Man evolves to machine
Ray Kurzweil's latest book imagines a future where machines are the next step in evolution
Ray Kurzweil, the well-known artifical intelligence entrepreneur, lays out a roadmap to the future in which computers augment and eventually replace human brain functions. His future sci-fi scenarios are by turns outlandish, scary, and fascinating, but they're all highly intelligent and fun to read. (3,200 words)
Ray Kurzweil invents technology-based products that push the envelope of technology's ability to interact with the natural world. In the late 1970s, he developed an OCR engine and a speech synthesizer as part of the Kurzweil Reading Machine, a device that scanned text and read it out loud for vision-impaired people. He subsequently licensed the OCR technology to Xerox, which still uses it to this day in its scanners.
Some of his products were eventually eclipsed by others -- Fairlight and Synclavier synthesizers; Caere OCR; Dragon Systems and IBM voice recognition -- but he always took the boldest steps of innovation. He brings the same thinking-outside-the-box flair to The Age of Spiritual Machines, his third book and the successor to The Age of Intelligent Machines from 1987. There have been many writings about how technology is changing the nature of our society, some of which (like Michael Dertouzos's What Will Be) have been reviewed in this column. But this book looks further into the future -- and into the past -- than any of them.
The Age of Spiritual Machines is a treat for the intellect. It reminds me a bit of Douglas Hofstadter's classic Gödel, Escher, Bach: An Eternal Golden Braid, (GEB) which is a fixture on the bookshelves of nerds everywhere (including this one). If you ever read GEB, you may remember the first few chapters, which attempted to cover a mid-level college course's worth of mathematical logic. Unless you were a real math genius, you either slogged through these abstruse chapters as best you could, hoping that eventually you would understand, or you gave up. If you persevered, you were rewarded with Hofstadter's rich, tapestry-like meditations on intelligence, aesthetics, and mathematics.
The speed of life
Similarly, Kurzweil's book starts off with some incredibly dense theories about time, evolution, and chaos, which set up his subsequent, somewhat bombastic rationalizations about how we are nearing a transformative moment in the history of the universe. These, in turn, pave the way for the meat of the book, where he discusses actual and predicted advances in technology that support his theories.
I won't risk misrepresenting or trivializing Kurzweil's theories by trying to summarize them here. However, one of his more interesting philosophical points is that time is not linear; instead, it's a record of salient events in the progression of the process that takes the universe from chaos to order -- a small subset of which is the process of evolution of life on earth. Examples of salient events in the process include the big bang, the creation of matter, the creation of life, the creation of life forms that can use tools, and so on. Kurzweil asserts that the "speed of time" should be measured by measuring the time betweeen salient events. If we look at salient events in evolution, we see that they are getting closer and closer together in time (coincidentally, Kurzweil's name means "a short while" in German), meaning that time is speeding up. He goes further in claiming that this speed-up effect is exponential: that evolution is getting faster and faster all the time. (Stay with me; the payoff will come in a minute.) After establishing that evolution is speeding up exponentially, Kurzweil then goes on to say that technology -- moreover, computing technology -- is an inevitable result of evolution. First, evolution produced entities that could perpetuate themselves (i.e., life forms). Then came species that had various skills. One of the most important of these was the ability to make and use tools. Next comes technology, which Kurzweil defines as the ability to make tools and record a process for making them, which others can then use and improve on. Computing technology turns out to be inevitable because it is an ideal means of recording, replicating, and modifying processes.
In other words, Kurzweil is actually saying that the inexorable acceleration of computing technology is part of the evolution of life on earth. He makes logical arguments that human technological creations will eventually take over the process of technological invention -- and hence the evolutionary process -- from humans. That is, machines created by humans will be the next stage in evolution, leaving humans as the inferior species. Furthermore, he says that this next stage will occur within the next century or two.
will be the next stage in evolution
making humans the inferior species.
That's a big statement, especially considering that most of us think of evolution as something that takes hundreds of thousands of years to work. How does he justify it? First of all, he points out that evolution has been speeding up ever since life began. Dinosaurs evolved from the primordial ooze over a period of hundreds of millions of years; humans evolved from primates in mere millions. Our own species, homo sapiens sapiens, emerged only 90,000 years ago.
Kurzweil uses a neat trick to tie this accelerating pace to the growth in technology: he extends Moore's Law backward in time. Moore's Law is the well-known prediction of Gordon Moore, a founder of Intel, that computing power will double every 18 months (recently revised downward to 9 months). Moore's result is based on the increasing number of transistors that can be put into an integrated circuit of a given physical size.
Kurzweil extrapolates Moore's Law by going back before integrated circuits, and even transistors, to Babbage's Analytical Engine, which appeared around the turn of the twentieth century. He figures out the computing power of computational devices throughout the century, from Babbage's to the present, and plots them on a logarithmic graph. Surprise: the results fit a curve that is close to a straight line, thereby establishing that technology has grown exponentially -- moreover, at the same exponential rate -- for at least a hundred years. (I wonder if Kurzweil could extrapolate this model even further into the past by considering arithmetic calculation devices, going back to the abacus.)
