So what do these three have in common?
- Thomas J. Watson Sr, CEO and founder of IBM (100 years old this year). Currently has a computer named after him.
- Alan Turing, mathematician and computer scientist (100 years old next year). Has a famous test named after him.
- Aurthur C. Clarke, scientist and writer (100 years old in 1917). Has a set of laws named after him (and is also the creator of the fictional HAL computer in 2001: A Space Odyssey).
Unless you have moved into a hut, deep in the Amazon rain forest you cannot have missed the publicity over IBM’s ‘Watson’ computer having competed in, and won, the American TV quiz show Jeopardy. I have to confess that until last week I’d not heard of Jeopardy, possibly because a) I’m not a fan of quizzes, b) I’m not American and c) I don’t watch that much television. To those as ignorant as me on these matters the unique thing about Jeopardy is that contestants are presented with clues in the form of answers, and must phrase their responses in the form of a question.
This, it turns out, is what makes this particular quiz such a hard nut for a computer to crack. The clues in the ‘question’ rely on subtle meanings, puns, and riddles; something humans excel at and computers do not. Unlike IBM’s previous game challenger Deep Blue, which defeated chess world champion Gary Kasparov, it’s not sufficient to rely on raw computing ‘brute force’ but this time the computer has to interpret meaning and the nuances of the human language. So has Watson achieved, met or passed the Turing test (which is basically a measure of whether computer can demonstrate intelligence)?
The answer is almost certainly ‘no’. Turing’s test is a measure of a machines ability to exhibit human intelligence. The test, as originally proposed by Turing was that a questioner should ask a series of questions of both a human being and a machine and see whether he can tell which is which through the answers they give. The idea being that if the two were indistinguishable then the machine and the human must both appear to be as intelligent as each other.
As far as I know Turing never stipulated any constraint on the range or type of questions that could be answered which leads us to the nub of the problem. Watson is supremely good at answering Jeopardy type questions just as Deep Blue was good at playing chess. However neither could do what the other does (at least as well). They have been programmed for that given task. Given that Watson is actually a cluster of POWER7 servers any suitably general purpose computer that could win at Jeopardy, play chess as well as exhibit the full range of human emotions and frailties that would be needed to fool a questioner would presumably occupy the area of several football pitches and consume the power of a small city.
That however misses the point completely. The ability of a computer to almost flawlessly answer a range of questions, phrased in a particular way on a range of different subject areas, blindingly fast has enormous potential in fields of medicine, law and other disciplines where questions based on a huge foundation of knowledge built up over decades need to be answered quickly (for example in accident and emergency where quick diagnoses may literally be a matter of life and death). This indeed is one of IBM’s Smarter Planet goals.
Which brings us to Clarke’s third law which states that “any sufficiently advanced technology is indistinguishable from magic”. This is surely something that is attributable to Watson. The other creation of Clarke of course is HAL the computer aboard the spaceship Discovery One on a trip to Saturn that becomes overwhelmed by guilt at having to keep secret the true nature of the spaceships mission and starts killing members of the crew. The point of Clarke’s story (or one of them) being that the downside to a computer that is indistinguishable from a human being is that the computer may also end up mimicking human frailties and weaknesses. Maybe it’s a good job Watson hasn’t passed Turing’s test then?