America's Autism Crisis and How AI Can Fix Science with NIH Director Jay Bhattacharya

24:40You said a few minutes ago that a lot of the grants in the NIH are going to older ideas, and there's, you know, lots of data that shows they're also going to, you know, more established, you know, older scientists, you know, at the very high, you know, highly regarded institutions. Um, the equivalent of that would be if we only funded, um, 30-year executives that came out of, you know, large established companies and ignored, you know, the young up-and-comers, you know, coming right out of the university, or dropping out of school or whatever. Um, you've talked a little bit about, you know, that question, Like, how do you reform the process, the, you know, the execution, to use Vanita's phrasing, on selecting for the innovation that, if you will, bubbles up from the bottom?

25:26It's a hard question, actually. It's something that's at the top of my mind. And actually, what you just described is exactly what we've been doing in science for a long time. So, the data out of the NIH is that in the 1980s, if you were 35, you actually had a chance of getting a large NIH grant. Like that was the median age of the first large NIH grant. You were 35 years old. Now you're in your mid-40s. We tell young investigators, you've got to do first-doc one, first-doc two. To be clear. Mid-40s, super young. I just want to be clear about that. Super young. I mean, I'm 57, so they all seem like babies to me.

26:03But the thing is, you have, just like as in Silicon Valley, the new ideas come from younger investigators, right? So I did a study a few years back where I looked at, it turns out that the age of the ideas in your published work ages by one year for every year of chronological age. So my ideas get one year older every year that I age. The very best scientists fight like crazy to stop that. So every two years of chronological age for Nobel Prize winners, their ideas in their papers age by a year. If you want the newest ideas, you have to let the young people have a try. And we're just bad at that.

26:46Like young people, we fund them and then they drop out and they leave for other places. That wasn't true back in the 70s and 80s. The culture of biomedicine says you have to have one, two, three postdocs before you have a shot at an assistant professor job. And as a result, the ideas that we support are just, they're just older. I mean, that's not necessarily a bad thing. I mean, of course, you should, in the portfolio, have some support for older ideas that are still promising. But if you don't also fund some of the newer ideas, the portfolio is going to produce fewer advances as a whole than if you do, right?

27:24You have to diversify in that sense. To solve that problem is hard. So the NIH has been trying to solve this now for two decades, and we made no progress. So first, we have to, I think, I mean, I'll just give you some sense of where we've gone backwards. You know, we used to have a system of peer review where in order to be a peer reviewer, you had to have a large grant. Now think about that. I got a large grant. I'm in my 50s, and I see an idea that challenges my 30 years of work. and I'm a reviewer on a panel, it's really hard to like open your mind and say, well, I might've been wrong. That system, that got changed, but now that's so that we don't longer have that rule, but like it's the mindset you, you have to allow.

28:19So what I've done is I've asked the institute directors, I've given them the authority essentially to expand what they can do in terms of the portfolio. I'm not going to judge them to make, just like within Silicon Valley, I'm not going to judge them on does every single grant succeed. I'm going to judge them on the portfolio as a whole. Does it translate over to better health for the people, for the disease that they're trying to address or the diseases they're trying to address? Does it result in big advances in biological knowledge? I'm going to assess the portfolio as a whole. And then the other thing is that does it match the strategic vision of the institute?

28:54The institutes have these fantastic strategic plans. You go look at them and say, your eyes will say, You look at them and you go, like, your eyes will get big with the science that they're proposing. And yet what they actually end up funding based on their peer review panels is often you'll get 10 great proposals on one part of the strategic plan and, like, nothing on another part of the strategic plan. And so, like, I'm going to encourage them to be able to pick the portfolio so that it matches the strategic plan. I'm going to reward them for rewarding and empowering early career investigators more.

29:27Right. So I'm going to build incentives into the decision-making by the institute directors so that they have incentives to solve these longstanding problems. We have to solve the new investigator problem. And I'm going to start to evaluate long-established investigators because I do believe they play a pretty fundamental role still. But in like how well do they advance the careers of the early career investigators that work with them, right? So if they're good at that kind of mentorship and career advances, I'm going to reward them in their grants. I'm going to start evaluating the grants for that too.

30:03Because the grant portfolio has to be sustainable in the long run in producing new ideas. I mean, if we don't have the early career investigators sort of getting the support they need, we're going to start to stagnate. I love to hear the interest in advancing early career investigators, but we can't have that conversation without talking about the universities from where they tend to come. And so, you know, I was a product of NIH MSTP funding. I did my MD-PhD with the generous support of the NIH. And my peers and colleagues, you know, in my class and, you know, decades behind coming up get trained on those grants today.

