid robot industry. The country's historic legacy in robotics has centered on mekatoronikusu, or mechatronics, a fusion of mechanical and electronic engineering, and it expects that its future will be chock full of autonomous walking robots of the humanoid persuasion. The country is in the midst of a ten-year national project to build just such a robot, which, says the University of Tokyo's Hirochika Inoue, "will walk in an unstructured environment and perform complex tasks." Japan's hope is that, among other applications, such a robot will fill the need for elder care in their aging population and situate the country at the forefront of a new humanoid robotics industry.
Inoue, a respected roboticist and a principal member of the project's governing committee, calls the humanoid project Japan's grand challenge. Though some researchers, such as Shigeo Hirose of the Tokyo Institute of Technology (see page 89), question such a robot's viability, Inoue is fiercely supportive of the project, which he sees as a technological lifeline for a country left behind when the software industry took off.
Posted December 22, 2003
They can climb stairs, juggle balls, open a door, smile engagingly, hear and see, swing like a monkey, crawl like a crab and swim like a fish. Who? Why the robots, of course. This startling picture book explores the amazing scope of robot capabilities. The photographs of the robots and their creators provide a unique picture of the dawn of these intelligent machines. The narratives are brief and to the point, explaining just enough but always remaining as support for the pictures. As I thumbed through this book, it became clear that the development of humanlike robots will come one project at a time, not by a thunderous breakthrough from a single genius working in a dark lab. Definitely buy this book; there¿s nothing quite like it.Was this review helpful? Yes NoThank you for your feedback. Report this reviewThank you, this review has been flagged.