Dorothy Chou | How tech can earn public trust and why 21st century science needs 21st century funding

[00:00] Dorothy: I've worked in tech for over a decade. If you look at the public sentiment in reaction to technology, it will tell you something, and I think we like to be in our bubbles and tell ourselves that we're doing good things for the world and they'll understand it, but that's actually not true. There needs to be some translational element between what the tech companies are building and helping people understand what's happening, and then letting people decide if that's something they even want.

It used to take a PhD four years to find the shape of a single protein based on an amino acid sequence, and with AlphaFold, we could do it all in one go, all 200 million of them. Why, in the wake of Google DeepMind winning the Nobel Prize, is there less and less investment in biotech? It's not something that's rational or logical.

It's not just a Google DeepMind problem, and it's not just a Google problem. It is a much bigger and systemic challenge that we need to fix.

[01:00] Beatrice: I'm very excited to be here today with Dorothy Chou. You've spent 15 years working at, I think, where technology meets society, and you've been at some of the biggest companies that we have. You've been at Google, at Dropbox, at Uber, and the last nine years at Google DeepMind.

[01:16] Dorothy: That's right, yes.

[01:17] Beatrice: And now you're leaving Google DeepMind. So nine years at DeepMind, what makes you want to work from outside a lab now?

[01:27] Dorothy: Well, I'm very lucky to be able to transition into a strategic advisor role while I explore a few things on the outside. I've spent my career bridging worlds.

I'm not an engineer, I'm not a scientist, but I did grow up in a family of biologists. My parents both worked in pharma and then in biotech, and then my brother's a doctor. So when you grow up as the person who was not good at the things that your family is good at, to participate in the conversation, you actually have to keep up.

And you have to learn what they're doing. So I've always seen myself as a person who's trying to bridge worlds, bridge how I think with how other people think. And I increasingly see that the world of what AI can do and where tech is sitting and how the public feels about it, or how our funding systems are approaching it, the gaps between that, it's not just a Google DeepMind problem.

And it's not just a Google problem. It is a much bigger and systemic challenge that we need to fix. So what I'm trying to do is spend more time with organizations that are sitting at the intersection of real world needs and what we're building in technology. One of them is, I'm the chair of an organization called UCL Partners, which is trying to figure out how we bring technology into the NHS in London.

Really serving some of these dire real world needs and connecting the dots between the two. So I'm excited to spend some time doing that, and also connecting and thinking through how our financing ecosystem can better support those types of things.

[03:12] Beatrice: Yeah, I love that. That makes a lot of sense. We're hosting a workshop later this year on AI and democracy.

One of the things that is really important for us is to have both the people building the cool AI tools, because those are exciting, but then also the people that are actually working with implementing democracy right now. Because for them it feels very far away, I think, to be able to implement these sort of things.

[03:40] Dorothy: Yeah. It's hard. It's two fundamentally different kinds of questions. And I think people are so used to working among people who unfortunately or fortunately think like them. So when you're a scientist or an engineer, you are trying to engineer towards the right solution, or in science, deducing towards the right solution to your hypothesis, or proving your hypothesis.

The question that we're trying to answer in society writ large, in democracy and politics, is how do we live together? Those are two very different kinds of questions, and they take very different types of thinking. So for me it's about translating the constraints and the incentives of each party to the other, which really starts to hopefully bridge some of these gaps.

[04:26] Beatrice: And yeah, we're going to talk in this conversation about how we can fund science and the age of AI. But before we do that, I just want to have a brief overview, because I think one question that a lot of people really grapple with is: how can we really use AI for good? What do we need to do to actually realize its potential?

If you had to give a brief overview of the challenges there, what would you say?

[04:54] Dorothy: Well, for me personally, what I've been looking at is why, in the wake of Google DeepMind winning the Nobel Prize in AI for biology — I mean, it was a chemistry award, but AI for biology — is there less and less investment in biotech?

It's not something that's rational or logical. In the wake of the Nobel Prize, what we should be thinking about and changing is moving from a patient, linear, and discipline-specific way of funding biology to something that is much more around underlying infrastructure and conditions, and we're not used to doing any of that with the existing financing systems we have today, whether it's government, which is much more used to that linear and discipline-specific way of doing it, or venture, which is much more used to a five to ten year exit window for the things you're financing in software.

