Until next week,Dateline: 03/16/97
HAL was born on January 12, 1997.
For the benefit of those of you who don’t watch movies or read science fiction, HAL was a superdupercomputer conceived about 30 years ago by science fiction writer Arthur C. Clarke and Hollywood director Stanley Kubrick for a movie (and a subsequent book) called 2001: A Space Odyssey.
He (HAL was given a male persona; it's interesting to speculate on whether the story would have been different if she were female!) was an artificial intelligence, capable not only of monitoring and flying the spaceship on which he was installed, but also of playing chess with the human crew, chatting with them, expressing emotions,… and planning and committing murder.
Having been programmed to ensure the space mission’s success at all costs, HAL bumps off the crew members one by one when it decides they are jeopardizing the mission. One crew member manages to survive, and "kills" HAL.
In the book, HAL is declared to have been "born" in 1997. As the title makes obvious, the space mission takes place in the year 2001. For the arithmetically challenged, that’s four years hence. So the question is: Will there be an artificial intelligence with HAL’s capabilities by 2001?
This is an aspect of the fundamental question posed by Professor David Stork in his recent book, HAL's Legacy, and answered, in the book, by expert witnesses such as Marvin Minsky, Doug Lenat, and a dozen other top names in the AI world.
If there’s any consensus among these folks at all, it’s probably that HAL will not be with us in 2001; that it will take much longer. But Dr. Minsky, as wise to the AI world as anyone, is prepared to concede that it could happen in four years… or four hundred.
Keith Devlin, author of Goodbye, Descartes: The End of Logic and the Search for a New Cosmology of the Mind, believes that the goals of AI, formulated at about the same time as Clarke and Kubrick were conceiving HAL, have been a flop. But, he acknowledges, there remains a spark, and you can find it in . . .
". . . your Web browser, the program you use to find your way around the World Wide Web. Browsers work by sending out small pieces of software, often called "softbots" (for software robots). Once released from your browser, these agents have to fend for themselves in the complex world of cyberspace. Their success depends on their ability to make "intelligent" decisions on their own. Softbots are a direct descendent of AI research.
With softbots, we have the operational equivalent of an ant. But few of us would be prepared to regard ants as intelligent."
True. But Prof. Douglas Hofstadter’s charming character, Ant Hilary (in Hofstadter’s magnum opus, Gödel, Escher, Bach: An Eternal Golden Braid), might argue that Professor Devlin is missing the big picture. Ant Hilary would say that while the ants themselves are pretty dumb, their collective behavior makes for a demonstrably smart mind, which can talk with its crew (causing them to go off and, say, repair a hole in the nest wall), monitor and maintain its own life support systems, plan for contingencies such as a rainstorm or a visit by an anteater (such as Ant Hilary), and plan murderous food-gathering raids on the surrounding environment.
Hofstadter’s point is that higher-level intelligence (an ant colony; a human mind) is a feature that emerges from the activities of lower levels of intelligence (ants; neurons).
In 2001, it seems to me, Clarke and Kubrick missed the crucial prophecy that HAL would be just one node in a network of millions of nodes. Even in space, HAL was communicating with Earth, albeit sporadically and with long time delays.
The day might come when the power, connectivity, and memory capacity of a standalone machine like HAL rivals the power, connectivity, and memory capacity of today’s entire Internet; but so what? The machine will still be just one (very smart) node on a network of millions of other (very smart) nodes. If HAL has the brains of an ant, then the network of HALs could have the mind of a colony. But if HAL has the brains of a human, then the network of HALs could have the mind of civilization.
In short, HAL’s emergence as a true artificial intelligence endowed with the capabilities described in 2001 may depend more on the development of connections—network—among him and his buddies than it will on human programming and machine-building skills.
Feel free to agree or disagree!
Until next week,
NEXT WEEK: The Electronic Brain Revisited. Professor Theodore Berger and colleagues expect to produce chips that can be implanted in animal brains in about five years, and in humans a decade or so later. What does this have to do with AI? I dunno yet, but I’ll have it figured out by next week. Seeya then!