HAL on Earth
[Ed. Note: No changes without author consent.]According to Arthur C. Clarke's 2001: A Space Odyssey, this past Sunday, January 12, 1997, should have been the date on which the HAL 9000 computer became operational." It was an arbitrary date: Stanley Kubrick's film, written with Clarke from a couple of the author's short stories (Clarke's own 2001 was spun off the screenplay), set HAL's birthday back in 1992, because among the many things Kubrick didn't predict was the relentless pace of technology in the decades to come. At any rate, both birthdays have come and gone, and alas, no HAL. The fictional machine to die the most poignant death since Robbie the Robot has, however, generated a flurry of attention from newsroom to classroom to lab, most of it lamenting our sluggish progress toward "the dream" of artificial intelligence (AI). The University of Illinois at Urbana-Champaign has planned a March Cyberfest dedicated to the Urbana-made HAL; MIT press has just released a book of essays, HAL's Legacy: 2001's Computer as Dream and Reality, edited by Stanford HAL-expert David Stork. "The big dream of creating a HAL failed," MIT's Rodney Brooks told the Los Angeles Times. "HAL, wrote a reporter in Crains, remains an elusive dream." January's Wired pushed a little harder: When will 2001's dream become reality?"Wait a minute when did HAL become a dream? 2001 reads to me more like a nightmare, and when the hatch got stuck on the recent Columbia mission, we should all have been grateful that it was a matter of human incompetence and not a diabolical computer's revenge. (Nevertheless, a recent NASA press release carries the headline "Software for the First Millennium Mission Closest Yet to HAL 9000.") It's hard to believe that we've forgotten panicky Dave Bowman demanding, "Open the pod bay door, HAL!" Or HAL's ominous response: "I'm sorry, Dave, I'm afraid I can't do that." The really curious thing about January 12, 1997, is not that HAL or a HAL-like substance hasn't been brought into being, but that we're all so disappointed about it. It's a little like lamenting that 1984 came and went without Big Brother making more of a fuss.Of course, HAL wasn't really bad he was just drawn that way. His breakdown was caused by humans who had programmed him to tell nothing but the truth and then instructed him to lie to Dave and his fellow spaceman Frank Poole about the intent of their mission. Naturally, the internal conflict drove him nuts. The moral of the story: If flawed beings create a machine in their own image, the machine's bound to have a few problems. If those beings build a machine that's smarter and more efficient than their own brains, watch out.Dream or nightmare, HAL may not be as far off as some people think. At the U.K.-and Arkansas-based Neutronics Technologies Corp., a group of engineers and scientists claim to have designed -- but not yet built -- a robot-like creature called the ENTICY1. I've been following NTC's CEO, Lee Kent Hempfling, around in the Usenet newsgroup comp.ai for a few weeks, wondering if he's for real, noticing that his frustration has been building along with his tendency to correct people on what seem to be minor issues. When, for instance, one user posts a message about "thinking machines" and proceeds to define machine, Hempfling jumps in to remind the reading audience that machine has carried the same definition for ages it's thinking we have to worry about. In an email to me, Hempfling elaborates: "Every single protocol of AI in the past has been based on making a digital machine actually think. But no one can agree on a definition of thinking. "That has resulted, he continues, "in the results of intelligence being duplicated -- not the process itself."Hempfling goes on to describe how "neural net technology" has relied only on binary thinking: on or off, one or zero, black or white. Neutronics, he says, uses fuzzier wave-computation quantum mechanics -- a model much more like the process of the human brain. Trouble is, nobody's listening. "We were shooting for a debut of a quantum computation, speaking, hearing, conscious machine by HAL's birthday, he writes. "And it would have happened. But it was ignored by DARPA, laughed at by the Arkansas Science and Technology Authority and has been ridiculed only because it is not digital." ASTA denied Hempfling's request for a $50,000 grant in May of 1994 because, according to vice president of finance James T. Benham, "the proposal failed to show how this approach to AI would have advantages over conventional computer technology." Last fall, when Hempfling pitched his story to a national news program, he was asked why "Joe Sixpack" should care. "I replied," Hempfling says, "that I gave Joe Sixpack more credit than they did."Given the legacy of 2001, The Terminator and Rod Serling, a better question might be why Joe Sixpack shouldn't be afraid. Very afraid. "A while back I rented 2001 for my family," Hempfling admits, "and the overall feeling was perplexed concern." But NTC isn't bent on global domination -- ENTICY1 was designed for the express purpose of studying human brain disorders. "We build intelligent machines to better understand what makes humans tick, says Hempfling. "Psychology and psychiatry will change drastically when they finally treat the brain for its causes and not its effects. The future is bright for mankind, he concludes. "Not scary."
