The Age of Intelligent Machines
"Our Concept of Ourselves"
by Ray Kurzweil
We know what we are, but know not what we may be.
William Shakespeare
What will happen when all these artificially intelligent computers and robots leave us with nothing to do? What will be the point of living? Granted that human obsolescence is hardly an urgent problem. It will be a long, long time before computers can master politics, poetry, or any of the other things we really care about. But a "long time" is not forever; what happens when the computers have mastered politics and poetry? One can easily envision a future when the world is run quietly and efficiently by a set of exceedingly expert systems, in which machines produce goods, services, and wealth in abundance, and where everyone lives a life of luxury. It sounds idyllic - and utterly pointless.
But personally, I have to side with the optimists - for two reasons. The first stems from the simple observation that technology is made by people. Despite the strong impression that we are helpless in the face of, say, the spread of automobiles or the more mindless clerical applications of computers, the fact is that technology does not develop according to an immutable genetic code. It embodies human values and human choices... My second reason for being optimistic stems from a simple question: What does it mean to be "obsolete"?
M. Mitchell Waldrop
As I discussed earlier, I believe that a computer will be able to defeat all human players at the game of chess within the next one or two decades. When this happens, I noted, we shall either think more of computers, less of ourselves, or less of chess. If history is a guide, we will probably think less of chess. Yet, as I hope this book has made clear, the world chess championships is but one of many accomplishments that will be attained by future machine intelligence. If our approach to coping with each achievement of machine intelligence is to downgrade the intellectual value of the accomplishment, we may have a lot of revision to do over the next half century.
Let us review some of the intellectual domains that machines are likely to master in the near future. A few examples of tasks that computers are now beginning to accomplish include the following: accompanying musical performances, teaching us skills and areas of knowledge, diagnosing and recommending remedial treatment for classes of diseases, designing new bioengineered drugs, performing delicate medical operations, locating underground resources, and flying planes.
A more difficult task for a computer, one that we shall probably see during the first half of the next century, is reading a book, magazine, or newspaper and understanding its contents. This would require the computer to update its own knowledge bases to reflect the information it read. Such a system would be able to write a synopsis or a critique of its reading. Of comparable difficulty to this task is passing the Turing test, which requires a mastery of written language as well as extensive world knowledge.
Of at least comparable difficulty would be to watch a moving scene and understand what is going on. This task requires human-level vision and the ability to abstract knowledge from moving images. Add the ability for a robot to imitate humans with sufficient subtlety, and computers will be able to pass a more difficult form of the Turing test in which communication is not through the written word transmitted by terminal but rather by live face-to-face communication. For this achievement we have to go at least to late in the next century.
It is clear that the strengths and weaknesses of machine intelligence today are quite different from those of human intelligence. The very first computers had prodigious and virtually unerring memories. In comparison, our memories are quite limited and of dubious reliability. Yet the early computers' ability to organize knowledge, recognize patterns, and render expert judgements - all elements of human intelligence - was essentially nonexistent. If we examine the trends that are already apparent, we can see that computers have progressed in two ways. They have gained even greater capacities in their areas of unique strength: today's computers are a million times more powerful in terms of both speed and memory than their forebears. At the same time, they have also moved toward the strengths of human intelligence. They are nowhere near that goal yet, but they are certainly getting closer. Today computers can organize knowledge incorporating networks of relationships, they are !
beginning to recognize patterns contained in visual, auditory, and other modalitie
s, and they can render judgements that rival those of human experts. They have still not mastered the vast body of everyday knowledge we call common sense, and ironically, they are particularly weak in the pattern recognition and fine motor skills that children and even animals do so well.
Computer intelligence is not standing still. Radical new massively parallel computer architectures, together with emerging insights into the algorithms of vision, hearing, and physical skill acquisition, are propelling computers closer to human capabilities and also continuing to enhance their historical areas of superiority. While machine intelligence continues to evolve and move in our direction, human intelligence is moving very slowly, if at all. But since we have computers to serve us, human intelligence may not need to change.
Thousands of years ago, when the religious and philosophical traditions that still guide Western civilization were being formed, a human being was regarded as special. We were different from animals and certainly from material things. The ultimate intelligence in the universe, God, knew about us, and cared about us. Later on as we learned that the earth on which we stood was not the only celestial body in the world, we imagined that all the other entities in the sky revolved around us. In this world view we were special because of our central location. The sun, the moon, the stars, the comets, and other celestial objects all paid homage to us. Still later when we realized that the earth was just the third planet orbiting an unremarkable star located on the arm of an unremarkable galaxy, our view changed again. Then we were special because of our unique intelligence: we could derive knowledge from information. We could create patterns with aesthetic qualities. We could apprec!
iate those qualities. True, animals shared in this intelligence, but to a much lesser degree, which only reinforced the uniqueness of the level of intelligence we possessed.
Now we are entering an era in which this latest concept of our uniqueness will be challenged once again. To be sure, this challenge will not arrive on a single day. By the time one can seriously argue that computers possess intellectual capabilities comparable to the human species, it will have been at least a century from the invention of the electronic computer in the late 1940s. We should have time to adjust. Perhaps we shall return whence we started, with an appreciation of the inherent value of being human.
Copyright © 1996, Kurzweil Technologies, Inc.