"Third kind of computer" learns cancer medicine
IBM's Watson computer is building its knowledge-base for its first 'serious' uses
By Stephen Bell, Singapore | Friday, 19 October, 2012
IBM’s Watson computer — best known for having competed successfully with human players in the TV quiz game Jeopardy — is building its knowledge-base for its first ‘serious’ uses, in the medical world and the finance industry. Both are of great potential benefit to humanity as well as promising markets in the billions of dollars to IBM, says Manoj Saxena, general manager of the group developing Watson.
The machine is “learning the way a child learns”, he says — by reading and by asking and answering questions of human experts — in restricted fields of medical and financial knowledge. The field chosen initially in medicine is breast and colon cancer.
Watson represents an attempt to imitate patterns of human thought and learning in hardware, software and stored information; something Saxena sees as a significant third thread of computer development.
First we had mechanical comptometers that performed only a restricted range of operations built into them, he says.
These were succeeded by programmable computers. The emerging generation of computers, Saxena says, will function like Watson does. However he acknowledges the emerging generation will not replace present-day computers as these replaced the comptometers; there will always be a role for computers programmed to perform particular tasks, he says.
There is little programming in the conventional sense involved in Watson, he says; beyond basic skeletons for the representation and interlinking of data and metadata, its ability to arrive at answers in human domains depends on a lot of reading — 7 million pages on cancer alone, plus much more on general knowledge and general medical terminology — and exchange of questions and answers with domain experts - in the medical case from the Memorial Sloan Kettering Cancer Center (www.mskcc.org) in New York.
“The hardware and software is like a car,” says Saxena. “It’s obviously not enough just to have a car; you need petrol — the stored information - and you need a driver — the expert who teaches the machine. I say [implementation] comes in three phases — build Watson, teach Watson, run Watson.”
The initial client for Watson’s medical application is US insurance company WellPoint. In the front of the queue on the financial services side is Citibank, a long-term partner of IBM.
And there is a queue of other potential customers, Saxena says. “I feel like I’m having to put my shoulder to the door sometimes, to keep out the crowds saying ‘we need a Watson’.”
Watson at present has knowledge in the cancer diagnosis and treatment field equivalent to a first-year medical student, he says. It still needs the years of practical experience that student will go through before being allowed to treat patients unaided. Not that Watson will ever be allowed to do that in Saxena’s vision.
“It will never replace the doctors; I see it as a decision support system for them.
“It’s important for us not to get carried away with the hype,” he says.
• Bell attended the IBM Interconnect conference in Singapore as a guest of IBM.