Biology doesn't fit any of those things, especially not AI for biology. So what we need to do is not only shift the way we're doing government and philanthropic investment, but also think through how we tie that in to the commercial sector in a way that makes sense for all these parties.

We have very different constraints and very different incentives.

[06:12] Beatrice: Yeah. And one more question before we dive deeper on the science funding: one of the things that you built at Google DeepMind was the Public Engagement Lab, which I think is a really interesting thing. What is a Public Engagement Lab,

and why does a lab like DeepMind have one?

[06:34] Dorothy: Well, to be honest, to your point, I've worked in tech for over a decade, and I honestly am not sure that the sector overall has ever done public engagement well. Right? If you look at the public sentiment in reaction to technology, it will tell you something, and I think we like to be in our bubbles and tell ourselves that we're doing good things for the world, and they'll understand it, but that's actually not true.

There needs to be some translational element between what the tech companies are building and helping people understand what's happening, and then letting people decide if that's something they even want. A lab like ours is trying to help create more social and narrative conditions to understand where the public is coming from, what their priorities are, and serve them so that we can have legitimacy to continue doing what we're doing.

And I'll give you a downside example, which is: if you look back a couple decades ago, when genetically modified organisms were coming into play, those types of developments in science are actually crucial to things like food security. But a company called Monsanto was basically demonized in lots of ways for creating alien foods and things like that, and that actually set back research about a decade or more.

It's a public misunderstanding of what's going on, but the company never invested in really engaging the public and bringing them along, and I think we're at risk of that happening with AI as well. If we can't prove to the public that we have verifiable and obvious beneficial impact, why would they want this in society?

[08:24] Beatrice: Yeah, you can see the backlash happening a bit already, people boycotting AI and things like that. So it feels like it's—

[08:32] Dorothy: Yeah, it's a really hard subject to broach. I think that the assumption that people understand science and technology out the gate is actually one that's faulty.

It would be like assuming that any idea I had is immediately translatable to another person, and I think there needs to be much more work done at the intersection of an AI-curious and interested public and what the companies are building. It's about leadership and bringing people along on that journey.

If you look at NASA, for example, and what they've been able to achieve with the public fervor around space exploration and going to the moon in the '60s, that's the kind of thing that we want to be able to create — a societal project around progress. What does that look like?

Especially at a time when, if you look at polls in most developed countries, parents think their kids will be worse off than they are.

It's creating the conditions for the idea that we have agency and that positive change can lead to something greater. That's not an easy thing to do, but you need to do that if you want people to buy into what you're building.

[09:44] Beatrice: You have this post out where you write about how we sort of are funding 21st century science in a 20th century way.

[09:53] Dorothy: Totally.

[09:53] Beatrice: And one of the — maybe we could also just do a little brief on what the whole AlphaFold example is, if listeners aren't familiar, because this is an example that you keep coming back to.

We want to have more AlphaFold moments. So what was the AlphaFold moment?

[10:10] Dorothy: So AlphaFold was our Nobel-winning system, and what it did was — if you remember, during COVID, the reason it was called the coronavirus was because there was a spike protein, so it looked like a little sun.

The corona. That spike protein was what we were binding therapeutics to, and the shape of that protein really determined the ability for us to bind around it. What we were able to do was predict the entire world of 200 million-plus proteins in one single go, because we were able to take amino acid sequences and figure out what that determines about a protein's shape.

So it's really crucial. And now it was interesting, because if we look at what was happening before, a lot of the people working in protein folding sometimes don't recognize that that term or phrase, outside of their field, is not immediately legible to most people, but it is such a crucial part of — diseases like Alzheimer's, Parkinson's, and Huntington's all have protein-shape-related disorders associated with why these diseases are so difficult, and if we can unlock some of that, it would help us figure out better how to treat it over time.

So it used to take a PhD four years to find the shape of a single protein based on an amino acid sequence, and with AlphaFold, we could do it all in one go, all 200 million of them. About 3 million researchers in 90 countries around the world have accessed it and are using it.