30:47How can you work with, you know, with the administration to ensure continuity for the training grants that, you know, NIH does believe are going to fuel the pipeline of early career investigators who, as you say, you know, are perhaps most likely to bring change, big ideas, you know, and take big swings. Yeah, we, as you know, you're biophysics, right? So we have a range of ways that we support early care investigators. So there are these awards for pre-docs, pre-docs meaning undergrads. And that's really important. We want to make sure that the very talented undergraduates who are interested in biomedicine and research biomedicine have the support to do this.

31:40if there's also like support for postdocs, right? So for people who are getting their PhD and then postdocs, I want to, it's going to be hard, but we have to structure things so that the range of investments we make actually translate over to people wanting to stay in biomedicine. I mean, we have a lot of people who drop out, but I think the RAID problem isn't that support for the early, I think we have a lot of portfolios pretty good on that. We could do better, but that's pretty good. The problem is, like, after you've had this career in biomedicine, how do you, like, do you have this research training?

32:21Do you have support to, like, make the next leap into an assistant professor job? And too often, it's too hard to do that. You can't get the support you need to do that. There's these K awards that we have that it's really difficult to get them. I think we have to do better at that. And we have to reward universities that are better at that. There's problems all across the system, but I think that missing link is really the, you know, you finish your MD and your PhD, and then can you get that assistant professor job? Or are you going to be asked to do 17 different postdocs before you have a chance?

32:56Right now, that system is set up to make it difficult. You mentioned earlier that we're not making advancements in life expectancy. Why are we lagging? why are some European countries doing better? And what are the highest leverage points you think to get back to improving there? Well, I think the key thing is we have to, a lot of our science is, you know, this replication question we talked about earlier is very important. We have to solve that. That will help a lot. And then this portfolio thing, I think both of those things actually will address the scientific rigor problem and the sort of conservatism problem.

33:33as far as like addressing life expectancy that that really needs to be it's in a sense not just a scientific problem like we have to we have to essentially get a message from the people that we want that they want scientists to address those problems like that's just we talked about earlier the political political um nature of of of that kind of allocation decision um but you know that's exactly what the mahab movement represents the mahab movement is a it's basically a cry for help from the American people saying, look, all these crime and disease problems, all these problems with our kids and we're sick, we're doing much worse than folks in Europe in terms of our health.

34:12And that essentially is a call for the NIH to reform itself to address those problems. And to me, it's a tremendous opportunity. And this is why I agreed to take this job. I mean, I was perfectly happy being a professor. But it's a once in a lifetime opportunity to make really work for the American people. You know, and I think having that political movement behind us is really important for that. Last week, you announced some really interesting initiatives around academic freedom. And many folks know your voice kind of reached the national stage, in part because of your ardent desire to see academic freedom respected, protected across the country.

34:57And, you know, it sounds like you're looking for ways to improve publishing fundamentally so that people feel freedom at all levels, including early career investigators, to share their view on science that they think might be interesting. And we need to figure out, to your point earlier, how to make the point that anything published is not necessarily fact, but it's one opinion, backed by one set of data and one set of analysis and one set of perspectives, and you'd like more of those to flourish in the public arena. Say more about the role that you want NIH to play in protecting academic freedom.

35:35First, at the NIH, I found out that a lot of folks, the internal investigators at the NIH, in order to publish their work, had to seek permission from their supervisors. I changed that. Like, no more permission. If you're an NIH researcher, you have a scientific paper, you don't have to get permission from me. People are going to publish research that I don't agree with. That's wonderful. They should be able to do that. Also, the places that like the universities, I think, need to be absolutely committed to academic freedom for excellent science to happen. And, you know, like there's been a lot of like angst over the administration's actions with the universities over the last few months regarding holding them to high standards regarding, you know, anti-Semitism and so on.

36:22But there's also been a message that we really do want academic freedom at the universities. Scientists will be able to say what they think and explore where they will, or else they're not good environments for research. As far as journals, that's a complicated question, but the problem right now is that the scientific journals, there's essentially a duopoly. Very few companies, for-profit companies, control a very large number of journals. and they charge tens of thousands,$10 ,000 per article for science that they didn't do, that the American people paid for. They actually had a sort of a policy where if a regular person wanted to go find a scientific article, they had to go, there was a paywall where they pay like$50,$100.

37:11We got rid of that paywall for NIH-funded research. There's still a lot to do in this area. We need more academic freedom. and we need more openness in scientific publishing, and I'm working on policies to do that. So, Jay, you know, one of the... one of the key questions, you know, for the American public, they're looking for better outcomes, better health. One of the big avenues that, of course, this country uses and has really had as a gold standard in the past is, you know, having this extraordinary public health infrastructure. But I think what's also true is, over the course of the last several years, there's a lot of mistrust now in terms of public health.

