IBM's STG Is On Track With Smarter Computing
December 21, 2011
IBM describes Watson as a new class of analytical advisor and not an application per se. Watson is sophisticated software running on commercially available IBM Power 750 servers optimized for rapid, deep analytics of massively parallel problems. IBM uses hardware optimization, such as exploiting Power’s robust memory architecture (page sizes, multi-page) and cache sharing between cores.
IBM discussed healthcare efforts it is pursuing in conjunction with WellPoint where Watson takes patient information (such as medical records, medications, and tests), and reviews a large body of medical information (textbooks, articles, etc.) to provide an evidence-based analysis whose output is presented to physicians. Unlike its Jeopardy! performances, Watson does not come back with a single best answer, but may present a series of suggestions based on probabilities, not certainties. Moreover, Watson is not bound by the constraint that it has to answer a question in 3 seconds or less, an essential point, since queries may include a series of interrelated questions (such as those an oncologist or other specialist might pose). That Watson does not make decisions is a critical point. Watson merely complements the analysis of medical data and the medical decision-making process. Watson’s output can also lead to further action (such as additional medical tests whose results can help improve Watson’s ability to perform differential analyses).
This can lead to important results. Dr. Herbert Chase, physician guest of IBM who is a professor of clinical medicine, told a story. Early in his career, he was faced with a difficult case where a woman was suffering severe debilitation from what appeared to be a muscle disease, but specialists could not diagnose her condition so no treatment could be put in place. Access to literature in those days was difficult, but probably through luck and sheer determination, the physician found a non-intuitive and unlikely solution that turned out to be the answer.
He had had three clues that led to his successful diagnosis and used them to test Watson. Watson came back with a list of five or six possibilities, the second of which was the correct diagnosis. A simple standard blood test could have been used to rule out the first option. The treatment for the correct option was simple and straight forward but not one that the specialists would ever have thought of recommending. Think of the earlier improvement in the quality of life of the patient as well as the cost savings (the woman was in the hospital for weeks while various diagnostic tests were made). This is only one illustration of how valuable this approach can be.
Dr. Chase felt that doctors will accept the Watson-type technology. They simply do not have the time to review all the literature in their own specialties, let alone across other disciplines. He feels that medical education is going to have to change from memorization of information to how best to analyze information. So Watson is likely to have a major impact in many ways — saving lives, improving healthcare outcomes for better quality of life, cutting costs and enhancing medical education — among them.