And we made it available open source. But I would say what I'm most proud of is a lot of the early validation around AlphaFold we did with the Drugs for Neglected Diseases initiative, on two parasitic diseases, leishmaniasis and Chagas disease, which are both primarily found in the Global South, but the treatments for them are incredibly antiquated because there's just not enough money going into these things.

If you think about how much it takes to fund a four-year PhD just to get one shape of a protein — it's really interesting for me to see that there's a global impact and unlock when this happens, not just for the diseases that can raise the most money.

[12:37] Beatrice: So if we go back to the point that we're trying to fund 21st century science with these 20th century methods or structures, what does that mean?

[12:49] Dorothy: If you look at the conditions for why AlphaFold happened, there were three main ingredients. One is that there was a publicly available database called the Protein Data Bank that scientists had been contributing to openly for a very long time.

So we had a highly validated data set to use. We also had an agreed benchmark — all the teams that were trying to figure out how proteins fold or take structure could test their systems to see how effective and accurate they were, and it was called CASP.

So both a data set and an agreed benchmark and evaluation method. We also had a very clear problem definition — this is the problem, it's how these amino acid sequences take shape and fold into a protein. That was all already agreed, and there was this culture at DeepMind of not just working on single disciplines.

We work at the edges of each and try to understand each other and cross over. And that set of conditions is what enabled a Nobel-winning breakthrough. So for me, when you look at funding science that breaks the mold — like I was saying, usually it's linear, discipline-specific, very tight circles versus these expansive ways of thinking about things.

And I think science's next unlocks are all going to be at the intersections of these different disciplines. So you start to think through: okay, if you deduce from AlphaFold, what should we be funding? One is the infrastructure — the data foundations are crucially important for different fields.

Two, actual validation methods are also quite important, things like CASP. It's so important to have those community validators to make sure that it's not just building for building's sake, but building for effectiveness in those specific fields. And on top of that, really creating that interdisciplinary space and culture — that's really not how universities are built today.

It's not how labs are built.

[15:02] Beatrice: I'll share a quote that Tom Kalil from Renaissance Philanthropy told me once: "The world has problems, but universities have departments."

[15:13] Dorothy: Yes. And that's what we think of as the classical definition of a wicked problem. It takes people from all different expertise and disciplines to come together to actually think with each other.

And after years of working with some really amazing people, I still think that's one of the hardest things to achieve.

[15:30] Beatrice: And so what we kind of need to do to make this happen — I mean, you were touching on it a bit earlier — maybe VCs aren't going to fund the data sets that we need because they're not really incentivized to.

So how can we make this happen?

[15:47] Dorothy: I think it's hard, because sometimes I feel like when I work with people in academia, or more the public sector, they sort of demonize the private sector. And I would say if we just all suspend value judgment about each side and just try to understand the constraints and incentives that each different mechanism has, you start to see where you can unlock some synergies.

So what I would say is philanthropy and government can really fill gaps that VCs can't, because VCs have fiduciary responsibilities that are legally mandated to return to their shareholders. So if you think about their constraints, and then the constraints on government and philanthropy, you can start to see where the edges can actually create more of a flywheel versus being stacked against each other.

I do think that we need to think collectively about how we fill gaps and align incentives, especially when there are areas where there's both public and market interest to solve something. I'll give you an example — during COVID, going back to that we needed vaccines, and we needed them fast.

So what governments did was say, "We are going to guarantee that we will purchase these vaccines if you produce them." And then pharma companies went back and said, "Okay, we're going to see what we can do." And they worked more quickly than we've ever seen that sector work, to produce these vaccines.

Now, the question I have for governments is: why aren't you doing that across the board to direct where you want AI to go?

[17:34] Dorothy: What I'm seeing is a lot of governments say, "Here's all the things I don't want you to do with AI."

Whereas I would say it's more powerful to say, "Here are the sectors where we think, from a social impact perspective, you should go with AI, because this is what society needs and wants from a democratically elected government. This is what society needs and wants. We want AI to go in this direction. And by the way, to do that well, here's how you should build it."

Very different from a stick approach — the carrot approach. That's what I want to see more of. And I actually think that leads to a lot of very interesting dynamics in the market.