37:48Um, how do you sort of rebuild that trust for the public? Because obviously, you know, if there's no trust, the message can only be so effective. And so how do we build those bridges back to the extent that you think they need rebuilding? You know, I think the problem with public health and the lack of trust in it, you have to point to the pandemic. You have no choice, right? If you look at, you think back to the pandemic, and you remember the plexiglass that was everywhere? There's still, every time there's a sea of plexiglass, It fills me with rage, but that's another story. And there was no science behind that, right?

38:21There was like the, you wear a mask when you walk into a restaurant and you take it off when you sit down. You know, again, no science behind it. A whole host of like things, and they were really damaging things like closing schools, where again, the science was so weak that it, and now kids are like years behind in their education as a result. And they'll be paying the price for that for years. and so a lot of American people have lost trust in public health for reasons I can completely understand and so the question then is what can we do about it and to me the key thing is there's two things that have to happen like two very broad things like one I think we have to restore gold standard science like that presidential EO on gold standard science is so important because what it says is it articulates things that we thought all scientists already knew or committed to.

39:14Like replication is really important. Unbiased peer review, humility in how we talk about the limitations of our scientific findings. There's a whole host of things where you read it and go, wow, I thought science already did that. So if we actually do that, I think that's a major part of this. The second thing is we have to, just like we talked about earlier, about the role of the people and politics in deciding what areas of science to fund. and then scientists to decide what priorities within the science areas to fund and the portfolio analysis. We have to convey to people that we are their partners in scientific investigation and in public health.

39:56Public health, folks in public health are servants of the people. And too often during the pandemic, it came across like we were sitting above people, right? Telling you what to do, telling you if you don't take this vaccine, you can't go to work, you can't get a job. But, you know, I mean, it was heartbreaking to watch because if I believe very fundamentally that when science works as partners with people, like, has this almost servant attitude toward people, you can do a lot of good. You can do a lot of good. But I think, really, that kind of humility and a return to sort of gold standard science, that's the way to solve the problem of trust.

40:35It's gonna take a long time, though. Because, I mean, I've talked to so many people around the country, and it's not... We're nowhere near solving that public trust problem. Dan, I think it's an especially challenging thing as you look forward, and I'd love to hear your thoughts on, you know, how do you convey, you know, recommendations and guidance in the face of uncertainty and incomplete information, right? Because going back to your point, like, in an ideal world, you're always resting on top of, you know, gold standard science, you know, but, you know, a lot of times, science, you know, There's a lot of unknowns in the science.

41:11Science is hard, going back to what you were saying earlier. And so, how do you communicate to, you know, a population, a nervous populace, you know, a sense of a recommendation or even guidance in a world where you yourself have incomplete information? I think you just have to be honest, right? So, if I asked a question about... I mean, God forbid there's another pandemic during my watch, and then I'm asked, okay, how should we manage... is it right to wear a mask or something, right? And I don't, there's no good scientific evidence. I'm going to just say that. You know, the analogy I grew up, I was a medical student once, I was an MD.

41:51So like, I can tell you this from firsthand experience. The first two years of med school, you do a bunch of classwork. The third year, you finally get to see patients, right? So you go walk into a patient room and you're wearing a white coat and you know nothing or very little. You know, you could fill with knowledge about biochemistry, You can write chemical equations so your fingers get tired. But what you can't do is understand what a patient really needs. And so you sit down in front of the patient. They tell you their story. It's wonderful. They put their trust in you. And you are tempted to tell them things to answer their needs that they're asking you.

42:29But you don't know the answer. You just don't. Because you're a third-year med student. Of course you don't know the answer. And there's a, like, because you're wearing the white coat and because you have someone looking at you wanting the answer, putting their trust in you, you feel this urge to, like, say things you don't know. You know, you start, like, freelancing. And that's just a terrible mistake, right? As a third-year medicine, you learn that you should just say, I don't know. I'm going to look it up. I'll look the answer for you. I'll get back to you. I'll consult with people who know more than I do.

43:00You have to be humble. and especially in the face of new things, you know, new pandemic or genuine scientific uncertainty, we in public health have to be humble and say, look, we're not sure, but here's how we're working to try to get an answer. And we have to convey that uncertainty and we can't blame the public. I've gone around and talked to lots of folks in public health and science and they're like, well, what we have to do is we have to teach the public more about science and make sure they understand that science isn't always perfect and science moves, you know, you may have eggs are great one day and eggs are terrible one another day.

