It's been a lively month's mail -- so much so, that I could omit my customary statement of policy, and nobody would misunderstand. As a creature of habit, I'll repeat it anyway: I print the letters that add information, give me a scrap, or entertain.
Another matter: I'm not above running in less colorful substitutes for words which would get this publication involved in pointless broils. I save my sufferance quota for controversies that are important to me.
Comments of my own are enclosed in double parentheses ((like this)).
| May 10, 1971 |
It's always a pleasure to receive "A Word In Edgewise." You not only know what's happening in AI, but obviously love to communicate; such a combination is pretty rare.
Now for the rumble. You say you'd like to see "an artificial intelligence that is perfect and imperfect in pretty much the same ways as human beings." My advice to anyone who is that serious about AI is, first of all, to put away all the clever little heuristic toys and games, the "programmed bags of tricks" as you call them, the robot wheels, the ELIZA dolls, and all the rest of the nursery school supplies. Next, invest $8.95 in a copy of Beyond Reductionism (Koestler and Smythies, Macmillan). Read it -- especially the chapters by Waddington, Piaget, and Hayek -- until the message sinks in. (There have been many such books around for years, but this is one of the broadest and most up-to-date.)
What's the message? Well, there are several, but the main one for us is that the great ball game of intelligence, adaptation, language, and thought just cannot be played in a sandbox. After 20 years, we obviously don't need fancier sandboxes. We need some people who know how to lay out a ball field. In other words, it may be okay to say "intelligence is as intelligence does," but let's at least try to be among those who are inquiring what intelligence does.
I am convinced that the important concepts needed for this job are both inter-disciplinary and non-reductionist. Mathematical and simulation techniques are already available, if we're willing to dig for them. Moreover, it's possible to begin on a small scale, provided we can keep the objective in mind, and not get sidetracked by anything that is obviously inconsistent with real life processes. Unless AI workers can keep this kind of faith, I'm afraid their efforts are going to remain ad hoc / ad nauseam all the way.
|
Paul B. Post Norwalk, Connecticut |
|
|
((Hey, there's one that livens up the party! For once, though, I'm speechless: I haven't read the book, and won't comment until I do. I will say that there's an abundance of smoke, anyhow: you're the second person to mention that book to me, and the third to confuddle me with this talk about "reductionism." At the risk of deflating your gracious buildup, I'll admit that I didn't know this was happening.))
| A Word In Sideways |
I suppose you would solve the population problem for those who refused to leave Earth by developing smaller and smaller robots. The total power consumption of a family could be limited, forcing parents to trade in on a smaller chassis before ordering roblings.
You portray a world of simulated immortality -- not of the flesh, but of the personality and the memory. For his survivors, a saved letter from the person or a photograph (or for the more dramatic, a movie or an autobiography) would be at least as satisfying as a robot model -- if the mere memory of the dear departed and the works he left behind were not enough. And these media never make demands, never propose simulated sex.
The vibrations of an electro-metallic shell are cold; too much time in cars has accustomed you to them. Let us develop intelligent machines to assist organic life -- not to replace it -- and in death (whether by nuclear holocaust or an auto accident), be content to reunite our flesh with Mother Earth.
|
A Warm Heart, Larry Tesler Menlo Park, Calif. |
|
|
((It's warm for a distressingly little while. Besides, as heart transplant surgeons have been reminding us, it's just a pump. We can't be too grateful that it's there, under the circumstances, but it's just a pump. For that matter, the vibrations that keep it warm are just muscle spasms. You can't be serious that that's the real you.
You're begging the question when you keep calling everything "simulated." Obviously, survival that was only an illusion of the spectators wouldn't be worth much. Repetitious assertion that it's got to be illusory, however, is no sort of argument. Can you show probable cause? And is there any test or experience which would satisfy you that it wasn't an illusion -- or is this just some unchallengeable prejudice which would leave you feeling "simulated" and miserable even after a successful transplant?))
| May 3, 1971 |
... Random thought: what if we set up John Campbell as a scientific/technical entrepreneur with a $1 billion/yr. budget?
|
John K. Dixon NIH Bethesda, Md. |
|
|
((Like as not, he'd only while away the hours writing to impostors about crackpots. See below.))
| May 12, 1971 |
Your "Word In Edgewise" item about the Columbus problem misses the really lovely complication of that deal -- and a lot of crackpot-genius cases.
1. Columbus held that the world was round -- as did all educated people of that time. (Take a look at Archimedes' original proof of his principle of flotation: he argued on the basis of water having a spherical surface, because he knew the Earth was round.)
2. Only the dolts in the street thought the Earth was flat. (Remember that a survey showed that 20% of Americans don't believe we had men on the Moon; they think it was all another episode of Star Trek.)
3. BUT Columbus was a crackpot with the nutty idea that the Earth was only about 8,000 miles around, and that he could sail westward from Spain all the way to India. Everyone who knew anything about navigation knew damn well the Earth was over 20,000 miles around, and that the westward distance to India was nearly 10,000 miles.
