That man, Arthur C. Clarke, saw great things for mankind beyond the Y2K mark. He envisaged that within the next year humans would have embarked upon far-flung space travel guided by a Heuristically programmed ALgorithmic (HAL), a machine that can reproduce "most of the activities of the human brain, and with far greater speed and reliability." Neither of his imaginings is close to becoming a reality. In truth, artificial intelligence has had its ups and downs since the 1968 release of 2001: A Space Odyssey. Computers can beat humans at the computationally intensive task of playing chess. But HAL's coherent replies to a range of human questions in "perfect idiomatic English" still eludes the best that computer science has to offer.
Despite a decidedly mixed track record, optimism about the prospects for machine intelligence still reigns at Carnegie-Mellon University, a longtime AI bastion. To mark the opening of a new computer center that bears the name of AI pioneers Alan Newell and Herbert Simon, the university held a one-day conference on October 19th that brought together experts from inside and outside the university (including Arthur C. Clarke, if only in video presence). Their mission: to answer the question of whether computers would help or hinder the building of a good world in the year 2050.
The assembled panel of experts was mostly true believers. If the CMU pundits are right, HAL will have arrived in 50 years' time: computers will reach or surpass humans in their reasoning ability. The main debating point centered not on whether we are capable of gestating machines smarter than us but whether those machines will do our bidding or destroy us instead. Will the movie we live in be The Jetsons or Terminator 2? No one entertained the idea that the truly intelligent cyborg might simply prove a bust. Remember that Nobelist Herbert Simon, a speaker at the conference, predicted in 1965 that machines would be capable of doing any work humans can do by 1985. (A graduate student could probably still get a doctorate from CMU for designing a robot to do something as simple as climbing stairs without tripping or stacking plates and cutlery in a dishwasher, an assignment that might be evaluated by how few crystal goblets were broken when loading the machine.)
CMUs artificial intelligentsia foresee accelerated evolution of cybersmarts. Author and speech recognition innovator Ray Kurzweil told a sympathetic audience that we have arrived at the age of exponentials. Every year, Kurzweil says, we experience exponential growth in the rate of exponential growth for computer and telecommunications technology. But exponentials, he notes, also apply to brain scanning, genome sequencing, the Internet and even human life spans. By 2030, he suggests, machines will be as intelligent as a human. In 50 years, as reverse engineering of the mind continues, we will be able to achieve full-immersion virtual reality.
Forget helmets and data gloves. Nanobots, robots the size of a molecule, will travel through the bloodstream of your brain beaming messages to neurons that will enable the simulation of sight, sound, smell and hearing as well as emotion and sexual sensations. Youll also be able to travel to St. Barths, attend every game of the World Series or engage Al Gore in a debate. This goes way beyond an accurate graphical rendition of the cherry in the tropical rum punch on the beach in St. Barths. The sense of self will become infinitely fluid at the same time as the concept of personal space will take on a much more literal meaning. What kind of postmodern succotash emerges when you trade my id for your ego? Dr. Kurzweil, meet Dr. Derrida. Kurzweil did not speculate on the prospects of this development for a Nike or a Budweiser. Imagine an inner voice--"Whaaassup"--exploding in your head just as youre about to doze off. Or what if Philip Morris could tweak remotely a nicotine receptor in your brain?
In contrast to this vision of reality as television, Sun Microsystems co-founder and chief scientist Bill Joy depicted a dystopian world in which the hubris of the human race will engender nonprotein life-forms that will multiply and threaten the planet. "Do we want to make a robotic intelligence that exceeds us?" he asked. Joy renewed his call, stated earlier this year in a Wired magazine article, for checks on unhindered development and dispersion of potentially harmful technology: "Just as with nuclear weapons, we wouldnt want everyone to have a nuclear weapon in their garage when they have a bad day." Joy spoke in the format of a "synthetic interview," a taped digital interview that could then be annotated and queried with questions typed into a computer.
