How Watson, descendent of Big Blue, is learning medicine instead of chess

Alex Wong/Getty Images

The smart computer, an artificial intelligence ticking away in the background until awakened, inevitably goes feral in science fiction blockbusters, causing big problems for the human race.

But in the real world, computers have increasingly taken on processing and analysing data too large and too complex to be done by human minds.

Much of the groundwork for this started with Big Blue, the IBM computer which in 1997 beat grand chess master Garry Kasparov over a six-game match.

It was the first time a machine had beaten a human master. But, with all respect to Big Blue, that particular electronic marvel wasn’t that smart.

Dale Potter at IBM’s Watson Group says Big Blue was where cognitive computing really started, at least as a dream, an aspiration, a vision for the future.

“As powerful as that story is … COMPUTER BEATS GARY … the computer couldn’t do anything but play chess,” Potter, IBM’s Partner, Healthcare Transformation, says.

The next step was the build a computer to emulate the way a human being thinks. The thought process goes something like this: analysing information, or curating it, pulling it together, then reasoning around that information. This leads to a hypotheses which is then tested before producing a conclusion.

“That’s generally how a human being thinks,” Potter says. “So we started down that path in research.”

It was a big project. First they built a platform which could understand the English language, a complex system in itself. IBM gathered a group of linguists to teach the computer how to read.

The platform was called Watson, not after Dr Watson, the sidekick to the fictional detective Sherlock Holmes, but after the Watson family, the founders of the IBM company.

That was just the start.

“Watson at the base when you first install it is like a child,” he says. “It has the ability to understand your language and it has the ability to learn and that’s really it. Then you have to teach it something.”

Adam Berry/Getty Images

The computer had to be taught what the researchers called ground truths. An example that might be the concept of time, which is different to different people depending on the direction you look at it from.

“My 18-month-old has no comprehension really of time .. . it’s right now,” Potter says. “My three-year-old is now at an age when he is starting to understand time and agonise about it because everything takes too long.”

So a single concept can have different meaning. And the teaching continues.

“Watson knows how to read English and understanding what it is reading,” Potter says. “Now we needed to get Watson to emulate a profession.

“So we sent Watson to medical school.”

The computer needed to know that the sedative valium has 147 different names and that there are many different cancers and that these tumours come in many sizes and grow at different rates.

“There is a whole bunch of parameters for cancer that a physician would take into account to decide what the appropriate treatment plan is,” he says. “So Watson has to know these and determine what is meaningful about those parameters.”

ChinaFotoPress/ChinaFotoPress via Getty Images

Watson now has more than 12 million pages of text on file, and growing. There are more than 200 medical text books on different topics and then there are national cancer care guidelines from the US. Last year doctors spent 5000 physician hours with Watson.

When people read they apply filters to decide whether the literature is meaningful to them or any good.

Watson has been given more than 50 filtering criteria including the credibility of the journal the information is published in.

When Watson looks at a patient, it starts filtering, starting with the sex and age, size of the tumour, type of cancer, a holistic approach. And for this patient, what is the relevant literature which is related to all the attributes of this patient and this type of cancer?

For a doctor, it’s very difficult to keep with with the latest medical advances.

Information is being created and collected so quickly that about 90% of the information in the world was created in the last two years.

Medical literature is at least doubling every seven years. “So what’s happening in this particular area of expertise is very difficult for any physician to keep up with the newest literature and evidence,” Potter says. “They tend to be three to five years behind.”

So what Watson is doing is bringing a very up-to-date outlook and checking everything. “It ensures that all the relevant evidence has been presented,” Potter says. “Doctors tend to focus on the cancer because that’s what they’re there to do. Whereas, there’s an opportunity to focus more on the health management of the individual.”

Doctors don’t have to read a mountain of journal papers to stay on top of the latest advances. Watson will do that for them.

“It’s a very powerful enabler for them to be able to provide the very best care in an evidence based way, personalised to a particular patient,” says Potter. “That’s really what Watson does.”

What drives the process is a framework which like the one used by people to make decisions: Observe, Interpret, Evaluate, and Decide.

And it takes Watson just seconds to search through millions of documents, pulling together all the relevant information, interpreting, evaluating and then making a decision.

The reality behind Watson is a powerful computing platform. Many think it’s a big monolithic type of computer, along the lines of HAL 900 in the movie 2001: A Space Odyssey, but the actual computing platform is not much bigger than a pizza box. But it is very powerful.

Currently there are more than 300 Watsons in different data centres across the world.

Business Insider Emails & Alerts

Site highlights each day to your inbox.

Follow Business Insider Australia on Facebook, Twitter, LinkedIn, and Instagram.