Artificial intelligence has been portrayed in movies both as friendly robots and cold-blooded killing machines. But John Daley, Vice President of IBM Watson Health, wants to help industry harness its power to improve public health, help doctors make more accurate decisions, and even treat cancer.
Artificial intelligence (AI) is a complex computer programing system, in which its algorithms and coding can change and improve on their own. In 2011, IBM introduced an AI program named Watson with a television debut on Jeopardy!, defeating two of the all-time champions in a three-day match. Since then IBM has applied the program to a variety of fields, using its data processing capabilities of to solve some of health care’s biggest challenges.
After working in quality assurance (ensuring that medicines and devices are produced safely and effectively) with several startups and Fortune 500 companies, John Daley joined IBM in 2016 to get involved in what Watson Health was doing. “This really is a chance to have an impact on the industry,” said Daley. “AI has the potential to change healthcare delivery, accelerate medical breakthroughs, increase efficiencies, and help lower costs.”
Enthusiastic and optimistic, I got the opportunity to speak with Daley on what artificial intelligence can do for the future of healthcare. The following interview has been lightly edited for clarity.
Chris Anderson: A lot of people have heard about Watson for beating humans on Jeopardy! How did IBM researchers go from asking a computer trivia questions to applying the technology to decisions about health care?
John Daley: The Jeopardy! appearance showed that Watson could really take unstructured data, understand it, make inferences, and give the correct answer. So from there, we asked, “what are areas with lots of unstructured data that we could add value to?” One area where you have that is healthcare.
You’ve got handwritten notes, voice annotations (from doctors), images, etc. We thought that we could really help there and so Watson Health was born. So they went ahead and bought Merge, a repository for some 30 billion medical images, Phytel and Explorys, companies that have tons of claims data on hundreds of millions of cases. We then started to use that data and develop an artificial intelligence system that can help process all that data.
CA: So what is an example of the impact Watson has had?
JD: We worked with the Barrow Institute, one of the world’s premier ALS research groups, to find genes associated with amyotrophic lateral sclerosis (ALS), Lou Gehrig’s disease, that have never been found before. Watson looked at the genes and protein through a process called genomics ( the study of gene sequencing).
In a few months, we discovered 5 genes that they had no idea played a role in ALS, which gave researchers brand new targets to study. You can just imagine what it feels like to do stuff like that, it’s like, “Oh man, let’s just keep going. It’s really cool!”
CA: What about public health?
JD: We just announced last week that in Ohio that one of the local juvenile courts is using Watson in case management. Think about the courts: tons of unstructured data. If a judge has a juvenile in front of him, he or she needs to know if they are in a stable home, if they have an addiction, the treatments they have taken, their probation status.
So the pilot program that has just launched will show a snapshot to the judge with all the relevant information for anyone who comes into their court. It will really speed things up and allow efforts to be put towards actually talking with a child.
CA: So the real genius of Watson is not so much that the program is “thinking” per se, but it’s taking all that data and organizing it and presenting it in a way that’s useful. Our brains didn’t evolve to process all that data, but Watson can.
JD: And you don’t have to go in and rewrite the program. It just knows.
CA: These are really cool, specialized applications, but if you just go to a doctor for your yearly physical, how is Watson Health going to affect you?
JD: We have a product that’s not out yet, but would have far more universal application. What it’s going to do is go through the electronic medical records and do the same thing we just talked about for a judge but for a doctor. It will learn to show information of a patient based off what the attending physician looks for. Say the patient is diabetic. It will show the last few weeks of blood sugar levels when the doctor goes in for your check up. It will learn just like a resident does.
CA: What is the most exciting development that you are seeing within the field of artificial intelligence, either at IBM or elsewhere?
JD: Watson for Oncology is out there and it goes into your medical record and pulls all the information related to your cancer, age, weight, type of cancer, progression, etc. It then, based on the research at Memorial Sloan Kettering, determines the current standard of care and recommends treatment options.
CA: So all the treatment data is based off how Sloan Kettering is treating cancer patients?
JD: And we do want to expand that. More data is always better, so if we can get other centers and more cases, we can only get better.
CA: Because there is a whole, pop culture, sci fi, the connotation of artificial intelligence, people are uneasy about it. Now we know it’s that you are going to see Dr. Terminator, but what does scare you about the future of AI?
JD: One thing that does make me nervous is that someone in good faith will put out an app or a program and something really bad will happen and it will set the whole industry back. One extreme, extreme, case that no one would have seen coming and hits the press. So my fear is that one really bad instance.
But in healthcare, we have the advantage of being regulated, so you have to slow the pace down, spend more time making sure the product works. And the regulators are really good at asking you the questions you haven’t thought of.
CA: Let’s say you are a student who thinks that artificial intelligence is really cool. What is something that kid should be doing if they want a career in AI?
JD: Just learn how to program! No need to specialize in AI at first, just get good at programming and familiar with different languages. Robotics is good, too. Also, take as many math classes as you can. The next generation workforce will need the critical data literacy skills required to be successful in a data-rich world.