AI in Healthcare: Lessons Learned from IBM’s Watson Health
It was not all that long ago that Watson Health, the IBM company that promised to revolutionize healthcare with artificial intelligence (AI), was making bold promises about finding cures for cancer. A decade after the Watson supercomputer was born, IBM is looking to offload their struggling healthcare company to a private equity firm or a tech company with a large checkbook.
So what went wrong? It wasn’t just one thing. Despite Watson’s promise of better record-keeping, better diagnoses, and advanced predictive analytics for better preventative medicine, the project was doomed to fail from the start. Proponents overestimated what AI is capable of. It can do a lot, but it is not a magic pill that will solve all of healthcare’s modern ills.
AI Isn’t Really Intelligent
At the root of most AI struggles is a fundamental misunderstanding of what the technology can actually do. To begin with, consider a sentient being with the ability to think. Sentient beings are intelligent in that they can recognize a lack of knowledge and then figure out ways to obtain that knowledge. Even the most advanced computers cannot do that. Thus, they are not really intelligent.
IBM’s Watson doesn’t know what it doesn’t know. It only knows what its programming tells it. Watson’s programming cannot account for a lack of knowledge because there is no artificially intelligent way to do so. Thus, the best AI can do is crunch existing data and reach conclusions from it.
Watson is certainly capable of doing that. It can do all sorts of amazing things. But it cannot think on its own. It cannot assess what it does not know and then figure out a way to obtain that knowledge.
AI Is Very Expensive
Another problem Big Blue ran into developing Watson Healthcare was the price tag. They spent some $4 billion to build Watson Healthcare into a company with revenues that exceed $1 billion annually. And yet, the company still doesn’t turn a profit. How is that possible? How do you bring in more than $1 billion annually and still come out in the red?
The reality of the matter is that AI technologies are extremely expensive. They are expensive to design, build, and run. In terms of operational costs, they go up the more powerful an AI machine is. IBM could not afford to spend the money necessary to develop Watson to the degree that it could cure cancer. Can anyone?
Interoperability Is Still a Problem
Perhaps the most important lesson of all is that interoperability is still a problem. That was one of Watson’s primary downfalls. It should not be a surprise in the healthcare industry. Interoperability has long been a bone of contention in the healthcare technology arena.
Dallas-based BenefitMall says they expect AI to play a bigger role in day-to-day healthcare over the next year or so. It likely will in relation to things like electronic billing, electronic health records, and medical transcription. But in terms of predictive analytics and finding actual cures, do not expect big things for quite a while.
Interoperability between systems limits AI’s contributions to the more advanced goals Watson Healthcare had in mind. The technology will get there someday, but not before interoperability problems are addressed. Intelligent machines must be able to intelligently work together or they will not get much done.
It is a shame that Watson Healthcare hasn’t met its lofty expectations. But it’s also not surprising. Those who follow AI technology religiously knew IBM’s order was a tall one. It is back to the drawing board for whatever organization decides to buy the struggling company.