So BioNTech, which had one of our big COVID breakthroughs, bought a company founded by a North African founder called InstaDeep, an AI company. When I went to South Africa last year to talk to some of the students we had sponsored at Google DeepMind with their master's degrees in machine learning, they were all interning at InstaDeep.

So what's interesting to me is you now have this pipeline of students in the Global South, in one of the places with the most advanced research on tropical diseases that are inevitably going to move north because of climate change, deploying their machine learning skills towards biotech, because they know we're going to need all these skills.

And that's what that advanced market commitment towards COVID vaccines really unlocked. It's not just the single vaccine, it's the entire pipeline of what can take place in an ecosystem, ensuring that the highest talent in these regions goes towards problems that matter versus going towards slop.

[19:21] Beatrice: Yeah. Are there any other exciting projects that you see happening on this now? I saw, for example, that Astera Institute launched something to fund—

[19:36] Dorothy: Yeah, yeah. I think there are a lot of different types of new catalytic funding initiatives starting to spring up. At google.org, we worked with the team there to launch an AI for Science Fund that's now in its second year.

The first year we did 20 million worth of science grants, and this year we're doing 30 million. The projects we're financing — both for-profit and nonprofit, anybody can apply — we're looking at how we can catalyze more teams to be using AI-enabled tools in different ways.

So at the Innovative Genomics Institute, for example, they're looking at the intersection of AI and CRISPR to figure out how we can help cows be methane-free, which is apparently extremely important for the environment. There's a startup called Spore.Bio in France that is working on antimicrobial resistance and detecting it in a matter of hours versus days.

They were basically a contamination detection company that was AI-enabled, and the grant is going towards antimicrobial resistance and that sort of work. And there are other grantees, like the University of Liverpool, which is trying to find new materials that can help with carbon capture.

So I do think there are so many different interesting areas and interesting financing that can go towards that. Certainly Google.org is not the only one. But I think all of those things need to be connected eventually with thinking through what the overall ecosystem is for how we finance that, and finance it sustainably over time.

[21:27] Dorothy: And what is the role of government versus philanthropy versus venture, especially when what I'm seeing is a lot of money loves to go towards simulation — simulating which new molecules would be good for this or that, or new materials. But when we need to do validation in real life, in a wet lab, or constructing new ways to do manufacturing for new materials, that's where things start to get really hard.

It's too long for venture. Philanthropy's not sure if they should fund it. What do we do? How do we help more scientists bridge that valley of death so that we're not losing all this valuable innovation on the front end?

[22:11] Beatrice: Do you have any thoughts on how we do it?

[22:14] Dorothy: Yeah, this is the part that I'm most passionate about — trying to figure out how we experiment to fill the gaps within. It's the highest expense, building physical infrastructure.

Simulations are not as expensive, and so figuring out where government can potentially step in, where philanthropy can help bridge those gaps, where we can have some blended tools between philanthropy and venture to really go after these problems, is what I would really love to see.

I'm looking at, for example, questions about — if we need to onboard different types of technology into the NHS here in the UK to help patients with preventative health-related work, how do we finance that when government budgets year over year and wants to see results immediately?

Whereas venture, we know that it takes five to ten years, if not more, to see that technology really take root and then start to scale and grow. How do we make those conditions a lot more legible, a lot more attractive, and make sense on all sides so that you're satisfying the incentives and constraints of each funder?

It's something that I would love to welcome people who are experts in financial engineering to help with. I think it's going to be some of the bridge work of our time.

It's crucially important.

[23:42] Beatrice: Do you see that — you call this blended finance?

[23:47] Dorothy: Yeah, it's one way to look at it.

A lot of the companies that I have personally angel-invested in are in these categories, and what I see is that they use a combination of government grants, philanthropy, and venture to really bridge those gaps, get off the ground, and work on the problems that are most important to them.

I think if you go in only one direction, especially with venture, sometimes you risk needing to do something extra to generate revenue instead of focusing on the problem you originally wanted to solve. That's because venture needs a return within a certain time period so that fund can keep going.

