Robots Are Getting Closer - 10
The Long-Term Roles of Robots
Our collective fears of clockwork mechanical men hearken at least as far back as the 14th century to the clockwork figures that tolled the hours.
"The first detailed records of a mechanical clock are from Norwich Cathedral in England, and had "models of the Sun and Moon, automata, including 59 sculpted images...and a choir or procession of monks. There was much colouring and guilding..." and the clock cost about $250,000 (in 1975 dollars) to build. Construction was carried out from 1321 to 1325, not long after the records begin mentioning clocks."
"Haber theorizes that Descartes may have gotten his idea of animals being automatons from seeing the motions of the life-like rooster on the cathedral clock."
"Montaigne was particularly impressed by the organs that played music to the accompaniment of the fall of water and devices which imitated the sound of trumpets. He related how birds began to sing and how, when an owl appeared upon a rock, the bird song ceased abruptly. This sequence, borrowed in its entirety from Heron of Alexandria, was to be borrowed again several decades later by de Caus."
"This is like what happened with Vaucanson's famous mechanical duck, the duck which aroused such controversy that it still features today in the saying ``if it walks like a duck, and quacks like a duck, then it is a duck.'' In 1738 Vaucanson exhibited his marvelous mechanical duck to an astonished Paris. It had multiply jointed realistic wings, could move its head around and mimic the swallowing neck movements of a duck, ``eat'' grain, splash water, etc.. The Parisians were used to ingenious clockwork automata which played whistles, wrote with pen on paper, etc., but what astounded them about this duck, and convinced them that it was a real step forwards towards artificial life, was that it had guts made of rubber hose and actually shat evil smelling duck turds soon after eating. Unlike all the other ingenious clockwork automata of the time, this one seemed to approach the miraculously self-sustaining feature of life of (seeming to) get its energy from grains instead of clockwork springs. Although it was fastened to a large plinth full of the gears and pulleys that made it work, the press of the day, being just as gullible as today's concerning these matters, soon had it capable of walking and swimming and really nourishing itself on grains. And of course, everyone started asking ``If this is what can be done now, what on earth will be possible in another 50 years? If it walks like a duck, and quacks like a duck, will it really be a duck?''
Mary Shelley sounded this theme in her story of "Frankenstein". Karel Capek elaborated on it in his 1921 play "R. U. R" (Rossum's Universal Robots)."1921. In the 1954 movie, "The Day the Earth Stood Still" (?), someone asks Klaatu, the human-looking alien, something about his robot, "Gort". Klaatu replies, "You don't understand. He is the master. I am the servant."
In the 1950's,
computers were hugely faster and more accurate than people in performing
arithmetic, and in other rote-mechanical mental tasks, and they were getting
faster every year. It was thought that in a few more years, computers would
surpass people in all spheres of activity except having babies.
Didn't happen. Our fastest 2002 supercomputer is 10,000,000,000 times faster than the fastest supercomputer (UNIVAC I) of 1952, my desktop Athlon Thunderbird is 1,000,000 times faster than UNIVAC I, and they're both as dumb as fenceposts. (I'm glad. I have enough trouble uploading these web pages without having computer that goes on strike for a little more tender, loving care, or nags at me for making typographical mistakes.) They're nothing more than mindless machines. Hans Moravec has observed that Gary Kasparov felt that he detected a real personality in the IBM computer that bested him in chess, Deep Blue. But Deep Blue didn't fight being turned off, or give the first sign of being self-aware. And I'm sure it isn't.
Computers can convert speech-to-text and text-to-speech. They can recognize faces. Customer interaction systems will be able to answer pre-programmed questions, and to provide pre-programmed services. Computers can be run graphically, with mice, or by spoken commands. But none of this involves the slightest awareness on the part of the computer.
Dr. Moravec's visual navigation module, will, I think, allow a computer driven platform to build a three-D shell model of its environment, and to navigate through it. Still, this, in itself, would seem to me to be another enabling technology, like speech recognition (including the ability to understand the meaning of speech), and speech production (Including, ultimately, the ability to frame sentences).
I think it will be possible to give computers self-awareness and volition, but I don't think it will happen by itself. I think it will take a great deal of research and product refinement to get there. One of the questions is that of whether we want to get there. I think we'll want to be careful what we do. I think, though, that we'll gradually create more and more equipment with intelligent safety controls. For example, anti-lock braking systems take over from a human in emergency braking situations. Autonomous steering and vehicle controls will probably take over if a driver shows signs of running off the road, or of running into another vehicle. We don't really think of anti-lock brakes having volition and a mind of their own, but I suppose that in a sense, they do. One of the applications for human-like robots is that of companionship. Ideally, we would like our inorganic companions to develop an attachment for us. Robotic pets are a possibility.
I don't know just how far we are along this road.
A baby must somehow identify with her own kind, rather than with a tree or a pet kitten. She imitates others of her kind, and compares her performance with the performances of those around her. She must be serious, but mustn't concentrate permanently on one topic to the exclusion of all others. She mustn't stare at the sun, ever. She must acquire a fear of heights based upon limited experiences.
The point of this is that there's complexity in designing a system to be self-aware the way an animal is self-aware. A combination of internal stimuli ("I'm hungry." "I'm hot." "I could go to the bathroom.") and external stimuli ("What was that sound?" "That big thing is getting closer and closer.") are competing with whatever is engaging an organism's attention at any given moment. Which one wins out depends upon which one has the highest relative urgency. It's on the basis of whatever feeling is dominant that an animal acts.
What I'm driving at is that there seems to me a great deal of detailed work must be done to create an artificial intelligence that will emulate animals or humans. It must abstract from the environment, just as the animal kingdom abstracts from the environment. It must correlate perceptual abstractions with stored abstractions, either reinforcing them, or defining a new category of abstractions. Correlations must be weighted in accordance with the importance (severity) of the experience, the uniqueness of the experience, and the number of times it recurs. It must forget over time, so that invalid correlations fade away. Drives (pleasure, pain) will underlie emotions. Emotions will provide incentives to act.
To say it again, increasing computer speeds by a factor of 10 billion hasn't led spontaneously to self-aware, purposeful computers.
I don't think that our collection of enabling technologies, including speech recognition and visual navigation, are going to spontaneously result in anthropic robots.
I don't think that adding inference engines and other logical operations to artificial intelligence is going to give us a humanoid robot. Purpose, spontaneous learning, and other human characteristics are still going to be missing.
I think it can be done, but I think it may take teams of experts a while to do it.
On the other hand, we're getting closer and closer.
9 10 11