������ý

This is very emotional for me.

(Cheers: “We love you!”)

I love this place.

Twenty years ago, I sat where you’re sitting. I was 21 years old. Homesick. Scared about what was ahead. Anxious about everything I was leaving behind. It was a big moment. Now, I want you to imagine yourselves 20 years from now. The year is 2044.

(Gasps and some clapping)

And you all have done incredible things. Some of you have found treatments for slowing down aging. Some of you have achieved commercial fusion, providing a limitless source of clean energy.

(Cheers)

Some of you have enhanced brain–computer interfaces, allowing individuals with neurologic diseases to communicate naturally and have a much better quality of life. Some of you have created AI that takes diagnostic errors from being the third leading cause of death, as it is today, to negligible. And some of you have even established human colonies on Mars or the moon.

(Gasps)

You’ve each found your own way to continue the great march of human progress. I have no doubt you will accomplish fantastic things. The future is yours.

Twenty years ago, I wanted to believe this, but I wasn’t sure. And my day-to-day required some tools to get there. I'll share three that I found to be most valuable.

First, build conviction.

In 2008, I was at Stanford working in AI. A physician colleague reached out to me and said, “We’re beginning to collect all this data on babies in our hospitals. Shouldn’t there be a way to make use of AI to make sense of it?” Prior to then, most hospital records were paper records; they were written in paper charts. And this idea of collecting data electronically was very new.

I knew very little about medicine, but as I went into it, I realized how antiquated the practice of medicine felt and could see new ways in which we could use this data to forecast which babies might be at risk for complications so you could prevent them. It was a new kind of treatment and a new kind of care model.

Yet, it was surprising to me that they didn’t already have it. I ran the idea by several doctors and they said it would never work. I ran the idea by several hospital leaders and they said nobody would pay for it. And I ran the idea by several medical researchers and they said the data was likely junk.

(A chorus of shocked gasps)

I realized the doctors were pattern matching off past experience, and they had seen nothing like it, so couldn’t imagine this new approach. It’s true they didn’t have a way to pay for it yet, but that doesn’t mean there wouldn’t be a way to pay for it if we created value. And the researchers who said no had little knowledge of AI.

I worked in AI, and it was clear to me that it’s uniquely positioned to make sense of these types of very messy, biased data sets. And by doing so, we could enable new software-based diagnostics that could help doctors go from practicing reactive care to proactive care. By reasoning from first principles, I built conviction. Every time someone said no, I tried to understand the why. If I didn’t hear anything that violated the core principles, it deepened my conviction, and I pressed on.

Fast forward: It’s 2015. I’m now a professor at Hopkins, and my lab …

(Cheers)

And my lab has shown how this method could work in many different disease areas — in scleroderma, in Parkinson’s, in sepsis, to name a few.

I was writing paper after paper showing how these methods could beat the standard of care. And then something really tragic happened: I lost my nephew to sepsis.

Sepsis is a life-threatening complication that happens when the body’s immune system starts attacking its organ systems in response to an infection. Turns out, in sepsis, the most effective form of treatment is to treat it early. An hour can mean the difference between life and death. I knew I needed to turn my papers into a real system that any hospital could use. I hired a team. We worked day and night for a year and a half building it. Finally, the day when the system would go live was coming near, and I was so excited that all of our research, the studies and all of this work would finally be real and finally be able to save lives. The day came. We turned the switch. And guess what? Only two doctors used it in the entire hospital.

I was shattered. None of the work mattered because no one was willing to use it. But that’s also when I learned that it was way more than an AI problem. It was a systems problem. It was an incentives problem. It was a human problem. Accept the fact that there will be totally unanticipated problems that you have no skills for, no intuition for, and you'll have to figure it out.

Turns out, this is how all hard things get done. And if you have the conviction and you embrace the unexpected, you will get there. Today, our system is being used by thousands of doctors in hospitals around the country, and we are still not done. In fact, we’re only at the beginning. We celebrate every new doctor we onboard. We celebrate every new hospital we sign. And we celebrate every life saved. I’ve learned to fall in love with progress — putting one foot in front of the other.

When I was little, I was told, “Girls don’t study science.” When I was in college, I was told, “You're not well-suited to do a Ph.D.” When I was in graduate school, I was told, “You’re not meant to be a professor.”

It almost always came down to deep conviction, embracing the unexpected along the way and falling in love with progress.

(Cheers and applause)

Looking back, Mount Holyoke gave me the space to think, many nonjudgmental sounding boards, the confidence to challenge ideas and my very own cheering squad, who back then made sure I was awake.

(Bursts of laughter)

They reminded me last night who back then made sure I was awake in time for class. And since then, have continued to remind me when I doubt myself that I have it all together.

You are all so lucky to have this place as a launchpad for the rest of your life.

To close in Mary Lyon’s words, “Go forward. Attempt great things. Accomplish great things.”

Thank you.