The computing system that won 'Jeopardy!' is helping doc
Source: IBM
IBM Watson famously hit the big time when it beat two "Jeopardy!" grand champions for a million-dollar grand prize in 2011.
That victory brought the cognitive system a lot of publicity, but it doesn't tell the whole story. "'Jeopardy!' was only the beginning for us," says Dario Gil, director of symbiotic cognitive systems at IBM.
Last year, IBM announced a projected $1 billion investment in the Watson platform, which is available in dozens of occupational areas. The system is designed to help its users make better choices based on real information. In oil and gas exploration, it helps companies decide where to drill. It can assist lawyers in building legal arguments and police detectives in solving cold cases.
And now Watson is going into medicine. That's important, because the third-leading cause of death is medical mistakes borne out of poor decision-making.
As Gil explains, "Watson is explicitly designed around scaling and enhancing the expertise of human beings. It's a cognitive system designed around us." What he means is that Watson isn't an example of "artificial intelligence" in that it doesn't make decisions on its own. "It's more like a tool," Gil continues, "in the same way any machine is a tool that helps us to work better. That's the unique thing about cognitive computing."
With Watson for Oncology (developed with Memorial Sloan Kettering) doctors can use the platform ― once a behemoth the size of a bedroom, now slimmed down to roughly the size of three stacked pizza boxes ― to wade through mountains of data. The system asks questions and makes hypotheses based on the latest medical research, along with the patient's symptoms and electronic medical records.
Rob Merkel, the healthcare and life-sciences leader at the IBM Watson Group, says, "Imagine Watson reading through an entire patient history, all of the information, all of the notes, and then being able to help determine what’s clinically relevant for that particular patient. It doesn't just summarize or regurgitate information that already exists. It actually provides insights." In a way, Merkel notes, it's like "CliffsNotes for a physician."
And how about normal human prejudice and mistakes? Could something like that cause the system to falter? Merkel points to a solution called WatsonPaths. "The normal human brain, when you’re presented with a question, gravitates towards a primary hypothesis," he says. "If you’re like me, and most people are, you exhaust all efforts to prove you’re right. Watson explores every potential answer at the same time, though." Watson is designed, then, to avoid typical human errors.
The system also learns over time. According to Gil, Watson grows smarter in three ways. First, it gets taught by its users. Leading oncologists at Memorial Sloan Kettering worked with IBM to teach the system how top doctors work. The system learns from interactions with patients.
On top of that, Watson accesses the mountains of data out there. Ninety percent of the data in the world was created in the past two years alone, which means there's a great deal of knowledge that can go unnoticed. To keep abreast of the latest medical data, Watson has ingested all 23 million medical papers in the National Library of Medicine (MEDLINE), and can access that information in milliseconds.
As Gil says, Watson is "designed around people. It really is intended to help us." Merkel adds, "I can tell you without hesitating, the totality of what we’ve already accomplished around oncology and what we’ll continue to do is completely amazing." IBM Watson sure has come a long way from its "Jeopardy!" days.
| }
|