It’s been two years since IBM’s Watson computer beat “Jeopardy!” champions Ken Jennings and Brad Rutter on the trivia game show.
Since then, IBM created a commercial division devoted to Watson and opened the platform to partners in health care and academia, giving leading research institutions access to the system’s natural language processing, analytics and machine learning technologies.
In return, Watson will learn from experts how to understand questions, recognize trends and recommend answers in new fields of knowledge. By learning from oncologists at Memorial Sloan-Kettering Cancer Center, caregivers at WellPoint Inc., and computer science researchers at Rensselaer Polytechnic Institute (RPI), the cognitive computing system will be able to use its ability to answer questions in context to assist professionals in those fields.
“We’re going to go and partner with those in the industries who know the problems better than anybody else,” said Stephen Gold, director of worldwide marketing for IBM Watson solutions. “We made a great decision, and realized that we’re not the experts.”
Gold said since the division to turn Watson into a commercial product was founded in August 2011, IBM has managed to make the machine 75 percent smaller, while increasing its computing power by 240 percent.
There is still a huge amount of work required from IBM’s industry partners to turn Watson into a business machine. It has to learn language used in health care and scientific research settings, for example, and understand the context of how that language is used. Training Watson to be a tireless assistant starts with teaching it the terms of the trade, and training it to correlate simple questions with the right answers.
For oncology researchers, the goal is to have Watson scan historical cancer cases, assess patient treatment options and then make the same determinations for treatment that the cancer center’s doctors would have made.
According to IBM, Watson has ingested more than 600,000 pieces of medical evidence and 2 million pages of text from 42 medical journals and clinical trials in the area of cancer research.
Ari Caroline, the director of Memorial Sloan-Kettering’s quantitative analysis and strategic initiatives group, and his team have taken on tutoring Watson for the last several months. Caroline described the process as “incredibly exciting or very frustrating,” depending on the day.
Caroline said many people would consider a system that could just replicate the expertise of Memorial Sloan-Kettering’s physicians after a user hand enters case specifics to be useful. But in the view of his analytics team, Watson is capable of much more. “We’ve set our sights much higher than that,” Caroline said.
Caroline said the system is not ready for prime time yet. But the progress is real, and soon Watson will help guide oncologists for treatment options and new research avenues.
“It’s definitely a time and effort thing,” Caroline said. “There is no question that this is an achievable task.”
IBM’s other healthcare partner is WellPoint, the largest health care provider in the United States. WellPoint is working on using Watson for utilization management; the machine helps the review process between a patient’s physician and their health plan.
The system seeks to streamline the treatment approval process and help doctors choose the right treatment plan quickly based on current symptoms and patient history.
Computer Science Research at RPI
IBM also recently announced that the computer science department at RPI would have access to Watson’s object code, exposing the technical details of the 41 different systems that make Watson work. IBM is giving the university an opportunity to train Watson to help in the field of computer science research, but also examine it as a research subject into artificial intelligence and cognitive computing.
Jim Hendler, the head of RPI’s computer science department, said the opportunities are endless for fields of study using Watson’s brain power.
“It’s like being the first university to get a telescope,” Hendler said. “You don’t really know what you’re going to see when you point it at things. You just know there is tons of stuff to be discovered.”
Hendler said the university has worked with IBM on Watson for several months; a number of researchers on the original Watson team were RPI graduates, so the school has long had a connection to the project.
There are already three graduate level research projects lined up to help Watson grow. RPI will help Watson become better at pulling data from traditional databases instead of simply scanning language from documents. The university will also add computational abilities to its cognitive strengths.
“It doesn’t currently do numerical reasoning,” Hendler said. “You can’t ask it which two numbers are higher.”
The third project is to give Watson the ability to have a dialog. On “Jeopardy!” it was not important for Watson to remember what quiz show problem it just answered. If a question is similar to the previous one, it just starts the process all over.
“We want to use it in intelligent tutoring,” Hendler said. “Can Watson teach based on having all that knowledge? To do that it has to keep track of what it’s been asked, what has it answered, what it wants to teach next, that kind of thing.”
Hendler said there is already a huge demand for some face time with Watson; he envisions a contest for undergraduate computer science students where they pitch projects to put Watson’s thinking power to use.
“We’ll teach some of the students how Watson works, what kind of projects might work, and then use their creativity to help us figure out some exciting things to do with it,” he said.
What’s Next for Watson
Gold said Watson still has learning to do, especially in use cases outside of health care, but the investment in time for training Watson is quickly repaid by putting people to work in positions other than data entry and data management.
“It’s a one-time investment to really get to a very different place and capability for the business,” Gold said. “When you look at data-intensive industries, a fair amount of labor is going into the management of that information. What Watson represents is an opportunity to displace or augment that traditional process.”
Gold said IBM is now working with industry partners to bring Watson to the financial services industry, and testing other business use cases. The system can be run in a data warehouse or in a cloud environment.
That means Watson will be available to different size businesses and a managed service.
“It will be available to big business, it will be available to small and medium business, and ultimately it will be available on a relatively ubiquitous fashion to everybody,” Gold said.
Email Staff Writer Ian B. Murphy at firstname.lastname@example.org. Follow him on Twitter .