In February 2011, IBM’s Watson computer was victorious over human trivia quiz champions on the game show Jeopardy. Does this mean computers can be smarter than humans? Not even close, according to Carnegie Mellon/IBM researcher Eric Nyberg. He says we should focus on augmenting, not overcoming, human capabilities.
*
Eric Nyberg leans over his desk, his nose almost touching the window of his office on the 6th floor of Carnegie Mellon University’s Language Technology Institute, housed within the angular and ultra-modern Hillman Center for Future-Generation Technologies.
Nyberg cannot stop himself from solving problems, even when problem solving is out of place. A bad situation brews on the street below: At least ten police cars surround a city bus, and as officers emerge, hands on their guns, to apprehend a suspect, Nyberg observes and comments with the kind of pragmatism one might expect from a man whose work in artificial intelligence just vanquished the two greatest Jeopardy champions of all time:
“It looks like the guy ran away. If you wanted to get away – not that I plan robberies – this is actually a pretty smart place to do it, because if you run from Morewood and jump over the fence, you can go right down into Junction Hollow. There’s no way you can get there from here unless you go back to Fifth Avenue and then over to Neville, and then go down.”
Nyberg, a tall, thin, bespectacled 49 year-old with a white pony tale, is a deeply creative man with problem-solving talents to spare. He excels at his hobbies more than most people excel at their careers. A talented musician, when he is not helping design computers that understand questions in natural language, he is designing and building electric guitar amplifiers — complete with “Nyberg” brand name and logo — that rival the very best in the business.
Nyberg’s focus on solutions explains why he was instrumental in the victory of Watson – IBM’s trivia supercomputer designed to win against human opponents on Jeopardy. But when asked if machine intelligence might some day overtake human intelligence, Nyberg immediately zeroes in on the problems:
“Watson is just a trivia-answering box. If I asked Watson, ‘Where can I catch the 54C bus to go home?’ he couldn’t answer.”
“Watson learns only when there are a lot of past examples to learn from. Thousands of Jeopardy games are available online, so if you want to learn how to play Jeopardy or teach a computer how to play Jeopardy, even though Jeopardy is a tough game to play, it’s actually not hard to try to build something.”
Watson, based on IBM’s DeepQA (Deep Question Answering) technology, is a powerful search engine that reads and interprets texts to find precise answers to questions asked in natural language. It is a quantum leap beyond search engines that deliver thousands of candidate answers, and is likely to impact fields such as medicine and intelligence very strongly. What’s more, it may soon be available to governments, hospitals and other organizations.
In an interview with Computerworld Magazine, IBM Senior Consultant Tony Pearson said DeepQA will be on the market within the next two years. According to Pearson, the 90 IBM Power 750 processors used for Watson’s televised Jeopardy match cost around $34,500 apiece, but the cost of slower systems would be much cheaper – as little as $300,000 for systems that require 30-seconds to deliver targeted answers, or much less for systems with much slower processing times:
“It’s quite possible [to wait two hours for an answer] if you run it on your Power 750 at home.”
DeepQA technology is useful in many applications, but Nyberg has found himself needing to explain to the public the limitations of machine intelligence, and why concerns that it may rival human intelligence are misplaced:
“I get reporters asking me questions like, ‘Is this the end of the human race?’ and I just tell them, ‘No, Watson will not be getting the launch codes any time soon. Watson thinks grasshoppers eat kosher.’”
“I call this the tyranny of anthropomorphism. Any time you stand up a computer agent that appears to have a dialogue and uses a human voice, and almost appears to have whimsy or emotion — even if it does only one thing right — you will completely overbuy. We’re not trying to oversell, but you’re overbuying. Come up with ten questions any human could answer and see what Watson’s answers are to those. Then you’ll realize how much we bought in.”
Nyberg does not see human-level artificial intelligence on the horizon, but he has high hopes for augmenting human intelligence. He refers to a science fiction novel by Neal Stephenson, entitled The Diamond Age.
The book describes a highly sophisticated device — an information companion that grows with children. The device learns about the child who owns it so that the more questions she asks, the more the answers are tailored to her development.
“This is science fiction, but it has interesting implications. If you think about a model like that, the person is not being invaded by technology. Rather, the technology is helping humans achieve their highest potential.”
