Summary Bullets:

• IBM leveraged Deep Blue, a supercomputer, to victory over chess champion Gary Kasparov and Watson’s subsequent victory over Jeopardy champion Ken Jennings to highlight the promise of AI.

• IBM hoped to expand Watson beyond game playing to solving problems that had proven intractable to business managers. Seeking a challenge, IBM adapted Watson to diagnose and treat cancer.

IBM’s experience developing the platform for Jeopardy gave the company hope that it could broaden Watson’s capabilities. Key to the success of Watson’s victory at Jeopardy was the development of DeepQA, a specialized software architecture that, according to IBM, “generates a wide range of possibilities and for each [query] develops a level of confidence by gathering, analyzing, and assessing evidence.”

Using AI to treat cancer, a disease that worldwide, caused 10 million deaths in 2020 and costs $167 billion to treat would mark a critical milestone in healthcare and AI. IBM believed that it could develop DeepQA to help doctors diagnose and recommend the best therapy to treat individual cancer cases.

Building out Watson’s ability to diagnose and recommend treatments for Cancer, IBM partnered with several of the leading cancer treatment centers in the US, including MD Anderson Cancer Center (MDACC). IBM added doctor’s patient notes and over 23 million summaries from medical journals. Hoping to make better use of massive amounts of data that cancer care centers generate, MDACC sought to develop an “Oncology Expert Advisor” (OEA) to help their oncologist improve diagnostic capacity and apply the latest treatment options to patients. In 2013 MDACC revealed a collaboration with IBM to build the OEA.

Unfortunately, it soon became clear that both IBM and MDACC had vastly underestimated the challenges of the project. Despite some positive results and adoption by several other hospitals, MDACC canceled the project after spending $39 million in development costs. For its part, IBM spent over $5 billion acquiring companies to help train Watson. However, IBM’s failure has not deterred competitors such as Oracle, Google, and Microsoft that are looking to take a run at capturing some of the $3 trillion spent on health care annually. Not surprisingly, IBM has been criticized for its process of developing and marketing Watson Health.

Being a pioneer is never easy. IBM’s attempt to build a solution using technology still in its infancy and then apply to one of the least automated industries was risky. There were several areas that damaged Watson’s chance of success.

• Doctors complained that Watson’s provided “black box” treatment options that did not adequately inform doctors of the rationale behind the solutions.

• Watson Health utilized medical data exclusively from US institutions creating a bias to other proven non-US treatments. This bias made doctors in foreign countries suspicious of Watson’s diagnoses.

• IBM did not allow full-peer review studies to be published in medical or scientific journals of record. Almost every public study was either conducted by a paying customer or included IBM staff on the list of authors.

---