So how do we figure out how to surface those needs and what's valuable to society, so that we can plug some of those gaps with different types of financing? I think we're going to need a lot more creativity in that area, and especially investors willing to try new things.

[24:44] Beatrice: Yeah, do you feel like there's some traction for this now? Because I feel a bit hopeful in that it seems like there are a few actors thinking a bit more on the meta level now.

[24:59] Dorothy: Yes.

[24:59] Beatrice: Yeah. And do you see this happening?

[25:02] Dorothy: Yeah, for sure. I think the fact that there's a lot of investment in companies like Cusp AI and Orbital in the UK that are working on material science — that's a very new thing. And I think it's one of those fields where some of the data sets we need don't yet exist, but the gaps can be plugged. And because we all start to agree on what good looks like, things can move forward.

These companies have raised very big rounds, which is what they need. And I think what we need to realize is that they may not return on a venture-scale timeline, but that doesn't mean it's not a good investment for society. Now, if it hits the point where manufacturing gets hard, who's paying for that?

Is it corporate partnerships? Is it some sort of philanthropy that's sustainable over time? Can we figure out a way to recycle philanthropic dollars — if there's a return, it just goes back into that philanthropy to redeploy in the future? That's the kind of thing I'm really excited to see more of, and looking into with some really interesting partners.

[26:05] Beatrice: Mm. And if we concretize this a bit — say we manage to have a Protein Data Bank for X — do you have an example of, in ten years, what is that, and what has been built on top of it?

[26:21] Dorothy: Absolutely. If you look at the commercial success that's come out of AlphaFold, we have Isomorphic Labs, which spun out of Google DeepMind, but we also have companies like Basecamp and Latent Labs.

There are so many others — you just look at the number of citations for the paper, and you can see the overall impact. I do think it's worth studying that to see: how much time and money did that take, and if we want more breakthroughs in these areas, what are the types of strategic investments from a government, philanthropic, and venture perspective we need to go into together?

How do we de-risk that for venture to really take root and accelerate? I think climate modeling is going to be crucial, and also a really good example of that. A lot of the breakthroughs we've had around weather forecasting have so many different types of impact, whether it's agriculture all the way to direct finance.

I think that's really crucial. Genomics — I would love to see a lot more done around that, and I think it could have a really big shift. One of the companies I work with is called Orakl Oncology, and they partner with the French public hospital system to look across everything from genomics to — they take cancer organoids from the French hospital system, GI, pancreatic, lung — and what they're trying to do is test those organoids against some of the top candidates in clinical trials and see what you would respond to.

And when I talk to those founders, what's wonderful about them is they're saying, "What I don't want to do is pretend, be overbroad, and say AI will cure cancer. What I will say to you is the one metric I'm caring about is: am I improving this patient's quality of life?" I think that's such a powerful thing, and something I don't see from many other types of founders, and that's what makes me so bullish on backing science founders.

They will give you the real deal. And I'm really proud to be working with folks like that.

[28:33] Beatrice: We talked about capital and coordination a bit — is there anything else?

[28:36] Dorothy: Yeah. I think the one thing I would add is that I'm not excusing venture capital. I think venture capital itself also needs to start thinking about how it makes some changes.

It's very easy — when I talk to traditional tech venture investors, they're willing to invest in bio as long as it's tech bio, as in it's a platform, because what they want to see is scale, like SaaS companies — how do you do things really quickly across the entire sector?

But then when you actually need to validate whether that platform is creating anything useful, the traditional biotech investors only really care if you have a molecule that's useful. So that doesn't really comport with each other, and you're stuck again with this valley of death in the middle between "I have a really interesting platform that I can finally get financing for."

If you went to traditional tech investors with a single molecule, they'd say, "I don't know what to do with that." And then you have these biotech investors who say, "I don't care about the platform, I don't even know if it's going to create anything useful — just give me the useful molecule." Even in the venture investing field, these two different types of investors don't know how to talk to each other.

So I think it's, again, getting people from many different fields and disciplines together, instead of judging each other and each other's expertise, to think through: okay, here are your incentives and constraints, and here are mine — how do we line them up in a way that we can make this maximally useful for society?

[30:08] Beatrice: Is there anything else that you think we haven't touched on in terms of AI changing the game — like the rules changing, or the playing field changing because of AI?