4. With the marine technology available at that time, a voyage of 10,000 miles over the open sea that lay between Spain and India was completely impossible; all the water and supplies that could be carried on ships, with the food-preserving techniques of the time, would be inadequate. All hands would starve and/or die of thirst on that vast and trackless ocean.
5. Conclusion: Columbus was a crackpot of purest ray serene -- off his everlovin' nut -- to think he could cross the vast expanse of empty ocean. The educated people of Europe were perfectly correct in saying the whole idea was crazy, and Columbus would simply kill himself, his crew, and lose his ships.
Just take a look at a globe, and see how immense a distance lay between Spain and India, and you'll see he was nuts.
Because, of course, neither Chris nor anyone else in Europe had any idea that the Americas existed, and had no reason whatever for guessing it.
So Columbus was completely wrong, and the scholars were perfectly correct ... except that by the sheerest bull luck he accidentally ran into huge continents nobody knew existed.
And even then he was such a bull-headed crackpot he never did acknowledge he'd found new continents; to the day he died he maintained it was India.
The Columbus problem is, really, "What do you do with a crackpot who is completely wrong ... but accidentally by mistake does something valuable. And then won't admit it!"
|
John W. Campbell Editor ANALOG Science Fiction/Science Fact |
|
|
((What kind of rhetorical point can you make with a "Columbus problem" like that? Anyhow, I have my misgivings that history presented itself to the participants in any such light.
It's easy enough now to say who were "all (the) educated people of that time" and who were "only the dolts in the street." But how did matters look then? What was the insitutionalized wisdom of the time? When a policy-maker called in the kind of certified "expert" that policy-makers are wont to trust, which did he get: a "dolt" or an "educated people"?
Columbus frequented the courts of Europe for ten years before he got his chance, and was turned down all that time, including once by Isabella. Is it anywhere recorded that anyone who was actually consulted advanced the reasonable argument which you suggest? Do you know the name of anyone who stepped forward afterward and said, "I told you so!"? Or did his erstwhile detractors act embarrassed and confounded -- like maybe they'd been arguing that the Earth had to be flat, so that the Biblical prophecy might be fulfilled that "All flesh shall see it together"?
Is it on record that Columbus heard the argument about distances and rejected it? From whom? And lastly, if the true dimensions of the globe were so evident to everyone but Columbus, why did Magellan, twenty-seven years later, stake his fortune on a westward route to the East Indies? Even with the Americas for a reprovisioning station -- which was a problematic view of the Americas at the time --, it clearly wasn't going to be an economic route. Clearly, that is, if you foresaw the distance.))
| May 1, 1971 |
Your last A Word In Edgewise stimulated the poetry in me thusly:
|
In the beginning was God. (Creation). |
* Alternate title: EVOLUTION SQUARED.
|
Dick Boyajian Chicago, Illinois |
|
|
((With that squared term, it might be the equation of a circle.))
| May 2, 1971 |
Although I'm not at all sure that I'd like to be transplanted into a robot -- even if this would guarantee eternal "life" -- I can see that this may well occur in the future. How could living be half so sweet if there were no element of chance, no thought that all this might end with a boom or a whimper?
If a man wants to be immortal today he has only to leave works behind which are "worthy of preservation" (whatever that means, future generations will always judge).
|
Charlotte Hollenberg Uniondale, N.Y. |
|
|
((I'm as oriented toward that kind of "immortality" as anyone born under the current dispensation. Guess what worthy work I'd like to leave behind, if leave I must? But it was pointed out long ages ago why works are a poor surrogate for immortality:
|
Short-lived are both the praiser and the praised, and the rememberer and the remembered; and all this in a nook of this part of the world; and not even here do all agree, no, not anyone with himself; and the whole earth too is just a dot. ...See that you secure this present time to yourself; for those who rather pursue posthumous fame do not consider that the men of after time will be exactly such as these whom they cannot bear now; and both are mortal. And what is it in any way to you if these men of after time utter this or that sound, or have this or that opinion about you? |
|
Marcus Aurelius Meditations (VIII) |
It's unfair to quote Marcus Aurelius in the service of discontented strivings which he would surely abhor -- but for those who are less equable than he in the face of extinction, he does put a bit of vinegar on the pacifier, doesn't he?))
| March 29, 1971 |
Would you agree with my approach on the enclosed sheets?
((Accompanied by the following enclosure, dated November 15, 1968)):
SUBJECT: CAPABILITIES OF COMPUTERS RELATIVE TO CAPABILITIES OF HUMANS.
This is a note to set forth a concept of the relationship of what a computer can do to that of what humans can do. I do not claim uniqueness but I feel it is worth setting forth for consideration.
I assert that computers will meet and exceed human intellectual capabilities at some point in time. This leads to the conclusion that at some point in time computers will, in fact, control people rather than vice-versa.
The basis for my assertion is as follows:
1) Human beings are physical entities composed of chemical components that obey the laws of physics.
2) The intellectual capability of a human is delimited by his physical-chemical composition.
