Mind Children: The Future of Human and Robot Intelligence

Hans Moravec

Harvard University Press

Cambridge, MA

1988

Reviewed By: Erin Rhodes

Date: August 1998

Imagine this: It is the year 2048. You are an aging CEO of one of the more lucrative robot factories established in the quarries on Mars. On the eve of your death, you decide to invest your savings in a trendy but promising new operation which guarantees the transference of your mind to a machine. It will be your stab at immortality; or, at least it will prevent the wanton loss of knowledge you have accumulated over your lifetime to die with you. You have always wanted to visit the moons of Jupiter, and this new machine, in the style, color, and material of your choice, with your brain inside, will be able to travel there, unimpeded by the severe environmental conditions of space, acquiring new experiences and memories along the way. Later, these memories can be downloaded into the mind of your granddaughter, who will not only share your memories but those of the machine’s experiences as well. Although your mortal body will have long passed away, the rumors of your death will be greatly exaggerated; in fact, because your mind will be constantly amassing new experiences, your permanent death will be highly unlikely.

Soon after, scan your brain "While-U-Wait" chain stores crop up around the country, ready to deliver a cheap, quick fix to the masses when they absolutely need transmigration to a new, technologically advanced entity. 100 million served. Never mind if the new repository you’ve chosen for your mind to live in is fatally destroyed; no matter, you have been conveniently backed up on the future equivalent of magnetic tape, ready to be downloaded into a new computer, resulting in a new you, minus, of course, the experiences you enjoyed during your last reincarnation. Indeed, mind-to-machine transferal methods would not be limited to human beings; genetic information encoded in the brain structures of other living organisms, developed over billions of years, can be transferred to the data banks, so that we can build on what we have learned over the course of our hard-won evolutionary history.

But what happens to memory and experience when it becomes a commodity? Can the mind really be freed from the physicality of the brain&endash;and of the body? Wouldn’t multiple versions or copies of ourselves, and the prospect of immortality, cheapen the uniqueness of being human? Are consciousness, emotion, and intelligence particular only to humans (and perhaps other living things), or can they be instilled into a machine? Are these even relevant questions to be asking?

Transmigration of the human mind into a machine is only one of many fantastical predictions Hans Moravec, a roboticist at Carnegie-Mellon University, makes in his book Mind Children. Moravec speculates that the end product of transmigration, where our "postbiological" evolution will ultimately lead us, is "a supercivilization, the synthesis of all solar system life, constantly improving and extending itself, spreading outward from the sun, converting nonlife into mind." (p. 116)

Moravec opens his book with the following provocation: "I believe that robots with human intelligence will be common within fifty years." His intent in writing this book may be to justify the necessity of his own research, as he contends that the future embodiment of machines that think will be the intelligent robot. A more subtle intent may be to gap the divide between the traditional artificial intelligence (AI) community and the cybernetics researchers. What lies in the divide is Moravec’s approach to the intelligent machine: the imitation of the evolution of animal minds by tackling the obstacles of perception and mobility first, since sensorimotor knowledge forms the bedrock upon which human intelligence developed. Moravec contends that this "bottom up" approach will meet traditional AI research, which has focused solely on copying the conscious mental processes of humans performing tasks, halfway, and unite it with the cybernetic approach to building models of the brain at the neural level. By following a model of the biological path to intelligence, coupled with the exponential growth in computer power, the evolution of intelligent machines will take faster strides (10 billion times faster!) than the processes of Darwinian evolution. (p.17) Within the next century, machines will mature into entities as complex as ourselves&endash;our artificial progeny ("mind children") will evolve so swiftly that we will lose the evolutionary race to a new kind of competition. (p.2)