[30:27] Dorothy: Yeah, I think, like I was saying with the Drugs for Neglected Diseases initiative, I am so excited about the prospect for AI to level the playing field a bit.

What I hope is that, instead of money going towards the people who can fundraise the best, people with the best ideas have the opportunity to build and attract investment as well. I really think it could, if we do things right, level the playing field for a lot of diseases and areas — frankly broader than disease — that are desperately in need of financing and innovation, but wouldn't normally get that kind of attention just because they're not as sexy or the flavor of the month.

Something like AlphaFold opens up the possibility for firms that used to have to spend a lot of time and money on one thing to now really focus on building what matters.

[31:43] Beatrice: Is there a project that you funded or been engaged with that you really wish more people knew about?

[31:52] Dorothy: Well, the big one that people are talking about most, writ large, in these AI-and-science fields, is actually fusion and solving that. I think the output, if we can get it right, would be clean water. If we can use that as an energy source, I think it would really mean a lot for our future and for a lot of people. There's a company called Proxima Fusion, out of Munich, that I've invested in and gone by and seen the magnetic tape and everything else they're using.

And again, the hard part of the fusion conversation is actually doing everything in real life. It's interesting, because they don't really consider themselves a tech company — they're really a manufacturing company at the end of the day, trying to figure out how to put what they've discovered into production.

And Europe is a really interesting place for that, because we're much more open to small nuclear reactors. We have a better environment for it, and I think if we can solve that, it would do a lot of amazing things for the world.

[33:04] Beatrice: I agree, I agree. Another thing — when I was doing my research on you, you know, obviously running the Public Engagement Lab, you've thought about these sort of things, but you said that public imagination matters, maybe just as much as public policy.

[33:23] Dorothy: Yes.

[33:25] Beatrice: What do you mean by public imagination, and why is it important?

[33:30] Dorothy: Yeah, I have a bugbear about — I think public policy loves to look at the downside and basically play risk mitigation, and I think that's an incredibly depressing way to run society.

I just feel like we used to have leaders, elected leaders especially, where when you went into an election, you were thinking: well, this person is casting a vision of what the future could look like if we work towards it together, and this person is casting another vision, and my job as a voter is to choose between those two competing visions of the future.

I don't see that now as much. What I see is a lot of regressiveness — well, we should just go back to what that looked like, or everything is clouded in fear, like this isn't working for us, so we should just move in this backwards direction. And I find that incredibly uninspiring and unimaginative.

But I also think that when you're in a kind of survival state — I was just in Silicon Valley last week, and young people there are saying things like, "I have three to four years to make my money, or else I'm going to be relegated to the permanent underclass." And I'm like, wow, what a depressing way to think about your role in the world.

Stressful. Yeah, stressful. I graduated into a recession — I graduated basically the year Bear Stearns crashed. And in that moment in history, President Obama was getting elected, and we were going hope and change. I think there's just such a difference between those two types of mentalities, and what I believe is that when you believe, as an individual, you have the agency to choose and change the direction of something important, and to take responsibility for that — not just for yourself, but for everyone else —

We talk a lot in this AI space about having individual agency, but I think it's actually about collective agency. How do we bring people together in a very polarized environment and lead? How do we see across not just different political spectrums, but different areas — how do we see across science and policy?

Two very different fields, like we were talking about. How do we see across the arts and AI, for example, instead of just pitting them against each other, and come together to exercise collective agency over what the world should look like?

[35:57] Beatrice: Yeah.

[35:57] Dorothy: That's where I think we can make the most headway, and we need to harness public imagination for that.

[36:06] Beatrice: Yeah, this is a strong question for me as well, an important one, because I'm running this Existential Hope project, trying to get more people to feel like they have some agency and that we can actually build a better future. Do you have any ideas on how we can do this?

How can we turn this around a bit?

[36:27] Dorothy: Yeah, I actually think we need to build more institutions that sit at the edges of different fields and disciplines. I don't think we have enough of that. I think we're still in this — to your point earlier — departments, and then big problems. Who are the new types of institutions that are going to be bridge builders, fluent enough in both to —

or it doesn't have to be binary, but fluent enough in different areas, or curious enough about different areas, to bring everyone together? Who are going to be those facilitators? I think those facilitative types of organizations are going to be crucial to the future, but we typically don't have enough of them.