43:39That's because we have new science. To me, that's like blaming the public. It's not that the public doesn't understand that science is hard. They understand it fundamentally. Like, this is not a complicated thing in the sense of, like, I mean, everyone knows within the public that science is hard. The problem is that scientists conveyed certainty about things they had no business conveying and then changed people's lives for the worse as a result of it during the pandemic. I acknowledge that there, you know, that the pandemic was a particular challenge with respect to both communication and certainty in the midst of uncertainty.

44:17But how do we acknowledge that challenge and not lose trust in some of the bedrocks of public health advancement that we've made over the last several decades, whether that's newborn vaccinations, you know, HHS held a listening tour and an advisory update on Hep B vaccination in babies, and it's great that we're looking at all of the data holistically there. But in some of those cases, you know, some folks would argue there is substantially less uncertainty than there was in the wake of a new pandemic with a new virus with no data, with completely new infections, than there is in the context of something like a hep B.

45:03So how do we, and please don't feel the need to respond to that specific vaccine example, but how do we not make it so that even when you do have relative certainty and you come out and say, hey, this is not perfect, but we're pretty darn sure this is a good idea. How do you then make it so that people don't say, well, you know, last time you said you didn't know, so I don't know. Right. So I think I don't know is a good answer when you don't know. When you have a little more evidence, a lot more evidence, like just take the MMR vaccine. I mean, if you want to prevent measles, take the MMR vaccine.

45:40I mean, it's the best way to prevent measles. And measles can be a deadly disease. Like I vaccinated my kids with MMR. I was really happy I did. me too and I think that that you know I think that that kind of certainty you know it's science right so nothing is known like tomorrow that someone might come along and they overturn you know Newtonian physics and all of a sudden you're talking about relativity or something right you'll always leave open that possibility but some things we do know with like much more certainty I'm not saying that we should all have false humility I think we should have humility for the things we just should actually have humility about, right?

46:21But at the same time, when we have an area of more scientific certainty, we have to leave open room for academic freedom so that people can have their say, think differently. We don't cancel them. We just reason with them. And we say, look, you say X, Y, Z, but look at all this other evidence. MMR is a good example. Look at all this other evidence that shows you differently. and they'll just have a public discussion. It's okay. I mean, it's okay to have that contradiction. And then I think what we'll come across is when there is actual excellent science replicated, maybe I'm naive, but I don't think so.

47:02I think that wins scientific debates. And you can look, there's evidence for this, right? So the uptake of MMR in this country, the MMR vaccine is like 95 % of American parents vaccinate their kids with MMR. The evidence is, and I think it's like 13 % of American parents vaccinate their kids for the COVID vaccine. I think that reflects the scientific evidence regarding the relative merits of those vaccines. The American people are not stupid. In fact, they're quite smart. And when we talk to them in ways where we show respect for their intelligence with data, allow people to disagree, but then have the evidence right there in front of people.

47:41So I think people will respond with trust where the evidence actually leads. I mean, maybe that's just a matter of faith for me, but I don't see any other way forward. You mentioned that the three priorities of NIH that you have are nutrition, chronic disease, and integrating AI. Maybe can you flesh out a little bit on the last two, what you see as most promising in terms of reducing the disease burden and then also in terms of integrating AI? I've seen some fantastic new ideas regarding Alzheimer's disease, for instance. A colleague of mine at Stanford has this fantastic set of papers he published using an old shingles vaccine called Zostavax.

48:22He found that in excellent observational studies that if you had Zostavax, it reduces the likelihood of developing cognitive decline for Alzheimer's disease by up to 20%, 30%. I mean, it's pretty substantial for a pretty innocuous safe vaccine that's no longer used, actually, because it didn't work for shingles. I mean, imagine if you had a very simple, cheap way to prevent 30 % of Alzheimer's cases or delay Alzheimer's for years. There's all these, like, huge advances I've seen that, you know, just need a little bit of scientific love. I think we just need to focus on those, make our portfolios focused on those, be willing to take risks in terms of like on things that look like they're new ideas.

49:06And we're going to make a lot of progress. And AI, by the way, I think is going to play a tremendous role in that. I just, you know, everyone knows about the protein folding and alpha fold. That has done an amazing job in turbocharging biomedical drug development. Because now you don't need to like sit there and wait. You can just do your computations, figure out how the protein folds, what the target size will look like, and then ask which of these drug products are more likely to actually work without having to do very expensive biologics in lab work. You still have to do the lab work, but they focus on lab work in more promising ways.

