Jeopardy-winning supercomputer’s next mission: save lives

Jeopardy’s most winning human opponents, Ken Jennings (left) and Brad Rutter, were no match for IBM's supercomputer during a three-day man vs. machine competition. Watson, the computer, earned $77,147 compared to Jennings ($24,000) and Rutter ($21,600). (Photo courtesy of IBM)

At first, it sounded like something right out of Terminator. Watson, IBM’s supercomputer, was programmed to annihilate two of humanities brightest in a complex game of trivia – it did so quite handily.

But that’s where the computers-will-rise-to-crush-humanity comparison ends.

After toppling Ken Jennings and Brad Rutter on a recent episode of Jeopardy, Watson’s new directive is to ultimately save time and human lives in the process.

“Look at the amount of data we deal with,” says Carrie Bendzsa, IBM Canada’s manager of communications and a Carleton grad. “The amount just keeps growing and growing exponentially.” The total worldwide data volume is expected to reach 35,000 exabytes in 2020, compared to 1,200 exabytes in 2010.

No human being could ever remember everything and it would take a single human being an incredible amount of time to dig through it all.

That’s where Watson comes in. The supercomputer’s data harvesting abilities are unmatched. When asked a question, for example, the mighty computer can pore through the equivalent of about 200 million pages of information to retrieve an answer in just three seconds. This same task might take a single computer two hours to process. By contrast, the same amount of work could be done by about 2,880 brains.

Bendzsa says this makes Watson the perfect “collaborative tool” in the healthcare industry, helping doctors make decisions faster and more accurately diagnose patients by cross-referencing information like patient records and medical journals – more information than a single person could ever learn – in a matter of seconds.

“It could help guide a cardiologist to avoid common mistakes,” says Bendzsa, pointing out that the computer could be used to ingest and analyze treatment data and test results – looking for causes of concern like the overuse of diuretics.

Andrew Low, a Carleton computer sciences grad and one of the engineers who worked on Watson’s virtual machine, gives the example of a doctor treating a patient.

“(They’re trying to) give the patient the best advice they can,” he says. “But how do you get enough information to make the right decision?”

Watson could scour thousands of medical journals, he says, and tell the doctor what information is relevant. Unlike a search engine, it’s not just “looking at link popularity . . . it’s about understanding the information.”

Since Watson can analyze and interpret language in all its complicated glory – able to grasp riddles, subtle clues and context – it even has the potential to help patients diagnose their own problems with online self-service help desks.

“A lot of people say ‘How is Watson different than a search engine?’” says Low. “It can give you an answer, but also give you the reasoning for why. This didn’t really come across on (Jeopardy) but it is able to describe the path it took (to get the answer).”

Even though it generally dominated the competition on Jeopardy, Watson baffled its developers on the second day of the competition. During the Final Jeopardy round, Watson asked “What is Toronto???” when given the answer “Its largest airport was named for a World War II hero; its second largest, for a World War II battle.”

The correct question, of course, is “What is Chicago?”

If Watson has a Canadian bias, it shouldn’t come as much of a surprise. Key parts of Watson’s “brain” were developed in Ottawa, tracing back to work done by former Carleton Prof. Dave Thomas’ tech startup company Object Technology International, which is now part of IBM. Many other Carleton alumni worked on the project including Low, John Duimovich, Marcellus Mindel, Timothy Mok, Anissa Shaddy, Matt Bisson, Mike Hay and Rohi Rishi.

This entry was written by Daniel Reid and posted in the issue. Tags applied to this article are: , , . Leave a comment, bookmark the permalink or share the following short URL for this article via social media:

Daniel Reid

By Daniel Reid

Whether it’s scientific breakthroughs, political manoeuvres or loaded technical jargon, Daniel Reid loves to untangle complex ideas to make them accessible to everyone. He is currently an editor at @newsrooms and is a former web editor at @CTVNews and homepage editor at @TheLoopCA. You can argue with him on Twitter at @ahatrack.

Be a part of the Carleton Now community

Carleton Now strives to be an inclusive, relevant and informative publication focused on building and fostering an engaged campus community. You can be a part of our community by: sharing or voting for this article (below), joining in the conversation, or by sending a submission/letter to the editor.

Comments are closed.

Current issue