Moravec breaks down his work into six distinct chapters, each of which address the forces (biological, industrial/economical, cultural, and cosmic) that lead us inevitably to the intelligent robot and beyond, and, when relevant, entwines the history of AI and computers as they relate to each force. The chapters form a progression from the construction of a utilitarian robot to machines with intelligence and consciousness (who inherit the universe while we as a species are grandfathered out of existence) to the emergence of freely evolving digital wildlife&endash;mutations and viruses analogous to what we know in the biological world&endash;that will alter the course of evolution much as they do today. The stage is set in the first chapter, "Mind in Motion," which explores the development of a "robot for the masses" to help out around the house: an autonomous, functional, mobile robot, able to navigate, see, grasp objects, and learn. Moravec suggests that intelligence can be best understood by constructing robots which can explore their environment and live autonomously in the real world. Sensory data will provide the robot with concrete and abstract representations of the world, which can be stored by the robot and examined later as a basis for learning, or modified by new experiences. This will be the beginning of the self-aware robot. Competitive development of "world simulator" software installed in robots will drive research efforts in the robotics industry of the early 21^st century. These simulators will provide the robot with hypothetical scenarios which will prepare it for future situations and opportunities, and establish the robot’s capacity to dream, imagine, and remember. From here it is not a long stretch to the emergence of robots with feelings, flashes of intuition, and intellect.

Although Moravec criticizes traditional AI research for setting its sights on the mechanization of pure reasoning and ignoring the importance of sensorimotor skills, he certainly takes a "strong AI" approach in his predictions. He is a staunch proponent of Moore’s law, which guarantees that we will have the computer power and capacity to approximate the human intellect within forty years, based on Moravec’s impressive calculations in the second chapter, "Powering Up." The main premise of his book favors the "strong AI" mindset: if you look at robotics research today, we are very much headed in the direction of utilitarian robots; in fact, we already employ industrial robots for many tasks. Once we have the necessary hardware and software power in place, humans and robots/computers will engage in a symbiotic relationship, exploiting the benefits of each: computers will take advantage of the mobility and sensory capabilities of humans, and humans will take advantage of the superior memory, calculating power, and communicating range of computers. (p. 85) This partnership may take the form wonderful wearable hardware such as "magic glasses" for vision, "magic gloves" for touch, and a "magic wardrobe" for movement. As the computer assimilates this sensory information, Moravec suggests that the distinction between the natural and artificial partner will no longer be evident. Eventually, machines could be made of a mix of inorganic and organic substances, when appropriate, the ultimate extreme being the transference of our minds into machines, as discussed earlier. Machines will become more competent with each successive generation, devising ways to improve upon preceding ones, eventually leaving us behind as we become more and more unnecessary. Whether or not we transplant our minds, Moravec asserts, we will have no choice but to evolve in the direction he outlines in this book, as our survival will depend on how well we adjust to changing conditions and keep in step with our new evolving competitors. (p.121)

It is quite easy to feel ambivalent about Moravec’s work. His musings strike you as far-out and radical at first, but his arguments and future projections are very persuasive and based on scientific truths embedded in the history and industry of computing power, robotics research, and Darwinian evolution. Moravec is obviously not afraid to stick his neck out; the possibility of such a future deserves consideration, no matter how distasteful or exciting it may come across to you. Moravec has an extremely inventive mind, and his imaginative scenarios, set off from the rest of the text with italics, are enjoyable to read for their entertainment value and dreamlike qualities alone. His creative applications of future technologies will certainly appeal to many science fiction readers. However, he is not a scientist to be taken lightly. The book is intended for the layperson, and serves as an excellent introduction to the AI and robotics fields, as he weaves the historical context of these disciplines together with his own experiences at Carnegie Mellon, and his predictions of the future&endash;as long as you keep in mind this is his version of the future.

Perhaps what is most disturbing about Mind Children is that Moravec comes across as entirely enthusiastic about our demise as the superintelligent qualities of our artificial progeny usurp our place at the top of the intellect chain. I’m not so sure I really like our mind children, or feel a sense of pride in their evolutionary superiority. There is no reason to believe that human equivalence is the upper limit and that machines will not outclass us one day in the distant future. Still, it’s no fun looking forward to becoming evolutionarily defunct.

Check out Hans Moravec’s website (http://www.frc.ri.cmu.edu/~hpm)

Additional links to Robotics resources (http://www-robotics.cs.umass.edu/robotics.html)

A la carte: Marvin Minsky’s "Will Robots Inherit the Earth?" (http://www.ai.mit.edu/people/minsky/papers/sciam.inherit.html)