IBM’s intention for Watson was always to create an interactive system that would allow medical professionals to interact with artificial intelligence and communicate ideas and advice (“Watson Health,” 2018). However, before being able to apply it to Medicine, IBM needed to test it at a lower level. In the AI Magazine article, Building Watson: An overview of the DeepQA Project David Ferrucci and his colleagues defined Watson as a question-answering (QA) /machine learning technology that aids medical professionals in time restricted medical care decision making (Ferrucci et al., 2010). Machine learning is when a technology system is programmed to learn new information by itself, rather than needing a human to input the data. It is programming the computer to have its own “brain” by which it communicates with itself. Ferrucci and his colleagues determined the best way to test IBM Watson’s intelligence and accuracy was to make it compete in Jeopardy game (Ferrucci et al., 2010). Since Jeopardy has a 3 second time limit for players to answer a question, the test was to determine if IBM Watson was not only smarter than the human brain, but if it could accurately answer questions with detailed information in a timely manner (Ferrucci et al., 2010). IBM Watson follows a strict 3-step format when attempting to answer a question/provide feedback (Ferrucci et al., 2010):
Determine category of question
Identify and understand the clue or sub clues
Output an answer
An example, provided by Ferrucci is as follows (Ferrucci et al., 2010):
Category: General Science
Clue: When hit by electrons, a phosphate gives off electromagnetic energy this form
Answer: Light (or Photon)
After the Jeopardy Challenge was complete, data scientists determined that Watson answered about 30-60% of the questions correctly when it was 80% confident it knew the answer (Ferrucci et al., 2010). This means that Watson answered more than half of the questions correctly only if it was confident it knew the answer. Now, applying this information into the medical system, if a doctor needed to find an answer to a question using his own brain and no resources, it would take him longer than 3 seconds with less than 50% confidence to find one treatment option. IBM Watson provides a range of most beneficial options that a doctor can further research and compare to see if it will complement the patient’s health history and provide him or her the greatest chance of survival. The QA system is charged with Practical Intelligent Question Answering Technology (PIQUANT) software that was researched by IBM for 6 years prior to the Jeopardy Challenge (Ferrucci et al., 2010). The PIQUANT software was programmed to only access information from a local database and not the internet (Ferrucci et al., 2010). This is a crucial aspect of IBM Watson because when doctors are using this system they want accurate and board-certified information that is legal and ethical to use in their care. IBM Watson functions as database that analyzes a question, decomposes it to its main keywords and topics, filters its database, and provides an estimated or “best-possible” answer. However, DeepQA system follows multiple steps and is shown in Figure 1, below, developed by Ferrucci and his team (Ferrucci et al., 2010).
Figure 1. IBM Watson DeepQA Architecture
IBM Watson’s technology and interactions with itself are crucial to the success and performance of the database. However, there is another interaction part that is equally essential: interaction with its users. Users and can input data into the system, but how can others access it and how is the data transmitted? IBM solved this issue through its release of cloud capabilities, reported by Target News Service, “IBM Unveils Expanded Watson Platform for Health Cloud Capabilities, Introduces Watson Consulting Services at HIMSS17” (IBM Unveils Expanded Watson Platform for Health Cloud Capabilities, Introduces Watson Health Consulting Services Unit at HIMSS17, 2017). IBM Watson is a cognitive service that annotates clinical data and organizes physician notes, discharge summaries, and pathology reports, and uses the information about the patient’s and past patients’ histories about the disease, symptoms, and medications to produce insights for effective care (IBM Unveils Expanded Watson Platform for Health Cloud Capabilities, Introduces Watson Health Consulting Services Unit at HIMSS17, 2017). With the secure cloud, authorized doctors, scientists, and researchers have permission to access information on test results, patient history, medications used, and more information that is related to a specific disease or illness the scientists in researching. The IBM cloud allows doctors to share and access patient information to not only doctors within the hospital, but also across the country who use the same system.
The software system, PIQUANT and DeepQA, are the two driving forces in IBM Watson that runs the system, organizes and stores the data, and interacts with the system and the users (Ferrucci et al., 2010). The IBM cloud is the platform that focuses on the user interaction. Both software’s work together to form IBM Watson and aid doctors in providing precision medicine – ability for doctors to access precise data that aid doctors to understand the disease and select the most beneficial treatment(s) that compliments the patient’s health and body.
