Paging Dr. Watson: IBM and Cleveland Clinic Collaborate to Train Watson in Medicine

13 8 Loading

IBM’s Watson aims to help docs sift today’s massive medical literature.

Can you imagine telling someone a century ago that a hundred years hence a stack of electrified silicon would be studying for its medical exam? It’s the stuff of Star Trek. (“Computer? Early Grey, hot.”) And yet at the end of October, IBM announced its question answering machine Watson will begin study at the Cleveland Clinic Lerner College of Medicine of Case Western Reserve University. (Mouthful, huh?)

The students and staff at the Cleveland Clinic will do for Watson what the rest of us do for Google—hone its answering accumen by using it. In this case, Watson’s users will be a little more hands on, actually answering its questions and taking a red pen to its answers. See here for more on just how Watson will work with students and clinicians:

In IBM’s words, Watson is a “computer system that can directly and precisely answer natural language questions over an open and broad range of knowledge.” They call the system “Deep QA,” run it on an 80 teraFLOP/s computer in Yorktown Heights, New York, and used it to defeat a pair of Jeopardy!’s all-time best champions in 2011.

But Jeopardy! wasn’t the final goal; it was a proving ground. IBM has great expectations for Watson in medicine (Memorial Sloan Kettering and Wellpoint), finance (Citigroup), and beyond.

Watson’s latest project with the Cleveland Clinic is kind of a match made in heaven—medicine’s Dr. Watson and Sherlock Holmes. Physicians already think a lot like Watson does. A central medical skill is sifting the literature to form probabilistic logic chains backing a particular set of possible diagnoses. Watson does the same thing.

Problem is, humans lack the hardware to keep up with the literature. The total amount of medical information available doubles every five years. Worse, medical knowledge tends to go stale within four to eight years. Meanwhile, free time is a commodity most physician’s have precious little of—five hours a month at most.

What is that for the average human? Maybe most of a real page-turner? Never mind assimilating and retaining a hundred pages or more of dense technical text.

Whereas on Jeopardy! Watson arrived at a particular answer (“question” in the game, of course) and simply buzzed in, the behind-the-scenes reasoning process was largely hidden. In the medical model, the reasoning process will be as important as the diagnosis. How can a doctor agree or disagree with his or her assistant without kicking the tires on their theory?

And that’s really the point. Watson will be a highly specialized medical search engine capable of human-like communication. Using Watson, a physician may uncover much of the same research he himself would have found, but he may yet find logically sound research that is new, compelling, and outside his purview.

Doctors will increasingly need only focus on keeping their rationality razor sharp and less so on maintaining an impossibly encyclopedic knowledge of medicine.

Watson (or some other system like Watson) could theoretically transform medicine by reducing the number of missed diagnoses due to our very human information limit. (There will be 200 times more medical data than a human could possibly process by 2020, according to IBM).

Might this tech someday make it into our homes and pair with an invention like that envisioned by the Tricorder X PRIZE? Probably not right away. The echoes of future lawsuits are loud enough to reach our ears even today.

But maybe eventually it goes home with a sturdy disclaimer, years of proven analysis, and a doctor backing it up. Watson could do a pre-visit check-up, and when you arrive at the doctor’s office, your doc gets right down to business. Or maybe after checking Watson’s analysis, he saves you the trip—two advils and a little rest! You’ll be fine.

Or maybe none of the above. The possible uses of AI-like software seem endless, but it’s the uses invisible to us that will be the most world shaking. As science fiction author, William Gibson, notes, “Any imaginary future…is written out of the zeitgeist.” And then almost immediately becomes obsolete.

