2001 was a year burdened with expectation that blossomed with
disappointment. Thanks to creative visionaries such as Arthur C. Clarke and Stanley Kubrick, the first year of the 21st century was anticipated with somewhat unrealistic notions of what life would be like. Now, with the tail-lights of 2001 fading into the past, we’ve come to terms with the fact that nobody’s yet colonized the moon and that NASA isn’t sending manned missions to the outskirts of the solar system. However, we are still nagged by one question: where’s HAL?
According to Mark Shainblum, a science fiction author and head writer for Airborne Entertainment in Montreal, HAL served as little more than a mirror to 2001: A Space
Odyssey’s 1960’s audience’s discomfort with what they viewed as the dehumanizing influence of computers.
“HAL was our child coming back to destroy us,” Shainblum explained. “He was Frankenstein’s monster again in new clothes.”
Whether the literary and film creation of the HAL 9000 was an exercise in metaphor or a forecast of things to come, the computer on the Discovery in 2001: A Space Odyssey has captured the imaginations of not only science fiction fans and moviegoers, but of scientists, engineers and astronauts. Whether or not they are consciously tipping their hats to HAL, scores of academics and inventors have taken its capabilities seriously and have worked towards creating a real-life iteration of HAL – minus its diabolical tendencies.
Doug Lenat, president and founder of Cycorp in Austin, Tex., is one HAL fan who believes that the computer was much more than a literary device used for teaching a lesson.
“I think that HAL was an absolutely brilliant, well thought out prediction and roadmap for us to follow,” Lenat said.
Richard Bergquist, chief technology officer and senior vice-president at PeopleSoft Inc. in Pleasanton, Calif., agrees with Lenat.
“If you take a step back and look at HAL, it was supposed to be an aid for people to better do their jobs,” Bergquist said. “It was there to assist people to make better decisions, and I think that in those terms we’re actually marching along fairly well.”
Whatever the original purpose of HAL might have been, technology is making huge strides towards
realizing a computer capable of HAL’s abilities and
intelligence. In many respects, HAL is already here in a basic form, although, unlike Frankenstein’s monster, it’s still unassembled – it currently exists only in bits
and pieces.
The problem is less about the technology than about putting these pieces together, according to Masood Jabbar, executive vice-president of global sales operations for Sun Microsystems Inc. in Menlo Park, Calif.
“It’s a matter of connecting the dots,” Jabbar said, admitting that the technology and the knowledge exists, in at least a rudimentary level in what he calls “silos of information.”
One of the key pieces to the HAL puzzle is voice recognition. It was HAL’s conversational abilities onboard the spacecraft Discovery in 2001 that made the computer both so compelling and so creepy. The characters in the movie went so far as to leave the room to discuss HAL, only to realize – too late – that the computer was skilled in lip reading as well.
Voice recognition technology has made leaps and bounds since AT&T released its first attempt at artificial singing in 1962, which served as an inspiration for HAL’s death scene in 2001. The computer-generated “Bicycle Built for Two” blew its listeners away and voice recognition, auditory and acoustic applications have been a focus for developers ever since.
Steve Chambers, vice-president of worldwide marketing at SpeechWorks International Inc. in Boston explained how his company makes use of speech recognition technology.
“We have two types of technology: speech recognition, which is kind of self-explanatory, and speech synthesis,” Chambers said. “We digitize your audio and break it down into phonemes. The technology then matches the phonemes against a database.”
The technology is able to catch and clarify similar-sounding words – like Boston and Austin – and make appropriate corrections, according to Chambers.
In a keynote at Microsoft’s professional developers conference in late 2001, Rick Rashid, senior vice-president of research for Microsoft in Redmond, Wash., demonstrated some of the ways that auditory and acoustic technology is being used by playing the digital-sounding “Bicycle Built for Two” recording back to back with a more recent computer-generated version of “Scarborough Fair.” While Rashid admitted that it’s unlikely that any singers are going to lose their jobs to computers in the near future, there are useful applications for the technology.
“You may want to capture people’s voices that are famous singers or have a particular meaning to you personally or meaning to society for the purposes of being able to recreate singing or to do new things with their voices after they’ve passed away,” Rashid said.
Despite these kinds of breakthroughs in auditory and acoustic technology, Lenat asserts that there is a long way to go.
“Speech understanding technology has not advanced significantly in the last 30 years,” Lenat said. “Machines are faster now, but the success rate is not much better than it was
in 1972.”
