In January 2011, an IBM computer named Watson beat out experts on the TV quiz show Jeopardy!, marking a groundbreaking development in artificial intelligence. Four years later, Watson is at the forefront of cognitive computing with important implications for the future of healthcare.
To understand Watson, it’s important to put it in context. For years technologists have produced question-answering systems, but until recently their capabilities were limited to simply phrased questions. What makes language so hard for computers is that it’s full of intended meaning. Every sentence has allusions and connotations that humans can unpack but computers interpret literally. Watson is able to understand complex language by using algorithms based on statistical interpretations of huge sets of documents, thereby determining which words are associated with a given subject and essentially learning the meaning of words and phrases in context (if that isn’t enough of a mouthful check out IBM’s video How It Works). This process, called cognitive computing, makes Watson one of the first computers that can understand unstructured data like journal articles and blog posts, which is 80% of data today.
Although designed to win Jeopardy!, Watson has evolved into array of commercial interfaces that allow experts across industries to harness its analytical power. Watson isn’t a software program or a physical supercomputer sitting in a room somewhere; when IBM talks about “Watson,” they’re talking about a suite of computing capabilities that work together to derive answers from large, unstructured sets of data in the English language. While a given query can be performed from a desktop computer or mobile tablet, the computing power necessary to perform that function is located in the cloud, on IBM’s servers.
Watson’s ability to answer questions about unstructured data makes it a promising tool for medical research and health care provision. According to IBM senior vice president Mike Rhodin, medical information doubles every three years. As we discussed in an earlier post, many medical fields are experiencing an impending doctor shortage, making effective and efficient care even more important than in the past. Watson’s primary contribution is to narrow down information for physicians and researchers using high-speed analysis of a given body of literature or set of medical records. There are already several examples of Watson being used for this purpose.
At Baylor College of Medicine researchers used Watson to search through 70,000 articles about p53, a protein that can sometimes indicate cancer but is not yet well understood. Previous research efforts had returned roughly one therapeutic target a year, but by combing through the literature and suggesting proteins that control p53 activity, Watson allowed researchers to identify seven proteins within a few weeks that could lead to possible treatment.
At Memorial Sloan Kettering Cancer Center experts trained Watson to help oncologists identify treatment options. Watson analyzes a patient’s case attributes against a huge array of data, including physicians’ notes, lab results, cancer case histories, established treatment guidelines, and published research to provide supporting rationale for individualized treatment options.
Researchers are also exploring the role Watson could play in the quickly growing field of genomics. The New York Genome Center is working in partnership with IBM to have Watson analyze a patient’s DNA along with existing research to suggest treatments options for glioblastoma, an aggressive form of brain cancer. By searching relevant literature, Watson can determine which of an individual’s many genetic mutations might be relevant and which drugs treat those mutations – all in a fraction of the time it would take a physician to complete the same research.
While Watson’s capabilities are surpassing what anyone expected at the project’s outset, it is still just a powerful tool in the hands of knowledgeable experts. In 2013, IBM released two new Watson-based interfaces designed for medical experts. WatsonPaths enables the user to better understand the data sources and systems consulted and the path it took to offering an option, basically introducing more transparency for the user. Watson EMR Assist helps physicians to uncover key information from patients’ medical records, with the goal of improving the quality and efficiency of care.
However, developments in the past year suggest that we are just beginning to see how far Watson technologies could go. In August 2014 IBM upgraded its Watson Discovery Advisor (a commercial offshoot that packages some of its capabilities as a data analysis cloud service), making some say it can now answers questions before you ask them. Jokes aside, the service is designed to uncover new insights and patterns in large data sets, and it continues learning through training cycles. Keeping with the trend for cognitive computing, the new Discovery Advisor is also 240 percent faster than the previous version. Perhaps we are entering an era when computers will present us with new answers and insights locked away in volumes of data.