Watson and A.I. in Healthcare

Posted on May 9, 2014


Smart phone applications are great, ever improving in fact, but for healthcare they seem to lack complexity and legitimacy.  By legitimacy, I mean an outstanding mechanism (data base / algorithm) to provide specifically essential and trusted insight.  Emergency Department wait updates are nice, but one-trick ponies ultimately.  And who (hopefully not too many) needs to use the local ED with a frequency to warrant a fingertip-ready app?

Smart phones have facilitated major care giving improvements in the field (in rural areas away from hospitals and civilization) throughout the past several years; however, lack of a really good ‘pocket doc’ app is still left wanting by most.  Many in healthcare acknowledge that given the initial excitement around a WebMD-type idea (the website which never really achieved the potential or success it was built for), the internet in general has been disappointing from a healthcare helper standpoint.

In 1997, IBM succeeded in creating a computer that defeated the world’s best chess player, Garry Kasparov.  Its name was Deep Blue and it took IBM a retooling and rematch from their 1996 effort to win.  This was groundbreaking work for IBM because chess was seen as an appealing proxy for the type of complex but finite (rules-wise) and analytical arena in which artificial intelligence decision-making could be tested to mimic a human.  Successful chess is also not a linear, formulaic effort; it includes creativity and nuance affected by the personality of the player—an artful side—things which computers historically were not good at accommodating.  As a chess fan, pop culture and technology follower, this was an historic and philosophically unsettling event.

In 2012, Fast Company profiled IBM’s brilliant super-computer, Watson.  Watson gained fame in 2011 by beating another human in a knowledge game, in this case Jeopardy! game show all-time champion Ken Jennings.  Since that win, Watson had morphed into “Dr. Watson” in that its artificial intelligence was now focused on solving healthcare problems.  Where Deep Blue was built around statistics and probability analysis in the service of developing strategy, Watson was built to handle two different and very complex items:  language, and what is known as unstructured data.  A short 2012 comparison of the two computers can be found here.

The ability to process unstructured data is a big deal. This is what makes a human doctor valuable, or any professional that must digest and process text-based knowledge for that matter.  A human doctor must apply experience and professional judgment to extensive book knowledge (unstructured data) to help determine the best clinical care approach for a patient, i.e. the diagnosis.  Sometimes the result is mechanical (execution of a surgery), sometimes it is consultative (verbal recommendations).  Given the immense complexity of the human body, all the medical data, studies and research available—historic and newly generated each year—that is a lot of unstructured data.  In fact, the subset of data is so overwhelming, it is argued no one person / doctor can keep on top of it all for one specialty, let alone master multiple specialties.  Therefore, the most effective processor of this data could be a computer.

Dr. Watson will not replace any human doctor anytime soon because of the judgment and application aspect of medical data.  Processing unstructured data is Watson’s forte, but at this point Dr. Watson crunches and collate data and provides human doctors with ranked options for possible outcomes; it offers no defined solutions.  Watson cannot perform a surgery or conclude which option is best for a patient (because it cannot converse, empathize and reason with a patient).  That is the doctor’s job.  At this time, Watson is simply the world’s best clinical consultant—a phenomenal feat.

All of this leads to how Watson’s expertise can be tapped by the common man.  IBM has smartly initiated a Watson Mobile Developer Challenge to help apply Watson’s expertise in large-scale ways.  The apps to come forth from the competition are the 1.0 of development around Watson; they are baby steps, but progress nonetheless.  Becker’s Hospital Review profiles some of the apps here.

Watson is, by far, the most powerful healthcare resource in existence, with a mind-blowing upside, and the general public is largely ignorant of its existence—but not for long.  I think of those Intel advertisements that used to have the logo “Intel Inside” for computers powered by Intel microprocessors.  I see Watson having the potential for a brand strength that will be the analytical engine for clinical insight of the future.  Apps, medical equipment and devices, kiosks, all kinds of patient interactive encounters could be ‘powered by Watson’.

And Watson will only continue to improve.  I hope the learning curve for app development remains steep, and I eagerly await what the future holds.