IBM’s Watson Fellow: ‘Jeopardy’ Strategy Wasn’t Just a Game but Template for Future Computing

IBM's Watson supercomputer's Jeopardy championship should only be the start of many new commercial opportunities, the company's VP for Emerging Technologies said.

Watson needs to be trained and Watson training is never complete, IBM Fellow Rod Smith told IBTimes.

Next quarter, the powerful supercomputer is scheduled to start its first commercial trial, assisting medical professionals at WellPoint, the Indianapolis-based health services provider.

While Watson won't be doing patient consultations, Smith told IBTimes in an interview in New York, there's no reason why a WellPoint patient visiting a doctor might not fill out a questionnaire about her condition before being seen. Answers could be keyed into Watson, which might suggest she's suffering from a particular problem, speeding the physician's diagnosis.

It's not artificial intelligence. It's statistics, Smith said. The IBM Fellow explained Watson now is being crammed with reams of medical data, health and disease information.

Smith, based in Research Triangle Park, N.C. where part of Watson resides (the other part is in Austin), said the whole Jeopardy strategy for using machine intelligence came from an epiphany colleague and IBM David Ferrucci had about five years ago.

There are lots of problems we have to model the brain, Smith said Ferrucci realized. But we don't know the brain really well. Rather than try to replicate it using older software and the specialty called artificial intelligence, the group decided to feed the computer with masses of data and information that would enable it to rapidly sift through it and come up with answers.

It's how you use the data, Smith told IBTimes, how it can be broken down into smaller pieces you can use and have a confidence factor in.

Indeed, Smith said the group of around 100 IBM researchers determined that the system of Jeopardy categories would be quite difficult to master, because material is actually more complexly stored.

That's one reason why Watson --- actually 90 IBM Power 750 servers using 15 trillion bytes of random-access memory and 2,880 processors --- flubbed a Final Jeopardy question last February when it defeated two human champions.

Watson answered Toronto when it should have said Chicago to the answer: Its largest airport is named for a World War II hero, its second largest for a World War II battle.

Watson might have done better had the answer been more specific, such as adding U.S. city but the computer simply couldn't replicate a human thought process.

Smith said he was glad to see other companies are using similar kinds of techniques, albeit in smaller ways. He praised Apple's introduction of the Siri personal assistant in its iPhone 4S launch with mundane queries like finding a good restaurant.

Apple was showing how the iPhone's built-in GPS connected with Yelp, the IBM Fellow said. What it was doing was connecting with a specialized server to obtain information.

In an earlier interview, Manoj Saxena, general manager of IBM Watson Solutions Group, told IBTimes last month that the health care sector was low-hanging fruit in an unstructured medical data market estimated to reach $15 billion by 2015.

Smith suggested that besides health care, IBM might next target Watson to the call center market, where it could interface with real consumers, as well as the financial services sector. He declined to provide details but suggested that the same kinds of programming and preparation would be used for specialization.

Thirty years ago, my investments looked better than they did today, a customer might tell a financial adviser with a tie to Watson. The computer might then instantly recall financial markets of 1981, examine the client's portfolio, adjust for inflation and bring it up to date with ease.

Then the consultant could make recommendations.

Meanwhile, IBM will be happy to allow for computer and electronics manufacturers to help Watson out. Tablets are a great way to fill in information, Smith said. IBM sold its PC lines to China's Lenovo Group, which has been selling its IdeaPad to compete with Apple's iPad3 and others.

But it's not likely customers at WellPoint will be using tablets at first for data entry, Smith told IBTimes. To start, only WellPoint employees will be working with the supercomputer.

So far, Armonk, N.Y.-based IBM hasn't reported how much it's spent on Watson. The company has reported spending $3.15 billion on research and development in the first half of 2011.

Nor is the company solo in the field of unstructured data. October has already seen major announcements by two major rivals.

Hewlett-Packard completed its $10.3 billion acquisition of Britain's Autonomy, a software specialist with many applications in the field. Through earlier acquisitions, HP, in Palo Alto, Calif., has obtained technologies from the old Digital Equipment and Tandem Computer that are used in its data center.

Oracle, meanwhile, rolled out a series of machines using its Exalytics Business Intelligence Machine built on networked servers from its Sun Microsystems unit. During its Oracle World forum, CEO Larry Ellison illustrated how the system provides speed of thought interactive visualization. Oracle is based in Redwood Shores, Calif.

In late Friday trading, IBM was at $182.73, only slightly below its July 52-week high of $185.63.