Discussion — 8 Responses

  • James Hutton November 16, 2012 on 12:06 pm

    It’s interesting that Watson is quite a huge machine and only does 80 TFLOPs, I guess it’s a couple of years old now. Here: http://fireuser.com/blog/8_amd_firepro_s10000s_16_gpus_achieve_8_tflops_real_world_double_precision_/ is a desktop, (although its stretching the definition of desktop), that can do over 8 TFLOPs double precision, and it’s not too long before you could have this sort of power in a regular desktop. Then you could have local processing of diagnoses communicating with an even larger central machine. I think this will rapidly scale up, and soon all diagnoses will be confirmed with a computer, no matter how small or simple. I’d be surprised if the decade ends without your doctor having some kind of computer assistance in their office.

    • Sunshine2047 James Hutton November 17, 2012 on 8:11 pm

      GPGPUs are notoriously difficult to program. As far as I know Watson is a pure CPU system.

      • James Hutton Sunshine2047 November 18, 2012 on 3:15 am

        Yes I know that, but the software can always be ported. If you don’t like GPGPU’s you can always use a bunch of Xeon Phi’s instead, for a similar FLOPs per $ rate.

  • Ormond Otvos November 16, 2012 on 1:30 pm

    I, for one, welcome our new silicon death panelist!

  • Sunshine2047 November 16, 2012 on 10:09 pm

    The echoes of future lawsuits are loud enough to reach our ears even today.
    Well, you can always try it out in some of the world’s poorest region with serious lack of medical personnel before extending it to developed places. In Congo for example, Dr. Watson is better than no doctor at all.

  • PaulEpping November 20, 2012 on 11:42 pm

    Agree with you James, but think that we will get our diagnosis through our mobile device (smart phone I think) because tjis smart phone is also the ‘hub’ for potential sensors that can provide ‘Watson’ with additional info to make the diagnosis more precise. We ‘only’ need a connection to ‘Watson’.

  • Robert Schreib November 24, 2012 on 1:05 pm

    ?? Could Watson, like the fictional Watson in the Sherlock Holmes detective stories, likewise use its unlimited databanks and electronic speed of data scanning to help real life detectives solve countless ‘cold case’ files?

  • Mica Esquenazi January 25, 2013 on 6:35 pm

    Proposed relationship models between human action and machine/artificial intelligence have long concerned scholars in the field of science, technology, and society (STS). The production, utilization, and impact of machine/artificial intelligence must remain an important concern of society. Emerging technologies such as IBM’s Watson highlight these concerns (Collins, 1992: http://mitpress.mit.edu/books/artificial-experts). Correspondingly, integration of a machine such as Watson within the medical field requires attention and foresight. As developments with this technology continue, the potential of utilizing Watson as a diagnostic aid, or perhaps an autonomous diagnostic agent, is becoming more feasible. Speed and protocol efficiency in medical diagnosis are highly valued. In the future it may be possible for artificial intelligence to perform more strongly within this realm than physicians. This is a future projection often mentioned by scholars discussing IBM’s advancements in Cleveland. While this probability has the potential to garner strong support from the medical community, as presented by this article, there remains the potential for strong rejection from the medical community as well. This is a way to diminish human error and reduce liability by the physician or institution.

    Currently, robotic surgery has received very positive reception within the medical community. Conversely, the use of computer interfaces, be they desktops, laptops, or tablets, has received very poor reception because of negative repercussions to the patient-physician relationship. The emotional investment of physicians is crucial for proper medical practice. As such, the usage of tools such as Watson within the medical field must occur with biopsychosocial considerations (Engel, 1977: http://www.sciencemag.org/content/196/4286/129). It is probable that for future physicians to rely on technological diagnostic tools, an entire paradigm shift would be necessary within the medical field (Kuhn, 1962: http://www.amazon.com/structure-scientific-revolutions-thomas-kuhn/dp/0226458083). The potential of Watson and similar technologies to occupy a major role within the medical field is not simply incumbent upon technology, but also upon the medical community and society at large. The advancement of this technology seems to be on the rise; consequently, we must pay close attention to the future of these technologies with a biopsychosocial outlook in order to insightfully cater to human and societal needs (Bloor, 1976: http://www.press.uchicago.edu/ucp/books/book/chicago/K/bo3684600.html).