Man Vs AI: Can computer outsmart human?

Dr Yuk-ching Hon

15 Feb 2011

From 14 Feb, for three evenings IBM’s celebrated supercomputer Watson will stand against the Jeopardy champions Ken Jennings and Brad Rutter in the world’s first quiz competition between man and artificial intelligence.

Jeopardy is a long-running American general knowledge quiz show watched by millions of viewers. The grand prize is $1 million; second place wins $300,000; third place receives $200,000. Jenny and Rutter have pledged half of their winnings to charity and IBM will donate its entire prize, reported WIRED, the computing online magazine.

A demonstration round on 10 Feb shows Watson defeated Jennings and Rutter hands down! http://www.youtube.com/watch?v=WFR3lOm_xhE

IBM views the Jeopardy Challenge as a milestone in the development of artificial intelligence – a memorable point where computers and computer technology have approached humans. The company claims that they have created a computer system that has the ability to understand natural human language.

Named after the IBM founder Thomas J. Watson, the supercomputer was programmed by 25 IBM scientists over the last four years. Some 200 million pages of content – the equivalent of about one million books, movie scripts and encyclopaedias have been scanned into the system. The super system is powered by 10 racks of IBM POWER 750 servers running Linux, and uses 15 terabytes of RAM, 2880 processor cores and can operate at 80 teraflops. That means 80 trillion operations per second. So it is not surprising to hear that the machine is the size of 10 refrigerators!

But Watson will not shine just as a quiz contestant. IBM has great plans for it after the Jeopardy Challenge. "Now that we've taught Watson to play Jeopardy, we're exploring the best ways to use this kind of capability to solve real problems," announces IBM's Vice President of Research Katherine Frase. "We're looking into what more we need to do to make it work for businesses."

One proposed application is "Watson, M.D." – a concept that would use the question-and-answer system as a diagnostic assistant in doctors' offices. A doctor could ask Watson terminal questions, and the system could use both online resources as well as similar patients' medical records to help reach a diagnosis. Watson, M.D. could also be used to help find information about rare conditions that no one has ever come to that particular hospital for.

Further potential projects include a question-and-answer engine for call centres or an analytics tool for financial companies.

So if computers can win quiz contests, defeat chess grandmaster (IBM’s Deep Blue supercomputer beat Garry Kasparov, the world champion in 1997), is there anything left that computers can’t do? Do we have to worry that one day, scenes of I, Robot will come true?

Well, Roland White, a Sunday Times columnist wrote yesterday that we have nothing to worry about, at least in the short term. In his book, What computers Still Can’t do, Herbert Dreyfus, a professor of philosophy at the University of California, Berkeley argues that computers can never match human intelligence because there are human knowledge that relies on unconscious instinct and not just analysis of facts and figures.

For example, a computer could probably string together enough phrases and sentences to generate a story but it could never produce a moving, original piece of prose based on its own experience.

One day you might get a computer to produce a satisfactory painting but could you get another one to review the artwork’s meaning and value?

And what about robots mimicking human? While the Japanese National Institute of Advanced Industrial Science and Technology has been making amazing progress in its robotic development and we have witnessed the remarkable singing and dancing performed by the cybernetic human robot HRP- 4C, http://www.crunchgear.com/2010/10/19/videos-cybernetic-human-robot-hrp-4c-dancing/ but will there be a robotic Yao Ming or Li Na whose instinct makes them great in their sports?

Most of all, it will be a long time before scientists can teach robots to acquire a full range of human emotions. Of course, we are excited by the extraordinary feat of Nao, created by Aldebaran Robotics. This cute, little humanoid robot is learning how to detect, respond to human emotional cue and simulate human emotions under the direction of University of Hertfordshire’s Adaptive Systems Group. Nevertheless, today’s robot still have a long way to go before it could weep reading a poignant poem or feel a surge of emotion at the sight of a new born baby.

So, fear not. Just remember who invent the computer.