IBM officials set lofty goals for Watson's impact on the healthcare industry, but recent lackluster performance may serve as a reality check for the tech giant, according to an MIT Technology Review analysis by contributor David H. Freedman.
In reference to Houston-based MD Anderson Cancer Center tapping IBM to help develop its Oncology Expert Advisor, the general manager for IBM Watson Solutions claimed "IBM Watson represents a new era of computing" in 2013. However, in February 2017, the cancer center decided to put the project on hold.
"The breakup with M.D. Anderson seemed to show IBM choking on its own hype about Watson," Mr. Freedman wrote. He added, "After four years [Watson] had not produced a tool for use with patients that was ready to go beyond pilot tests."
Mr. Freedman acknowledged Watson's shortcomings weren't indicative of technological flaws. In fact, he noted the negative news surrounding Watson is often "a reaction to IBM's overly optimistic claims of how far along Watson would be by now."
In the case of MD Anderson, the problems seemed rooted in management and funding issues related to the Oncology Expert Advisor project as a whole, rather than IBM's technology. In addition, the extended amount of time it takes IBM to get its Watson projects off the ground is reflective of broader issues to deploying machine learning in healthcare.
To provide effective feedback to clinicians, Watson must be trained on certain types of data. But for some complex problems, the data might not exist. "[F]or potentially groundbreaking puzzles that go well beyond what humans already do, like detecting the relationships between gene variations and disease, Watson has a chicken-and-egg problem," Mr. Freedman wrote. "How does it train on data that no experts have already sifted through and properly organized?"
Mr. Freedman claimed IBM Watson will likely still be a leader in AI in the healthcare space. However, the company must stop providing the public with "unrealistic timelines and promises."
Click here to view the full article.