Interview: Using IBM Watson for smart decisions

In this video, Dan Cerruti, who is responsible for Watson commercialisation at IBM, speaks to Computer Weekly about why Watson is more than a game show winner.

Two years ago, IBM astonished the world with a seemingly intelligent computer called Watson that beat human competitors on US game show Jeopardy. This was a prototype – a one-off machine. 

The technology has now been commercialised, and IBM has begun to work with organisations to build applications based on the technology.

“We don’t think about Watson in terms of intelligence, in the same way you would think about a person. We did not set out to build a system that was smart,” says Cerruti.

Watson is about understanding human language, he says. 

There are many computationally complex problems, such as weather forecasting, that use a lot of data, but such problems are not textural data, which makes them unsuitable for Watson. “We’re looking for problems where humans make decisions,” explains Cerruti.

But unlike traditional decision support systems, which are keyword-driven, Watson is driven by an inquiry. It then draws on the wealth of information it has gathered, and returns a set of matches based on the probability that the result is the correct answer.

In terms of programming, Watson is a non-deterministic machine – its innards work on confidence levels, rather than executing a set of program instructions, step by step, as is the case with a traditional, deterministic algorithm.

Cerruti says Watson is helping doctors and nurses gain a deep understanding of a patient. It matches all the textural data on a patient against treatments for cancer. It has to learn the correct answer, which it then uses when given a new inquiry. 

“Watson is an advisor to help a doctor figure out what might be the most suitable treatment,” he says.

View All Videos