Watson, M.D.: ‘Jeopardy!’ Champion Turned Medical Student

0
1002

Gilberto Flores

Meet your new doctor: Watson, the IBM supercomputer.

Watson, the sophisticated question-and-answer system developed by the International Business Machines Corporation for the purpose of understanding the complexity of human language, has a new job diagnosing patients at the Memorial Sloan-Kettering Cancer Center in New York. IBM has been using Watson to help doctors at several hospitals diagnose and provide treatment options for their patients. The breakthrough of a supercomputer that understands human language, along with Watson’s razor-sharp accuracy, has the potential to help doctors eliminate unnecessary costs and procedures, as well as fundamentally change the United States’ $2.8 trillion healthcare industry.

But what exactly is Watson? According to IBM, Watson is “a computer system that can directly and precisely answer natural language questions over an open and broad range of knowledge.” Essentially, Watson is a room of computer servers combined with IBM’s DeepQA software to create a system that analyzes language and provides answers using a series of complicated word association algorithms. Watson is composed of about 90 IBM Power 750 servers, stacked together in a server room at the Thomas J. Watson Research Center in Yorktown Heights, New York. These servers make up the equivalent of about 2,800 powerful computers with a memory capacity of over 15 trillion bytes, or about 14,000 gigabytes.

Watson was conceptualized in 2004 when IBM Research manager Charles Lickel was having dinner with co-workers at a restaurant. The room had fallen silent because everyone in the restaurant was watching Ken Jennings play “Jeopardy!” on the TV. Jennings was in the middle of his famous 74-game run on the game show. Lickel was intrigued by the idea of using “Jeopardy!” as a possible challenge for IBM, and Watson was designed specifically for the purpose of answering questions from the game show.

According to Dr. Chris Welty, an IBM Researcher for Watson’s Algorithms team, the first version of Watson was a question-and-answer system that was unsuccessful in comprehending the complexities of language and word association. To be able to compete, Watson needed to understand simple elements of human language, like the difference between “a runner running in a race” and “a candidate running in a presidential race.” For the next seven years, researchers worked tirelessly to develop the technology needed for Watson to be able to compete on the show against some of the best players. In February of 2011, Watson made its debut on “Jeopardy!” and was introduced to the world.

On the show, Watson had to stand on its own like all other contestants. Watson was not connected to the Internet, so it had to rely solely on all the knowledge it had. Since Watson cannot hear or see, it received the clues as a text file at the same time that the host read them. Watson reads the question, generates thousands of possible answers, and narrows it down to three possible answers using complex computations and algorithms. Watson then used his computations to assign a confidence level to each possible answer, and if the confidence level for his top answer was not above a certain percentage, Watson would not buzz in.

Standing completely alone, Watson managed to use its 15-terabyte memory capacity to win the game with a staggering total of $77,147, besting its competitors Ken Jennings and Brad Rutter, both of whom scored around $20,000 each. All contestants were playing for charity, however, and Watson’s IBM team was given $1 million to donate to the charity of their choice.

Although it was an unprecedented win for IBM, the team had no intention of stopping with “Jeopardy!.” Dr. Joseph Jasinski, an IBM researcher for the Smarter Healthcare and Life Sciences department, stated that “there are lots of other domains where people want questions answered, healthcare is certainly one of those places… Suppose you are a clinician, a doctor, [or] a nurse trying to diagnose a very complex case. You have some ideas, but in order to confirm your hypothesis you need a lot of information.”

Updated to comprehend hypothetical scenarios instead of just quiz questions, Watson is now able to go through hundreds of thousands of pieces of medical information — from textbooks to medical journals, clinical trials, and thousands of other medical documents to find and pinpoint a diagnosis. In addition to a diagnosis, Watson is able to determine possible treatment options for a patient and assign each treatment a confidence level based on the likelihood of that treatment’s success. Watson’s amazing speed and accuracy make it capable of diagnosing cancer better than the average medical student and as well as a doctor.

So what’s next for Watson? According to Ginni Rometty, Chairwoman and CEO of IBM, Watson’s research team at IBM is currently in the beginning stages of developing Watson 2.0, “a Watson that can see, actually see pictures and interpret them, [such as] x-rays… and Watson 3.0, a Watson that can debate and reason.”