By Allison Proffitt

January 10, 2018 At a JP Morgan panel yesterday afternoon, what started as a discussion on the promise and future of artificial intelligence shifted to more fundamental discussions of drivers of health care. 

“One of the challenges of drug discovery has always been the lack of understanding of complexity of disease,” said Gini Deshpande, CEO of NuMedii. “We really need to understand disease at its fundamental level in order to be able to treat it. Our view is you can use AI coupled with really good data to look for subsets of patients, identify them at the front end, and then identify therapeutics that will help those patients effectively.”

Deshpande believes AI is most applicable early in the discovery process. Artificial intelligence is unlikely to speed up the clinical trials process, for instance, she said. “Our hope is that… these tools will enable us to ensure that the drugs that are identified are going to be effective in patients.”

Jamie Heywood, co-founder of PatientsLikeMe, agreed that AI is—at best—a tool, not a panacea for drug discovery. “The idea that you can just sort of observe everything, and plug in something you care about in a complicated, unstructured data environment and have the truth emerge is just not true. Where this is effective is where there’s clarity on an objective pattern that you’re interested in, and where you have data on metrics that are significantly driving the outcome of that pattern.”

In healthcare and research, we consistently over simplify both problems and solutions, Heywood said. “It’s largely a BS narrative. The system is so complex and so interactive and so resilient… that in order to modify the behavior of a complex biological system, or a complex social system, or complex outcome, you have to look at the interdependencies of those things.”

AI, for the first time, can consolidate multi-omic biological data, physician data, treatment and diagnoses, and patient input to reveal how resilient networks relate and interact, Heywood said. “That’s the part that’s going to be interesting… We live in a system that’s horse and buggies. We’re going to suddenly have helicopters.”

If only a tool, AI will continue to need operators. “There needs to be this human layer,” explained Jordan Shlain. “We still need that broader perspective and experiential understanding of what does this mean for your next week, month, year in the context of your life.”

Shlain is a primary care physician whose practice has offices in LA, San Francisco, and Silicon Valley. His practices are interested in longitudinal care, in being the organizing principle around care for you and your family, he explained.

Shlain argues that your most valuable asset is your health. "In our current healthcare system, we don't invest in our health." He compared our healthcare system with the way we view our financial investments. When it comes to money, we do research. We regularly check on our investments, rebalancing as needed. We pay far less attention to our health. 

"I'm out there building a model where you actually pay me, and I will be your health asset manager and advocate and advisor. We will not let things get past you... We keep you on the rails to make sure you're doing the right things." 

That requires a human touch, and it’s anything but artificial. But Shlain is hopeful that sophisticated pattern recognition will soon be able deliver the right information for you and your doctor to act on. “There’s too much stuff out there for me, as a doctor, to know about you and your life when you’re not here. If that could be collected in various formats and packaged up… it’s got to be actionable!”

Andy Slavitt, former acting administrator of the Centers for Medicare and Medicaid, agreed that AI or not, we must work toward more effective processes and programs to deliver healthcare. Since leaving CMS, Slavitt has been overseeing an investment fund, he said, and launching a “very large scale public policy effort, that I hope will bring sense back to Washington.”

“We’re not necessarily even good at—I’m convinced—using normal intelligence. Forget the artificial intelligence that can be created!” Slavitt encouraged healthcare to “look at the heart of the value chain where people are good at asking questions,” instead of stacking new solutions on top of old solutions.

Although NuMedii’s vision is to disrupt drug discovery with big data and artificial intelligence, Deshpande was frank about the most immediate opportunities. “To be very candid about it, there’s low hanging fruit in other sectors in the healthcare space where AI can have immediate impact, that’s probably closer to the patient. For instance, helping ophthalmologists spot patterns of diabetic retinopathy.”

Heywood was even more blunt. “Healthcare’s needs in AI are pretty weak. I don’t think the breakthroughs matter there. What matters is the breakthroughs in cost, scale of understanding of biology and the human condition.”