Can We Detect Cancer Before Symptoms Appear? Part 1

So yeah, one of the things that I think of in this space is, you know, over the last few years w- with my experience focusing on biomarkers and early detection, I mean, there's so many areas in this space that we can go through with AI and how it's gonna impact or already is impacting the oncology space.

And arguably, um, detecting it earlier is one of those pivotal moments of where we can make a huge impact.

And I've seen on, in just the last few years how AI in healthcare has really evolved from just, like, this simple automation into systems that can reason across all these data systems, generating clinical insights, increasingly operate in what many are calling, like, agenic workflows.

And at the same time, as you've seen, I'm sure in the work that you've been doing with the, with the groups you work with, and I know you do also do a lot of advocacy work and, and work on the policy side.

But healthcare is struggling with clinician burnout.

It's one of the reasons I actually shifted.

I mean, I wanted to do more, but it's, like, so easy to get burnt out in doing patient care all day, and so, um, the fragmented data systems, this explosion of all this information.

So I'm really curious on your perspective on, you know, what are, what really today, like, not the hype cycle, but where are we today with the state of AI in medicine? Like, what excites you at night? Like, what you can't go to sleep at night because you're so excited and, like, that moment.

Like, what is that for you?

You know, a couple things.

I think what I'm seeing for the first time is that, um, we're actually programming the human body, right? And to me, and, and what, what I find so interesting, Lindsay, is that, you know, I, I, I right now in my, uh, my role at Google, I work a lot with video models, and what's astonishing is that the technique behind the video models, it's learning physics.

P- and Navier-Stokes equations are, like, complex equations of how do you get a space- SpaceX ship to Mars? Like, all the equations for the gases and the rockets, that's all called Navier-Stokes.

It's a high-dimensional physics, hard to learn.

We're learning it through observation.

We're learning gravitational pull.

We're learning emotion through observation.

And what's interesting is as you learn that, what d- what works that shouldn't work, and it's still astonishing, is that now that you learn information theory behind physics in these models, which is sort of astonishing, that maps extremely well into biology.

And so there's something funky about the universe where actually as you understand the information theory about things that, that are pixels and language, there's something about the language of biology, and it's mapping very well to understand the genome.

And so I said, "Well, how, what do you see in the genome? You got three, three billion pairs.

You got these 23 chromosomes." Like, how do you start? And it reminds me of a, a quote from NASA.

He goes, "Look for the big holes." When we were trying to figure out where the minerals were on the moon, I said, "Why?" He goes, "What do you think hits a...

What do you, what do you think causes a big hole in the moon?" I'm like, "Um, asteroid." He goes, "Right.

There's a big rock somewhere.

[laughs] Go find it." So the point is a snip, there's something there, so that's the big rock.

And what we're starting to see is, like, at Google, we can go across, like, for a million and start to understand what that million...