“Look, Dave, you’ve got a lot of things to do.
I suggest you leave this to me.”
Lenat reasoned that the development of speech recognition technology has been retarded because in order for a computer to be able to converse, be it audibly or through the use of a keyboard, it must possess common sense. This is a problem that he has tried to tackle since 1984, when the former Stanford professor began teaching Cyc – artificial intelligence software – basic facts about the world in order to create a base of common sense.
“We use our knowledge of the real world, context and common sense to understand what others are saying,” Lenat said. “The computer can take knowledge and draw the same conclusions that you or I can draw by representing information in a logical form using predicate calculus.
“For example, you could input a piece of knowledge such as: Osama bin Laden runs al Qaeda, and that al Qaeda is a terrorist organization, and that terrorists sometimes attack civilian targets. If you ask the question, ‘Does bin Laden ever order people to attack civilian targets?’ Cyc would answer yes, even though it doesn’t have the piece of information that people attack civilian targets. It can reason its way to the answer,” Lenat explained.
“Reasoning makes an enormous difference. Without reason, it’s like having a library of books in Chinese, but you can’t read them,” he said. “They’re of no use to you, but if you have the same books in English and have read them, they’re of tremendous use. The task is not to figure out what to put down, but how to put it down in logical form.”
The first piece of information fed to Cyc back in the 1980s was “Napoleon died in 1821. Wellington was greatly saddened.” In order for Cyc to understand what these sentences meant, Lenat and his team had to explain concepts such as death, time, war and emotion.
“This snippet of information sprouted additional pieces of knowledge like ‘people hear about events that occur’ and ‘you don’t hear about things after you’re dead,’ so that when we input the question ‘Did Napoleon hear about Wellington’s death?’ the computer would know that the answer would be ‘no,'” Lenat explained. “By the end of the day we had
several hundred concepts mapped out. By the end of the week, we had several thousand pieces of information. We look at novels, newspapers and magazines – not for content, but to find out what the writer assumed the reader knew about the content.”
John Lam, a writer and developer of training classes for Wintellect in Toronto is
unsure of how eager he is to converse with his desktop, regardless of its ability to reason.
“Voice recognition is good for certain things – it’s good if we can’t use our hands for some reason – but I don’t really want to talk to a computer all day long,” Lam said. “It seems like these are just solutions looking for problems.”
Not so, says Lenat, who cites the Internet as a prime example of how common sense would make a world of difference in the interaction between a computer and its user.
“When you’re searching for something on the Web, why does it make horrible mistakes? Because it doesn’t understand the content of the documents,” Lenat said. “If you’re looking for articles on George Bush it will pull up a whole bunch of articles on George Bush and some articles that say ‘We are not going to talk about George Bush in this article’ because it doesn’t know ‘not.'”
“Daisy, Daisy, give me your answer, do.
I’m half crazy all for the love of you.”
Anthony Vranic, a solutions architect at Thinkpath Inc. in Toronto is worried about the philosophical consequences of this kind of artificial intelligence.
“As we build machines and systems and the entire social fabric in a more automated way, we will start relying more and more on automated decisions and forget how we got to that decision,” Vranic said. “Think of our parents. When they were in school they used slide rules and pen and paper because they didn’t have calculators, and to this day they have a better grasp of basic arithmetic functions than most people in my generation. I strongly believe that with increased automation, we’re getting away from what’s really going on.” He said it’s scary to think that before we even get to HAL, we’re going to become dependent on automation.
Shainblum asserted other philosophical issues around creating a real-life HAL.
“I’m not sure that true artificial intelligence is desirable. HAL was a sentient, thinking being, and with that comes the question of whether we have the right to enslave such a being. There are a number of ethical, moral and social issues surrounding a new form of life such as a self-aware, sentient computer. I’m not sure if it’s paranoia or not, but maybe we’re meddling in things that we shouldn’t,” Shainblum said.
As to whether common sense is the missing link to the creation of HAL, Shainblum is skeptical.
“Common sense has led humanity so astray for thousands of years. Aristotle came up with all of these common sense solutions that were totally wrong because common sense doesn’t always work in every realm,” he said.
Lenat, however, is so confident that common sense is indeed the key to realizing HAL, that he is willing to go on record with a timeframe.
“I’d be surprised if it’s less than five or more than 10 years away,” Lenat said. “All the pieces of HAL’s functionality are around in a rudimentary form, but common sense is the bottleneck – it’s the critical piece.”