Hans Moravec, founder of CMUs mobile robot laboratory and someone who has written of how our brains may one day be downloaded into a computer, traced a genealogy of robots, originating with the creation in 1950 of Elsie, the work of British biologist W. Grey Walter, who built it to have the IQ of a bacterium. Elsie would move in response to light, including beams that it would emit from its own tortoise-like body. When placed before a mirror, it would dance in front of its own image. Later robots, such as the 1990s vintage Xavier, could build an internal map of their surroundings and navigate down a hallway, though Xavier flustered easily. It might, for instance, get stuck under the hanging leaf of a potted plant. "There was a significantly high probability that the machine gets confused about where it was," Moravec said. "Typically, once that happens its very hard to recover, so the robot became lost." Moravec hasn't lost confidence in the rapid pace of robotic evolution: home robots will arrive by 2010. Autonomous vacuum cleaning may be one thing. But let your kids load the dishwasher, if you value the integrity of grandma's crystal.
David Gelernter, a professor of computer science at Yale University and a parallel programming specialist, asserted the primacy of technology in human behavior. "Fish swim; birds fly; we invent technology," said this victim of anti-technology zealot Theodore Kaczynski. Gelernters message, though, was not unambiguous. Similes for man and machine are often misguided. Mind is not to brain as software is to a computer. Software can be ported from one computer to another. Joe Schwartzs mind cannot run on Melissa Feinsteins brain. And machine intelligence will never be achievable until a computer can mimic the full cognitive spectrum, including emotion, dreams and the nuances of touch and sensation. "Youll never get a computer to think until until you've figured out how to get it to hallucinate," he says.
Raj Reddy, founder of CMUs Robotics Institute, began by saying that none of the predictions of Joy, Kurzweil or Moravec will materialize. "I dont see that society is willing to invest billions of dollars to create robots that exceed human capabilities." Reddys vision is less invasive than Kurzweils but no less utopian. Given the increasing capabilities of storage and bandwidth, people will, in a metaphorical sense, engage in teleportation and time travel, while achieving an immortality of sorts. A few petabytes of storage (quadrillions of bytes) will allow a record of every utterance and every movement from birth until ones last gasp for "less than the price of a pizza."
Petabit communication pipes and perhaps thousands of high-definition cameras will enable someone to manipulate a "soft camera" that will elicit a view from thousands of angles throughout a stadium dome or from down on the field. "This will let you watch the Super Bowl from the vantage point of the quarterback," Reddy says, deeming this ability the functional equivalent of teleportation. The lifetime video and audio transcript can be annotated and queried (as was the Bill Joy interview), bequeathing an intimate legacy for future generations. "Youll be able to converse with your great, great grandchildren," an experience Reddy equates with a form of time travel. Virtual immortality will come if the petabyte storage capability provides a "rapid simulated learning environment" that infuses your biological clone with the totality of your experiences. Maybe mind is portable after all.
Herbert Simon, the polymath CMU professor who received a Nobel Prize in Economics, enjoined that the human mission is to make the future, not predict it. The villain in society is more often the human being and not the misbehaving computer. "It shouldnt be 'Cherchez lhomme' but 'Cherchez lordinateur.'" As far as thinking machines, robotic soccer matches played on the CMU campus represent the present pinnacle of artificial intelligence, requiring machines to use sensory and motor capabilities to coordinate movement of players and devise strategic responses to the actions of the opposite team, says Simon. One day robots will triumph over humans in a 90-minute match, he predicted, a feat that should not preoccupy those with an anthrocentric bent. "We havent stopped playing chess," he said of Deep Blues 1998 victory against Garry Kasparov.
In a taped interview from Sri Lanka, Arthur C. Clarke agreed with those who imagine that virtual reality will become a surrogate for life as we know it. People may neglect the material world--with its myriad of woes ranging from Mideast strife to global warming to killer asteroids--for an existence that embraces the allure of simulacra. "Virtual reality, which is going to be the next big thing, will be so much more attractive than real life that people just won't make the effort; people will just sit back and enjoy the inputs into their brain," Clarke observed. "That could well be the end of the human race." Aliens watching terrestrial broadcast television signals leaking into outer space may see the world's billions perishing from what they will come to describe as the Couch Potato Famine. If humans become mere spectators, will elan de la machine rob us of elan vital? At the moment, neither Nobel prize winner nor microprocessing grandmaster can answer with authority.r