Watson, the Jeopardy playing super-computer, is out to take on new challenges. Rather than reworking its algorithms for a chance to compete on say “Deal or No Deal, IBM executives have other ideas in mind.
At The Singularity Summit 2011 over the weekend, Dan Ceruttti, who is effectively Watson’s business manager, told the gathered researchers, computer scientists and neuro-biologists that Watson’s 2880 Power7 cores would be crunching away on our nation’s healthcare problems. Cerutti mentioned a “Watson for Healthcare” initiative, which IBM has been talking up to the press soon after Watson became the first non-human Jeopardy champion.
Yesterday I was at Predictive Analytics World chatting with IBM’s Paul Hake at the Watson booth. According to Hake, the healthcare initiative is actually broken up into several “use cases”. While an evidence-based diagnostic or treatment app is the sexiest of the kind things that Watson can do for healthcare, Hake mentioned other problem domains that Watons will take on.
One of the first to be released will be an app for assisting providers in getting reimbursements–i.e., understanding complex insurance polices and deciding which treatments are billable. Hake also mentioned a consumer facing oncology app to help the public unravel cancer treatment options and answer other questions.
These are all worthy scenarios for Watson, but I had to circle back to Watson as a diagnostic tool. I had first thought that reading doctors’ and nurses’ hand-written case notes would be a deal killer, but Hake reminded me that lots of records, especially in larger medical institutions are on-line. IBM’s approach will be to partner with EMR vendors–the likes of Eclipsy and Meditec–as well as work with speech-rec vendor Nuance for an ultimate voice interface.
And by the way, the booth had a Jeopardy demo setup, and I fell for the bait.
I went O-2 against Watson, blowing a question about the EU–Istantbul was the answer–and my aging memory circuits forgot that Chablis was a 7-letter Burgundy appellation. On the latter, perhaps IBM could come up with Watson for Food-Wine Pairings?
This is not so far fetched since there are plans to release a mini-Watson down the road. This scaled-down version will have the same capabilities as the full-blown Watson but with more modest hardware requirements it will simply take longer to answer.
I look forward to a future hackathon in which there are … hackable Watson APIs!