Great future promise
Watson’s 2011 Jeopardy win was brilliant IBM marketing
Most of us are familiar with IBM’s Watson from its epic performance beating Jeopardy champions Brad Ruttner and Ken Jennings in 2011. At the heart of Watson’s win was its ability to access to 200 million pages of content totaling four terabytes of disk storage. Watson simultaneously executed hundreds of proven language algorithms to find correct answers. According to IBM, “(these) algorithmic techniques are used to analyze natural language, identify sources, find and generate hypotheses, find and score evidence, and merge and rank hypotheses.” The answer generated by the majority of Watson algorithms for each Jeopardy question was almost invariably correct.
Less well known at the time was that IBM’s primary commercial target for Watson was the medical industry.
Watson’s progress is impressive but has not met market expectations
IBM has invested over $4B in acquiring companies with extensive medical data to feed into Watson—including anonymous medical records and academic research data files. Any AI-based learning system requires extensive, complete data to achieve reliable results. Like any machine learning system, it’s only as good as the data it’s working from. IBM’s overriding objective has been to use medical data to help Watson make more precise patient diagnoses and to accelerate medical research.
In 2013, IBM announced Watson’s first medical software commercial application with the Memorial Sloan-Kettering Cancer Center, New York City in lung cancer diagnosis. It has also collaborated with the Mayo Clinic, the Harvard- and MIT-affiliated Broad Institute, medical-test giant Quest Diagnostics, the Jupiter Medical Center in Florida, a hospital chain in India, and 2,000 health care providers in a six-county area in New York. The goal of the New York project is to achieve 25% reduction in ER admissions and hospital readmissions.
Watson’s software is also being used by the Barrow Neurological Institute drug development research. That collaboration has resulted in the discovery of five new genes linked to ALS (Lou Gehrig’s Disease). Another related, amazing breakthrough–Watson’s findings have helped the Ontario Brain Institute identify 21 promising potential ALS drugs.
One notable setback has been with the M.D. Anderson Cancer Center in Houston. After four years that collaboration hadn’t created a viable patient diagnostic tool, and so the project was shut down. Internal management political struggles and funding issues account for much of this failure. However, IBM should accept its share of blame for this and a slower rollout of Watson health care delivery tools than expected. –It all boils down to its overly optimistic claim of how quickly Watson would evolve.
The main challenge for Watson has been ‘training’ the medical data that is available and filling in the gaps of still missing data. This problem is endemic to the entire field of health care machine learning,
Great future promise
Hopeful indicators of future Watson software breakthroughs—
- IBM stands above its competition in having the resources to be the first to make breakthrough advancements. In 2015, IBM expanded its data portfolio to Merge Healthcare medical imaging data for $1B.
- Research company IDC says the market for AI software and services will explode from $8 billion in 2016 to $47 billion by 2020, with Healthcare among the most profitable industries evaluated.
- IBM believes its platform will help doctors more efficiently diagnose and treat the 14 million Americans living with cancer, many of whom lack the resources to get the best treatments.
- Finally, IBM continues to excel at winning the trust of executives and IT managers. This is getting Watson inside many medical centers and other medical data collecting organizations to provide IBM with critical data required to the leader in AI-supported health care.