Yes, yes, we’ve read the Wall St. Journal and are well aware of the fact that IBM Watson’s leaked business results fall short of expectations. So what?
If Watson is the cognitive computing success we think it could be that won’t matter ten years from now. By then Watson will have been instrumental in teaching Cleveland Clinic medical students to practice the art of medicine (Watson will be handling much of the research and science) and to save more lives. It will have helped pharmaceutical and medical researchers bring new, safe therapies to market faster.
It will have helped retailers and service providers make suggestions that are actually relevant, by learning from what you do and do not like and get it right much better than today’s machine learning solutions. And it will enable customer support services to actually be helpful.
Heck, I might not even need to read the microscopic-sized serial numbers off of my devices when they break down. Watson will already know the make, model and other minutia I’m now asked for when I cry out for help.
Much of this can be done already, some of you might be thinking. At what level of accuracy, satisfaction and delight, I ask. Consider that every time I write about a vendor or technology, I’m deluged by ads for Enterprise software. Can’t anyone figure out that I’m extremely unlikely to have any use for it?
Machine Learning Beyond Ads
Now, mind you, getting people to click on ads isn’t on Watson’s agenda at the moment. Instead, it (the group of technologies which make up Watson) will initially focus on answering the world’s hardest questions, like finding the best approach toward treating a specific individual’s cancer. This, according to CEO Ginni Rometty and Watson chief Mike Rhodin.
While critics of Watson question whether it can go beyond spitting out answers to black and white questions by “thinking outside the box”, it’s worth noting that we’re not even leveraging most of the information “inside the box” at the moment.
According to Craig Thompson, President and CEO Memorial Sloan-Kettering Cancer Center, only about 20 percent of the knowledge that clinicians use today is evidence based. Pretty scary, right?
Better Than My Doctor Alone
And this isn’t because physicians are too lazy to keep up with their field or aren’t that into their jobs. It’s because it could take as many as 160 hours to read the studies being published each week. Given that a week has 168 hours and that more and more studies get published each year, the situation is likely to only get worse.
This being the case, if my doctor wants to collaborate with Watson, who can ingest and organize 160 hours’ worth of reading in no time, I’m good with that. And it’s worth noting that Watson, unlike Google, will have a relationship with my doctor and an ability to ask for more data (that could very well come from an additional medical test or physical exam) before providing an answer. It can also assign a probability for that answer.
Crunching and Combining Big and Small Data
But Watson’s value doesn’t stop there, according to Thompson. Watson can look not only at research data, but also at molecular and genomic data, a vast number of case histories, as well as an individual’s personal data based on medical history and such, before delivering a diagnosis and suggested treatments.
This sure beats diagnosing everyone who has a cough for a week or longer as having an upper respiratory infection and prescribing a Z-pack.
We'll look at even more about Watson tomorrow.
Title image from IBM.