What is pneumonia? No, that’s not a question, but it could be an answer, if a university hospital in Cleveland Ohio starts using Jeopardy star supercomputer Watson to help train its medical staff.
The IBM machine is set to move on from pitting its wits against challengers on the game show to a healthier goal, in which it evaluates medical case scenarios and diagnoses submitted by young doctors and either affirms or rejects it to a high level of probability.
The move will also be IBM’s first big push into the medical sector, worth trillions of dollars a year. In 2010 alone, the U.S. spent over $2.6 trillion--approximately 17.6 percent of its GDP--on health care.
IBM claims around one in five medical diagnoses are currently either wrong or incomplete, with as many as 1.5 million medication errors made in the U.S. every year.
With its “understanding” of natural language, ability to trawl through masses of data and aptitude for putting information into context, Watson would seem to be an ideal fit for finding answers quickly, allowing for more rapid patient diagnosis and knowing which treatments would be most effective for individuals, based on their personal medical history.
The supercomputer has already been tested at a New York-based cancer center, as well as by a major health insurance provider in the U.S.
The machine is already purportedly scoring well on test questions being administered to it off the United States Medical Licensing Exam, which all human medical students are required to pass in order to become fully fledged physicians.
Having Watson replace human physicians is not the plan just yet, though, according to IBM’s scientists.
Instead, it’s hoped that Watson will be able to boost a real medical department’s ability to keep abreast of new research, cut down on misdiagnosis and improve treatment suggestions.
Brave new world… but will it make medical professionals lazy? And what about its bedside manner?
In actual use we might get a situation more like HAL in the movie 2001, where they added special secret programming in the last minute:
"HAL, the symptoms are difficulty breathing and wheezing when subject is under stress."
HAL secret programming:
possibility 1: has a cold virus
possibility 2: has asthma
cost to treat #1: $25
cost to treat #2:
Further testing required: $1000
If asthma detected, $2000+/year
"Patient has a cold. Recommend decongestant."
Given the algorithms they developed, I would trust Watson much more than the average doctor to correctly assess my medical condition. Once tapped into the medical knowledge base, I could easily see a Watson connection in every doctors office so they could feed in the inital data and then get a list of useful tests to run to clarify the dianoses and prescribe the correct treatment.
I hope the AMA embraces the capability that Watson could offer to the public in more accurate medical treatment anywhere in the world.
Just a thought.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.