These theories about man, machines, and evolution take up the first part of the The Age of Spiritual Machines. If you read lots of science fiction, then you probably take these kinds of assertions in stride. If you don't, then you might be tempted -- as I was, at first -- to dismiss him as a crackpot. But even if you think he's a crackpot, you'll be tempted to read on.
If you do, then you'll get to the middle part of the book, which is a bit more down to earth and considerably easier to digest. It's a fascinating explanation of technologies that, when brought to fruition, will bring in the age of machine-enhanced man and eventually a blurring of boundaries between man and machine. Most of these technologies result from three basic ideas: digital replication of the human brain, nanotechnology, and quantum computing.
Ray Kurzweil is an unrepentant supporter of artificial intelligence (AI). He is one of those who laments the demise of this term: it went into disrepute in the 1980s when many AI companies failed to deliver impressive products and went bankrupt. As is typical of such people, he complains that AI is doomed to mean "whatever isn't possible with computers today," meaning that when a problem (like, say, computers beating human chess champions) gets solved, it's no longer considered AI.
He is also a staunch cyberneticist. He feels that one of the most important things we can do to advance computing technology is to determine the makeup of the human brain and use that as a starting point for superior computing technology. It should be possible within 20 years, he claims, to perform a nondestructive scan of the human brain -- using some MRI-like technology -- in enough detail to model it accurately enough to reproduce it. A project already under way in Japan has this goal.
to model the human brain accurately
enough to reproduce it.
There's a lot more to reproducing the human brain than just creating a map of it; you have to figure out how it gets trained, how it interfaces with its "peripherals," such as the nervous system and sense organs, and so on. Nevertheless, Kurzweil holds to the reductionist theory that if it's possible to model the human brain using computer circuitry, then because computer circuitry is several orders of magnitude faster than synapses and neurons, it should be possible to build a version of the human brain that runs a lot faster and has much more capacity, but is otherwise identical. Kurzweil suggests that within 30 years, computers will exist that are orders of magnitude more powerful than the human brain.
As an intermediate step, he suggests the idea of brain implants that augment existing brain functions. Already, crude cochlear implants of this type are available to help the hearing-impaired. He says that the first implants will fulfill this type of function for people who are sense-impaired or subject to memory loss. Inevitably, however, people will want such implants to improve their senses or their brain functions, and then we will cross the Rubicon into the area of computer-enhanced humans.
Kurzweil admits that there are tricky ethical issues surrounding these types of technologies. If you download your brain into a computing device, then are there two of you? If you decide to use the computer brain and turn off the biological brain, have you committed suicide? These are vexing questions. But they are predicated on the assumption that qualities like "soul" and "personality" -- and the most troublesome of all, "consciousness" -- can be copied, as long as we understand the brain to a sufficient level of detail. Other scientific philosophers, like Roger Penrose of Oxford University, disagree.
As a further example of Kurzweil's views, he cites researchers who claim to have found the part of the brain that is stimulated if the brain's owner has a religious (or, as in the book's title, spiritual) experience. In other words, it should be possible to create spiritual experiences on demand by stimulating that center of the brain. To say the least, this would not be likely to go over big with certain political elements in this country.
Nanotechnology is another important component in Kurzweil's view of the future. Nanotechnology is the idea of building machines on an atomic scale that can combine themselves into matter. If such machines can possess the intelligence of powerful computers ("nanobots"), then you get all sorts of interesting results, such as Star Trek-style food replication, tiny medical diagnostic devices that actually live in your bloodstream, and objects of all types that can morph into other objects at will. Add the ability of these devices to replicate themselves, and you get something pretty scary.
Blame it on the ether
Quantum computing, Kurzweil's third basis for the future of technology, is even wilder -- and more tenuous in its existence -- than nanobots. Quantum computing builds on the theory in quantum mechanics that subatomic particles have spins and positions that are ambiguous until they are observed by a conscious entity. For example: if you turn on a flashlight, it emits photons. If you shine it on a mirror, then the photons that strike the mirror from a certain direction bounce off it in all possible ways. But if you look at the mirror, you cause the photons to decohere, a phenomenon in which every path cancels out except the one "true" path, which is the one that leads in the opposite direction and same angle to the mirror.
Research into quantum computing says that this is a simple example of how quantum effects can be used to solve problems: in this case, the problem of figuring out the angle at which the photons bounce off the mirror. If you construct a quantum computer to solve a difficult problem -- say, factoring a 1000-bit encryption key or solving a 1000-node traveling salesman problem -- then it will simultaneously consider all answers and, when observed, decohere so that only the correct one shows. As a result, a quantum computer is an infinitely parallel computer, one that renders trivial all problems previously thought to be too difficult for computers to solve. (To be more precise for the computer science majors among us: a quantum computer of N bits is capable of infinite parallelism on all problems with answers expressable in N bits or less; therefore, it's capable of solving NP-Hard problems of up to N bits in one time unit instead of 2 to the Nth power.)
