7

When IBM Watson was enlisted in the battle against cancer, it needed to learn about the many different types of cancer, their symptoms and treatments. It also had to learn to distinguish various cancers from other diseases with similar symptoms. Before suggesting a course of treatment, Watson had to be able to take into account side effects that can vary from one patient to another. All this learning requires the guidance of human experts. They are key in developing a corpus of knowledge. Assembling the corpora starts with digesting masses of information from every available and relevant source. The real building then begins with domain experts assisting in evaluating the information and discarding any items that are out of date, irrelevant to the problems that are to be addressed, or held in poor regard by the domain’s peers. Next comes the training phase, in which the computer learns how to interpret information in a process known as machine learning. With supervised cognitive systems like Watson, experts will upload training data in the form of questions and answers. The purpose isn’t for the system to memorize answers but to uncover the underlying patterns of thought that lead to answers. These patterns tend to be unique to a particular domain or industry. The system continues to learn once it begins interacting with end users. There are regular updates of information and periodic reviews by experts. Over time, cognitive computing systems grow more capable of responding to complex situations. Users receive quick answers with an array of recommended choices backed by evidence. They even get help uncovering new insights and patterns that had been hidden in the mass of information the cognitive system has been fed. 7