Last Valentine's Day, instead of taking my wife out to dinner, I watched Watson, a computer built by a team of IBM Researchers, compete against past Jeopardy! champions on National TV.
With a team of about 25 researchers and engineers, we built Watson to advance the computer’s ability to understand natural language, the language humans use every day to capture and communicate their knowledge.
We fed it millions of pages of text from thousands of different sources so it could find and analyze the evidence in support of the right answers -- in just seconds.
Watson didn’t just find relevant documents, it analyzed language and produced answers with accurate confidences that enabled it to make good decisions about which clues it could answer and how much to bet.
While Watson winning "Jeopardy!" was exciting, we actually built it with far more practical applications in mind, like helping to tackle one of the world's biggest problems: health care.
The challenge is daunting as there are more than 12,000 diseases and some take years to diagnose. It's also important to find the right treatment that considers all the latest evidence, and that it is personalized for an individual’s specific patient history and condition. Such information is often buried in huge volumes of natural language content from doctor’s notes, in hundreds of massive reference books, or thousands of journal articles.
At times when health, well-being and lives may be on the line, emotions are charged, and patients and family members want answers. I know what it’s like to want answers, and I’ve seen Watson help find them.
Several years ago, when my father went into cardiac arrest on his 70th birthday I was asked to sign a Do Not Resuscitate (DNR) form at the hospital.
Whether to sign a Do Not Resuscitate form or not is a decision that millions of families across the country struggle with as family members become ill.
In my case, I refused. I felt that we hadn’t explored all the options for how to treat my father’s condition. I didn’t feel comfortable signing the form. I asked questions and ended up arguing with the chief cardiologist to get hard evidence for why I should sign the form.
I wasn't satisfied with the use of general statistics instead of specific and personalized evidence. My family members were scared for me and for my dad.
How could I argue with the chief cardiologist?
But without clear evidence my father was brain-dead and thoroughly exploring all options available to me, I felt I had no choice but to refuse to sign the DNR.
With my persistence, the medical team delved into my father’s specific history. Side-by-side, we manually reviewed his medical history and related research and made a series of crucial and very personal decisions.
Was a computer used or not? If not, why?
About 36 hours later, my father was sitting up in bed without any sign of brain damage. Smiling, he said, “So, I understand everyone thought I was dead.”
During the following years, he was treated for an ongoing heart condition that was constantly changing. Being well-informed was critical to treating his ailment and extending his life.
It was critical to have fast access to relevant knowledge, which was out there, but where? With access to data analytics, I believe, we would have had better answers and could have made more informed decisions.
When I think about the role of technology in my personal experience, I know that unfortunately, a surprising number of medical errors are made because a patient's disease is not diagnosed properly or fast enough -- or the wrong treatment is prescribed.
The use of IBM Watson to assist physicians could help to analyze patients’ data against the latest medical information, resulting in faster, more accurate diagnoses and better treatment. It could use its analytical capability to provide a doctor with different diagnostic possibilities for a patient's condition, recommend tests, suggest the most effective treatment and provide reminders about drug interactions -- right at the point of care. By analyzing as many as 200 million pages in seconds, it could provide the medical team with personalized recommendations.
The transition from the quiz show victory to physicians' offices will take some time, but with the complexity and the rapidly growing volume of information out there today, it’s an imperative that we advanced this technology forward.
We've made progress in the past year. WellPoint Inc., a large health care provider, is using Watson in a pilot. An oncology project involving Watson is also underway at Cedars-Sinai Medical Center in Los Angeles.
While Watson could never replace a doctor, it could serve as an invaluable tool for doctors to help deliver the best care – to help save lives.
David A. Ferrucci is an IBM Fellow