You can read Kurzweil's explanation of quantum computing over and over again, as many times as you like, and you may still scratch your head. It's like that well-known cartoon (reproduced in this book) that shows one scientist talking to another in front of a blackboard. On the blackboard are some arcane-looking mathematical symbols, followed by "Then a miracle occurs," followed by more math. The caption reads, "I think you should be more explicit here in step two." In quantum computing, it's still a mystery to me how the quantum decoherence process can be made to produce the desired result. Nevertheless, Kurzweil claims that two scientists have already produced a quantum computer that can add 1 + 1 and are working on building one that can factor small numbers.
Quantum computing rests on all sorts of wacky assumptions. Some of them seem to be principles that only physicists who specialize in this field understand; others are just plain weird. The most mystifying are the assumptions about the nature of consciousness.
Remember that quantum decoherence takes place when particles are observed by a conscious entity. Elsewhere in the book, Kurzweil dismantles the ideas of "consciousness" and "observance" by suggesting that they can be reduced to sets of machine operations, which in turn can be reduced to simple machine operations that patently have no consciousness. He seems to be saying that there is no such thing as consciousness -- or, at least, that there is no magic switch or threshold for consciousness versus nonconsciousness in a given entity. Yet there must be, because consciousness is a necessary precondition to quantum decoherence. Particles either decohere, or they don't; they are either observed by a conscious entity, or they aren't.
Is consciousness something definite, or is it -- like the concept of "ether" that predated quantum theory, or like Newtonian mechanics -- merely a shorthand that scientists and philosophers have dreamt up to explain whatever is inexplicable in a current theory? Or is consciousness -- as Buddhists believe -- something that exists separately from matter? Are Kurzweil's intelligent machines, which are modeled precisely on human brains, capable of causing quantum particles to decohere? Is "ability to cause quantum particles to decohere" a test for consciousness?
brains, capable of causing quantum
particles to decohere? Is this a test for
To his credit, Kurzweil represents different views on this subject (including his own) and says that while each one has some validity, no single one of them has prevailed. And admittedly, his predictions for the future do not hinge on the existence of quantum computing. However, his predictions do depend on brain modeling and nanotechnology coming to fruition. At the end of each chapter, there is a dialog between the author and a character called Molly, who is supposedly reading the book and asking questions about it. This is a clever device that gives Kurzweil a chance to go back over some of his more difficult points.
Molly is a teenager in the first two parts of the book. The last part of The Age of Spiritual Machines is a series of chapters predicting how life will be in 2009, 2019, 2029, and 2099. In each of these, he makes a few predictions about technology, politics, economics, education, and medical science; then he has long dialogues with Molly as she becomes a young adult, a mother with kids, and an old lady of more than 100 years.
These dialogues are quite imaginative. They paint a picture of the future that is mostly (though not totally) optimistic. Machines get more and more intelligent. Humans take on more and more technology, such as brain implants. Virtual reality and real reality become harder and harder to distinguish. The economy continues to be robust, and living standards become high around the world. Physical things, because they can be built with nanotechnology at any time, have not much meaning; everyone's material needs are taken care of; only ideas matter.
Perhaps the most intriguing of Kurzweil's extrapolations has to do with the nature of knowledge acquisition. As time goes on, more and more of people's lives will be taken up with education. The world will become more and more complicated, and human brains will become more and more powerful thanks to technological enhancement. These two factors will combine to create a world where, eventually, all work will be towards acquiring knowledge. The world will be measured in terms of ideas, not bodies or natural resources.
(As a corollary to this line of thinking: Kurzweil suggests that if alien beings ever come to Earth, they will not appear as human-scale entities in large spacecrafts. Any race advanced enough to send a party to Earth must surely have conquered physical limitations through nanotechnology and other means; therefore they would have to be microscopic in size, and they would be out to capture Earth's knowledge, not its people or natural resources. In other words, it is entirely possible that aliens have already invaded earth many times, but we haven't observed them!)
Another very rewarding part of the book is Kurzweil's discussion of computing paradigms. He says that there are three basic techniques for creating machine intelligence, each of which is inspired by some aspect of nature: recursion (based on mathematics), neural net (based on the human brain), and genetic algorithms (based on evolution). He goes to great lengths to explain these three paradigms in terms that nonprogrammers can understand, and as far as this ex-programmer can tell, he succeeds.
Overall, The Age of Spiritual Machines is a truly fascinating book, whether or not you believe some of its more outlandish predictions. Kurzweil's is one of those minds that is a real pleasure to watch: he has deep knowledge of many different fields, and he is able to deftly bring various strands together to prove his points. His writing also sparkles with light humor and personality, the one drawback being his inability to resist promoting his own products much more often than necessary. Next time you're in the mood for some brain candy, pick up this book.
Title: The Age of Spiritual Machines
Author: Ray Kurzweil
List price: $25.95
About the author
Bill Rosenblatt is market development manager for media and publishing industries at Sun Microsystems Inc.
If you have technical problems with this magazine, contact firstname.lastname@example.org