Watson, you may remember, is the IBM supercomputer that beat two of Jeopardy’s most famous contestants on air in February 2011. Watson is a natural-language processing superstar that can read around 200 million pages of text in three seconds and come back with a set of answers to a problem or challenge ranked by its confidence level in its answers. In Jeopardy, Watson would buzz in only if its confidence level was high enough for a given answer.
If you can believe this, Watson is now going to medical school at Memorial Sloan Kettering Cancer Center. Starting in March of this year, researchers are working with the elite supercomputer to teach it to learn oncology and have it suggest proper courses of treatments to doctors. Instead of answering trivia questions, it will now present a doctor and patient with handful of options for treatment, diagnosis or research. In addition, it will inform doctors of the specific sources it used to derive its answers. It’s not designed to tell the doctor what to do but guide in the decision process. “Health care is a lot less deterministic than Jeopardy,” said IBM chief medical scientist Martin Kohn. Watson has some limits, such as being able to read PET and CT scans to identify tumors. It can only read text, but IBM says it will apply some of its new image-recognition technology, that can interpret EKGs, echocardiograms and cardio MRIs. IBM also reports that about 90% of nurses in a clinical setting who utilize Watson actually follow its medical advice.
If image-recognition isn’t enough for your “wow-factor”, IBM has reduced the size of Watson from the size of a large bedroom to about the size of a pizza box server, all while increasing the CPU speed by 240%. Even though data has grown around 800% in the last five years (90% of it is from the last 2 years), the technology is so advanced it can fit inside a pizza box.