And if you look at, for example, the absorption and adoption of AI across critical sectors like health, energy, or education, it's happening very unevenly, and there are very few places dedicated to bridging the gaps between AI and these fields — both from a validation perspective, like, is this the appropriate way to use AI in this field, and also from a practical, step-by-step perspective: how do I retrain and figure this out?

So I do think establishing those types of bridgers is going to be crucially important. How to finance that, how to manage that is, again, going to be a very blended and interesting exercise for society to figure out, because it doesn't fit neatly into any of these given areas. And if it did, I think it would pervert the incentive structure a bit.

It shouldn't just be about sales.

[38:11] Dorothy: It shouldn't just be about this company or that company, or even — open source models have their own constraints as well. So thinking through what the right model for this is, how we build it together, and how we get people bought in and brought along on the journey is going to be really interesting.

I actually think — I've worked in AI in sports as well, and what I've found is that because the sports sector is much more open to data science overall, the adoption has been much more seamless than in some of these other fields. It's fascinating to see professional football leagues say, "Oh yeah, we were thinking about some tactical stuff on the field, and yeah, we can just integrate your open source system. That's what this would look like."

Versus — and I will also say it's a much less sensitive environment than health, obviously — but it says something about society if we can adopt something much more quickly in sports than we can in public health.

[39:11] Beatrice: Mm. Yeah, that's funny. I'd never heard of AI for sports. I'm going to have to look this up. Okay, so, last two questions. If someone young is listening to this and wants to start working on supporting this type of work, do you have any recommendation of where they can start?

[39:35] Dorothy: Oh, yeah. And this is speaking to more people like me. So some people are going to go off and be experts in different fields, and that's great. But if you're like me and you feel wholly inadequate about science because people around you are better at it, I would say it's okay to become the person who's fluent in a little bit of a lot of different things.

I would consider myself a generalist, even though I've worked for this long in science and technology. But basically, the things that you don't understand are also the things that other people don't understand, and I've spent my whole life trying to — because I had to explain to myself what my parents and my brother were talking about — later on, explaining to the world what the scientists at these different companies were talking about.

And that, I think, is a valuable skill, mostly because I'm curious. I'm curious about the people behind the science. I'm curious about why they're incentivized a certain way. I'm curious about why policymakers are operating a certain way, and I think that curiosity will take you to places that can be useful and very values-driven in your own way.

So I would say, instead of worrying about not being the best person at this or that, you can still find value in being the person who's curious about many things and connecting the dots between all of that.

[41:02] Beatrice: And it's so nice now that AI can also be your personal tutor, to afford you all these —

[41:07] Dorothy: Oh, yes. Look everything up.

Yes. Just look up — I mean, you have to validate it too, but look up all the things.

[41:11] Beatrice: Yeah, yeah. Last question is: what's the best piece of advice you've ever received, or just some good advice?

[41:18] Dorothy: Oof. For someone like me, who loves living life at a fast pace, seeing results and progress at a very fast pace, and puts a lot of pressure on herself to be driving that progress — some of the most important moments in my life have been when I've slowed down to speed up.

So really taking the time to — and it's not just for yourself, by the way — it's also because when you don't slow down, if you keep operating at that speed, no one can follow along with what you're doing. So on both a meta level and a personal level, I think taking the time to regroup and recalibrate with yourself, and with the people around you who you're leading, that will enable acceleration in a real way.

I think it's partly to understand yourself and be self-aware about why you're making certain decisions, as well as to understand the people around you. I think it's impossible to bring people along when you're not curious about them, why they're making certain decisions, and what motivates them, and what they see success as for themselves — it's going to be different for everyone.

So I think this idea of slowing down to speed up, in this weird 9-9-6 culture, is something that I hold very tightly, and I'm trying to be better at every day.

[42:48] Beatrice: Yeah, I think that's some great advice. And thank you so much for coming. It was so nice to talk to you.

[42:53] Dorothy: Thank you.

‍