In 2011, with its well-publicized Jeopardy win, IBM showed off Watson's ability to ingest a vast amount of information, and to answer sophisticated questions posed by humans. Since then, the company has been working behind-the-scenes to push Watson into the nation's largest industry sector: Health care.
The applications for health care are wide-ranging and diverse. Oncologists at top hospitals are experimenting with using Watson to pull up relevant clinical trials and medical research while treating cancer patients. Other health institutions are analyzing a vast pool of patient data from various sources -- and directing resources to those who are most likely to get sick. The company recently announced a partnership with drugstore chain CVS to reason through 70 million people's health records and identify the subtle signals of disease progression.
In June, we broke the news about IBM's latest health care collaboration. The company is working with a nonprofit called Way to Wellville, which is on a mission to transform a community’s health status in five places across the country, in just five years.
Improving health outcomes for thousands of people is one of Watson's most ambitious projects yet. KQED spoke with IBM executives and data scientists to determine if Watson is up for the task.
From Jeopardy to Public Health
A few years ago, IBM's Steve Gold had a fateful encounter with Esther Dyson, a prominent technology and health investor who has backed startups like 23andMe.
"She [Dyson] had a noble thought. I got to thinking about how we could help," said Gold, who works on the Watson team as a chief marketing officer.
That noble thought involved using technology to improve community health. Dyson and the team behind the Way to Wellville project, as it would later be known, opted to work with five communities across the U.S.: Lake County, California; Spartanburg, South Carolina; Niagara Falls, New York; Clatsop County, Oregon; and Greater Muskegon, Michigan.
Many of these communities are "high poverty hotspots," meaning they have high concentrations of residents under the poverty line. Across the board, these communities struggle with keeping their citizens healthy and managing ever-rising health care costs.
Health experts say that the industry desperately needs user-friendly tools to parse through a massive amount of data, which is often scattered among many different systems. "The problem of how to make health data meaningful is a very large one," said Ash Damle, a data scientist and the chief executive of Lumiata, a company that specializes in analyzing health data.
Hospitals and health clinics store a vast amount of data from doctors' notes, patient medical records, research papers and academic journals. It's a virtually impossible task for the human brain to search through this data and dig up the most relevant information for a patient -- but a computing system like Watson may be able to do it.
In addition, IBM spent many years pumping Watson full of medical knowledge. By 2013, Watson had analyzed 605,000 pieces of medical evidence, two million pages of text, 25,000 training cases and had the assist of 14,700 clinician hours fine-tuning its decision accuracy," Forbes reported. Much of this data is "unstructured," meaning it is text or video-based and can't be easily stored in a database like Microsoft Excel.
The Best AI for the Task?
Still, machine learning experts say that they aren't yet convinced that Watson is the best option for all health providers -- particularly those who lack ample resources.
According to Damle, the Lumiata CEO, health organizations need to invest in training for their employees to use Watson for the long-haul, as this kind of technology gets smarter over time. For those who can't afford IBM's consulting fees, this may prove to be an ongoing challenge.
Moreover, many Silicon Valley startups are building more affordable "open source" alternatives. Data scientists can view the computer code under the hood, and tweak it if necessary.
"IBM is taking a different tack: A more business-focused, closed approach," said Anthony Goldbloom, chief executive of Kaggle, a company that hosts data prediction competitions and works with hundreds of data scientists. Goldbloom added Watson's Jeopardy win was an "unbelievable achievement" but it's not yet clear whether Watson is the easiest-to-use "general-purpose" tool for health care applications.
Others take a more skeptical view.
"The IBM approach is to claim to solve every problem and scramble to write custom code where their core capabilities are lacking," said Matt Ocko, a venture capitalist at Data Collective, who specializes in data science and machine learning.
"But there may be a lot of startups with better machine learning technology, which take a more specialized approach, ready to solve problems where IBM is behind," he said.
IBM has agreed to work with the Way to Wellville team for five years, lending both its technology and talent.
IBM's Gold sees potential for the team to experiment with the new digital health technologies during this period. Watson will ingest data from the latest wearable devices, such as the Fitbit, the iPhone or the Apple Watch, so the community can monitor individuals' exercise levels.
"You could imagine putting a wearable device on a fifth-grader to track their movement," said Gold. "You'd need a good way to merge all this data into the cloud and draw some meaningful conclusions -- and that's where Watson can help."
IBM's partnerships director Toshio Mii recently flew to Spartanburg, South Carolina to meet with community health officials. Spartanburg is focusing its energies on combating childhood obesity.
It's still early days for the project, but Mii said the IBM Watson team is brimming with ideas. He has high hopes for wearable-technology and is currently evaluating trackers from companies that focus on kids, like Sqord. The community will pay for and distribute the trackers, Mii said.
According to Mii, Spartanburg is also exploring ways to improve the quality and nutritional value of school meals, as well as the access to green spaces.
Finally, the community will use messaging tools to help uninsured families in Spartanburg access information about improving health and tackling obesity. The community will also use this technology to remind patients to keep their doctors' appointments.
If any of these projects prove successful, IBM will help gather up some of the data and present it to policymakers and community health researchers.
"The goal is to replicate some of these projects elsewhere," said Mii. "We want the positive results to proliferate."
Get the best of KQED’s science coverage in your inbox weekly.