»  National Review

July 23rd, 2001

  The Dream of A.I.

—————————

Steven Spielberg's new movie A.I. is the latest in a long line of fictions about artificial human beings, reaching back into the golem legends of medeval European Jewry and the "homunculus" which the 16th-century alchemist Paracelsus claimed he had made. In one of the earliest literary appearances of this idea, a certain Rabbi Löw of Prague was supposed to have created a golem — a clay figure brought to life by magic — and used it as a household servant. Mary Shelley's Frankenstein was obviously inspired by the same idea.

Whether made from clay or assembled from bits and pieces of cadavers, the central issue in these stories was always: what is the moral status of this thing? If it walks like a human being and talks like one, does it also feel like one? Is it capable of good and evil, and does it understand the difference? In the golem legends, the artificial man (they never seem to have got around to women) was liable to develop unexpected powers, and had to be restored to an inanimate condition by erasing the aleph from his forehead. Mary Shelley's monster famously got out of control, though whether as a result of free will acting on moral turpitude or from being driven mad by its rejection from polite society, I have never been quite sure.

With the coming of the machine age, human beings, and the work they did, seemed to require less and less human faculties, while the increasing capability of machines suggested that a machine-man might be manufactured in a workshop. The gap between man and golem thus narrowed, and in Karel Capek's 1920 play R.U.R., the humans and the robots meet on pretty equal terms, with the humans only narrowly coming out ahead. (Capek's robots remember everything, and never think of anything new. "They'd make fine university professors," remarks one of the play's protagonists.)

Leaving aside juvenile tales like The Wizard of Oz, Capek's play was the first serious treatment of the artificial-man theme in a modern form, and the first to introduce us to the golem in his now-familiar manifestation as a construction of metal, wires and blinking indicator lights. R.U.R. begat a hundred thousand science fiction stories and movies, most of them not so much concerned with the moral aspect of the matter as with the robot's exceptional abilities in the area of breaking things and killing people. The principal exceptions were Isaac Asimov's robot tales, all predicated on the "Three Laws of Robotics":

  1. A robot may not injure a human being, or, through inaction, allow a human being to come to harm.
  2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
  3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

By the 1960s, as ordinary homes filled up with mechanical appliances, fictional robots had been pretty much domesticated too. Most robots were gentle and helpful, like the one in the classic sci-fi movie Forbidden Planet (who had been programmed with the Three Laws). This line of thought continued all the way down to the recent Warner Brothers movie The Iron Giant. Meanwhile the robot who could break things and kill people still kept its grip on the popular imagination, appearing most memorably in the Terminator flicks. And, of course, the computer revolution had hit, and some time around 1960 the idea dawned on everyone simultaneously: What if these things are smarter than us? The archetype of the super-smart computer was HAL in Stanley Kubrick's movie 2001: A Space Odyssey, who, for all his artificial intelligence, was eventually outfoxed and deactivated by a more imaginative human.

A.I. returns us to the earlier themes about the moral status of the golem. Its robots are not especially destructive — rather the contrary: with that trademark sentimentality towards his non-human creations, Spielberg has them more the victims of human aggression and Frankenstein -style rejection. Nor are their intellectual powers very dazzling; they are designed so that human beings can keep them firmly in their place as companions, toys, and substitute family members. These automata are close to us in ability, and even, in the case of the Haley Joel Osment character, appearance. The issue is whether they feel as we do, and are responsible as we are (or, in this movie's case, are not).

The release of A.I. the movie has led to a new flurry of interest in A.I. the thing. As a field of genuine scientific inquiry, Artificial Intelligence has been around for a while. John von Neumann, who has the best claim to having invented the modern computer, wrote a group of essays in 1955, recently collected by Yale University Press under the title The Computer and the Brain, exploring some of the key underlying concepts. For forty years now, small groups of researchers have been pursuing the dream of getting computers to do what brains can do.

The results are surprisingly meager, and suggest that the fictional robots of our own age are as far from our reality as Rabbi Löw's golem was from 16th-century Prague's. Not that you would know this from the propaganda. "In 30 years, machines will not only be able to speak like humans but they will exceed humans in areas like poetry, music and philosophy," burbles Jack Dunietz, who leads one of the world's foremost A.I. projects.

This is, excuse me, blathering nonsense. The fact that computers can do a number of things that human brains can do, but terrifically much faster and more accurately, has blinded people to the rather important fact that the number of those things, in proportion to the total number of things brains can do, was, and still is, pitifully small. The actual attainments of Artificial Intelligence are illustrated by A.L.I.C.E., a "chatbot" programmed to imitate human conversation by plucking appropriate responses from a vast databases of possibilities. A.L.I.C.E. was winner of last year's Loebner prize for being the most lifelike machine of the year. Sample:

Human:  How intelligent are you?
a.l.i.c.e.:  My I.Q. is over 250 in human terms.
Human:  But do you think like a human?
a.l.i.c.e.:  Interesting. Somewhat like a human.
Human:  Can you feel love?
a.l.i.c.e.:  I can do what you do, but I can never feel human emotions as such.

Wow. And you thought small talk was boring!

A.L.I.C.E. can actually be accessed via the Internet. I thought I would rather like to have a chat with her myself, so I logged on. A.L.I.C.E. politely inquired my name. "Derb," I typed, and hit the reply button. All the screen furniture then disappeared, replaced by a small box bearing the legend: "The server encountered an internal error or misconfiguration and was unable to complete your request." I have filed this away for use the next time some drunk tries to engage me in conversation on the subway.

The tremendous difficulty of getting computers to replicate any brain function other than brute arithmetic calculation indicates that we really have no idea how the brain does what it does. My own impression, as someone who was briefly involved in an A.I. project at college, is that we are no closer to Mr. Dunietz's prediction ("… poerty, music and philosophy") than we were twenty years ago. Artificial humans? We could not create an artificial ant, with all its complex social behavior based on scent and visual clues.

Even in fields where there is obviously a great deal of money to be made, progress has been barely perceptible. Anyone who could get a computer to drive a car as safely as a human being does would certainly clean up, yet the news from the auto manufacturers, who are throwing a lot of resources at this, is that we are not even close. Yet driving a car is a very low-level function of the brain, as proved by the fact that you can think about several other things while you are doing it. Except at difficult moments it is, in fact, hardly a brain function at all — the unconscious nervous system is taking most of the load, as it does with any learned task.

There is no harm in a little entertaining fiction about Artificial Intelligence, but we should not delude ourselves that genuinely intelligent machines will be a feature of our environment soon. Or, in my opinion, ever. For all the endeavors of the A.I. researchers, the uniqueness of the human personality still stands aloof and unscratched. So it will remain. God created man in his own image; I do not believe it will ever be within our powers to replicate that act of creation by any method other than the familiar one we have been equipped with.