3) Computers are also physical-chemical systems and there is no fundamental limitation (relative to the chemistry of a human) on what physical-chemical system can be put together to make a computer.
4) Therefore, a computer can be developed with the intellectual capabilities of humans.
5) One can reproduce computers faster than a complete human development cycle. Thus, once a single computer has been developed with the intellect of a human, it need only be a short time before many of these are combined to produce a greater intellectual capability and the computer's intellectual capacity exceeds that of the human.
6) As a result, computers will utlimately control people.
As I see it, the week link in this argument, if any, is in (2) above. However, I know of nothing that refutes this point.
I realize that the time when this computer capability will exist is not in the immediate future. However, one should remember that computers as we know them have only been around about 30 years and great increases in capability have been made in that time. I say, following Al Jolson, "You ain't seen nothing yet." This computer capability should exist by the year 3000 certainly and maybe well within 100 years. Thus, we might replace 1984 by 2068.
The means of putting together computers of this capability may well differ from what we do now. It may incorporate biological systems or other vastly different technologies. The only technique I rule out is that of human reproduction.
There may be an integration of man and machine physically. Then it would not be clear which (if indeed which is the operative word) would be in control. I still feel the machine ultimately would not need the man.
|
Milton J. Waxman Northridge, Calif. |
|
|
((I held this letter over because worries like that are the topic of the current issue. As you see, I pin my hopes on a particular variant of what you call "the integration of man and machine physically." To wit, the man gets transplanted into the machine at a certain stage in his career, and the machine is a continuation of the man.
If that's possible, then the "machines" will continue to need (untransplanted) men in the same sense as parents "need" their children: not as instruments, but as kin. Which is not to say that every older person is fond of every younger one that walks the earth; it suffices that most older people be tied by duty and affection to somebody or other in the younger generation, and then the question, "Do we need them (as a class)?" cannot be effectively raised. Of course, I'm assuming that the transplantation technique preserves existing affections; we shouldn't certify any method that fails in this particular.
Anyhow, if the machines are really us, it'll be we who do the "needing" or "not needing." If eventually we decide to reproduce ourselves by manufacture, that's no scarier than, say, a gradual abandonment of pregnancy in favor of test tubes and tissue cultures. (Women's Lib might even applaud.) It's not the same prospect as having something that is not us decide that our present form is expendable, while we're still stuck with it.
No doubt the transplanted elders would be in control. See how different it sounds when you say it that way, as opposed to, "The machines will control the men"? The whole thing is reduced to a "conflict of generations" problem. That's not to be brushed aside lightly; the lengthening of the life span and the enhancement of transplanted people's abilities would aggravate the problem beyond anything we have known. Still, it's at least a familiar problem, and there are reasons why it doesn't get out of bounds in the same manner as territorial conflict. It presents no difficulties that we couldn't back ourselves into by successful research in geriatrics.
If this sort of integration with machines is not possible, or if we neglect to prepare it while we're putting intelligence into machines, then I think we're in the sort of trouble you suggest. We shouldn't just drift.
My scientific predilections are more or less in line with your first five points, but I wouldn't state any of them without careful hedging. Specifically:
1) and 2): Anything that we eventually find to be part of man's composition and necessary to his working, we'll no doubt call "physical" or "chemical," as these are our words for whatever we understand about the composition and inner workings of things. It doesn't follow that the phenomena we now intend by these terms are sufficient to account for man and his intellect. The real question is whether phenomena yet to be discovered are so different from the known ones, and so necessary to the understanding of intelligence, that our present efforts are merely misleading. I don't lean to that supposition, but I have a healthy respect for what's an open question.
3) and 4): What's a "fundamental" limitation? It is at least conceivable that nature's prodigious use of components is a necessary part of the trick. It may follow that to construct a thinking device on a scale that we have room for on earth, or can make many of, we have to use components as microscopic as those used in the original model. It may then follow that the only materials worth consideration are exactly the organic ones selected by nature. And finally, it may follow that the cheapest method of fabrication is ordinary human reproduction. Is that a "fundamental" enough limitation?
Nothing could be more damaging to my schemes, and I hope it isn't so -- but one has to be honest about the possibilities.
5) Who says that thinking computers can be reproduced faster than a complete human development cycle? We know nothing about feasible processes of maturation and learning, let alone optimal ones. At the back of your mind, you must be reasoning like this: (a) A human being matures for twenty or thirty years, and arrives at a certain physical-chemical state. (b) By then, he is a mature thinking organism. (c) If we could figure out what it is in the physical-chemical state that is relevant to the thinking, we could reproduce the thinking ability without going through the personal history again.
But how do we know that the reproductive process would be something short, like automobile manufacture? It might be more like shipbuilding, or ten times longer than that. Indeed, for all we know, the appropriate process might be brewing. Physical-chemical states are not arrived at just any whichway.
Not that I really mean to argue with you; my best guesses, up to the sixth point, are like yours. I just like to cover all the bases.))
| Monsters, Slaves, and Honored Ancestors: The Three Lores of Robotics |
|
 |