Nybeg’s vision is ambitious, but he is not alone. The notion of augmenting human capabilities with technology is no longer confined to science fiction.
During the same month Watson was victorious on Jeopardy, the cover of Time Magazine read, “2045: The Year Man Becomes Immortal.” The photo on the cover was of a human with a computer cable plugged into the back of his head.
February 2011 also saw the release of a documentary film describing the life and work of noted futurist and inventor Ray Kurzweil, who was also the main subject of the Time article. The film, entitled Transcendent Man, follows Kurzweil’s life and career, and the movement he is at the center of. The movement, known as transhumanism, seeks to exponentially improve human capabilities through technology.
Transhumanists say that in the very near future we will use technology to direct our own evolution, replacing mutation and natural selection, and making humanity as we know it obsolete. According to them, the human genome will soon be as upgradable and transferable as software, and our interface with the tools we create will be so intimate that we will meld with our inventions and become more than human.
Some people do not think this would be a good thing. In the September 2004 issue of Foreign Policy, Francis Fukuyama, Johns Hopkins Professor of International Political Economy, called transhumanism “the world’s most dangerous idea” because, according to him, its first victim “might be equality.”
Fukuyama asks, “…. what rights will these enhanced creatures claim…. compared to those left behind? If some move ahead, can anyone afford not to follow?”
Nyberg is not as worried as Fukuyama, but still views Kurzweil’s cybernetic vision with caution. He supports the broad agenda of enhancing human abilities, but questions what he sees as the limited concept of technology promoted by most transhumanists:
“I think the discussion has been very narrowly focused on how to make people cooler, faster or smarter — like some kind of bionic man. Transhumanism should never be defined in terms of the technology or how you look or what you’re carrying around. It should be defined in terms of capabilities or behaviors, not in terms of people having things grafted on to them and plugged into the back of their brains.”
Nyberg also worries about improving technology without simultaneously improving morality, and wonders whether transhumanists are overly-focused on means when they should be focused on ends – ends that would make us not only more powerful, but also better:
“In order to be transhuman, you’d have to get to the point where people would be fighting to get the jobs that have the biggest opportunities for community service. That’s what transhumans would care about, not making the most money or having the most power.”
Technology may not be improving morality, but it has already augmented our abilities in measurable ways that would have been considered transhuman, at least in some respects, centuries or even decades ago.
For example, many say colorblind British artist Neil Harbisson, 28, is a cyborg because of the electronic equipment he wears in order to transform color into sound waves, allowing him to “hear” colors. Harbisson identifies with his “eyeborg” so much that in 2004 he successfully petitioned the British government so he could wear it in his passport photo. Hearing aids, pacemakers and other common devices may also qualify as early cybernetic advancements.
Nyberg even sees the dawning of transhumanism in things that are not attached to the body:
“I had a situation fairly early on with the World Wide Web where I was in Germany late at night. All the offices were closed, and I had no idea how to get the train to Stuttgart, but because I was a technologically savvy person, I just went to Google or Alta Vista or whatever the search engine was in those days, and had the right information at my fingertips. By conventional means I would have been at a loss.”
In addition to the technology of today and the technology of the future, Nyberg also suggests that we consider what role the “technologies” of the ancient past might play in expanding human potential in the transhuman age:
“While Kurzweil is talking about living forever through technology, and there are people designing drugs to make our minds go faster, there are other people who would laugh at this and say all we really need is to discipline ourselves, meditate and do yoga and so forth. You can make your mind go a lot faster this way — and you will achieve insights you would have never had otherwise.”
Nyberg is open to low-tech solutions because he approaches everything as a scientist. He is not interested in gadgetry for its own sake, but in solutions that can make life better and more peaceful.
On the street below, the police take their suspect into custody, and Nyberg shows his autographed copy of Kurzweil’s book, The Age of Intelligent Machines. The inside cover reads, simply, “To Eric, Welcome to the age of intelligent machines – Ray.” – but Nyberg is ambivalent. Echoing a minister’s criticisms of Kurzweil in Transcendent Man, Nyberg says, “This comment about how it’s our goal to find God, and not to become God, is maybe something we need to spend a little time on.”
Polished Writing 