49:46in the way that we deliver medicine, right? So you can have AIs help radiologists do a better job at making sure they catch things, catch everything. Even simple things like, you know, you go to your doctor, the doctor sits there looking at the computer the entire time rather than you because they're like filling out their electronic health records. Have an AI assistant listen to the conversation, fill out the form for the doctor. So they're just checking afterwards, taking them a couple of minutes, and they're spending all their attention on you, right? All of this needs research, by the way.

50:19I mean, is this going to help patients? We have to ask those questions. But to me, that's a tremendous promise. Like those simple things can transform biomedical research and how patients are treated. So that's why AI is so important to me as a potential tool. We does need research. I mean, I don't want to, we can't have AI hallucinating on us and then treating patients based on hallucinations. But, you know, that's a matter of research to fix those kind of problems. We heard that HHS rolled out across agency-wide an enterprise-secure version of ChatGPT, which seems like a terrific achievement from the perspective of internal HHS and NIH operations even, to be able to look up internally how new is an idea.

51:06Simple queries and data fluidity of that kind seems important. what's the future? Is an AI going to write the Institute's strategic roadmap and an AI submit a grant and an AI review panel review the grant? And, you know, where are we going to play a role as scientists? I mean, I don't. Okay, so that question. The answer is no. Yeah. I mean, I think AIs are really good at summarizing existing knowledge. The training data you give it It helps it. It's fantastic at that kind of thing. Really developing brand new ideas that challenge existing paradigms. I don't, I mean, I don't know your experience with AI, but they're not quite as good at that.

51:54It's really, we have, just to put a new policy in place where I'm limiting the number of new public applications you can have, like we can have six to cycle or something. We have people writing 60 applications, and very clearly AI generated. And then we have, you know, it's, I mean, what it does is overwhelm the system of noise. Yeah, so I think AI is really important, as I said. But we have to do research to understand how it can be used to help people. And I think people, scientists are still going to have a tremendously important role. The new AI system rollout in age is just exciting. We're actually been working on a new system also specific to NIH, again, to protect in ways that protect patient privacy and all that, but rolled out across the NIH so that people can interact with it in ways that help on NIH-specific tasks as well.

52:52So, I mean, I think that's all very exciting, but it's an augmentation of capacity rather than a substitution of capacity. It'll make people way more productive. It'll help us address some of the key problems, but scientists are still going to... I mean, we still have work to do as scientists. We do. If I could just end on one last question. If you had one message for the rising star scientist contemplating a career in science where they can bring the best of their abilities to making science better, smarter, faster. A scientist embarking on a new PhD in a brave new field. A scientist thinking about starting a new company to advance the work that they're doing.

53:36A scientist at the NIH running a lab. What is your one message to the individual scientist who's out there, you know, hoping to make the biggest impact they can? I mean, science is incredible. Like, it has almost limited capacity to advance human well-being. And it's the individual scientist who believes in their idea keeps knocking on the door even when the door is closed over and over again until it opens. That's who really makes a big difference in this world. I would say, please stay in science, keep knocking on that door, and change the world with it, because that's the only way the scientists can do that.

54:21I love this story of Max Perutz. I don't know if you've heard of him. He was a University of Cambridge researcher in the, I think, the 50s. And he had this idea that he could figure out the structure of myoglobin. It sounds like a very geeky kind of thing. It's like, but it's, uh, but back then there was no protein folding field really. I mean, it was like, and he was a student and all his professors kept telling him, pick an easier problem, Max, this is crazy. Why are you spending all your time? You're never going to finish. And for a decade at the university of Cambridge, you wander around, everyone knew he was a genius, but he was like, got nowhere.

54:50He's not just working at it until finally he figured it out. And it's just transformed like a whole host of things in biomedicine. Um, uh, and, uh, you know, eventually won the Nobel Prize. It's the kind of thing where I ask myself, do we have a scientific sort of infrastructure today that would allow a Max Perutz to do what he did back then? And I would love to make that happen through sort of the power of the NIH, to allow the Max Perutz of the world, the new ones who are now sitting there with great ideas, to be able to try them out and change the world with them. Fantastic. So maybe on that note, just looking to the future, if we end where we started, where, you know, you talked about the NIH's highest ambition is to improve the health of the American people, whether that's measured in life expectancy or the rate of chronic disease that Americans suffer from, if you had to guess where we're going to see the biggest and best gains, is that going to come from, you know, how we manage patients?

55:58so the management of disease, you know, new molecules for treating disease, or modifications in terms of how we all live? Yes. Yes, yes, yes. Yes to all of the above. I mean, you know, I am a big believer in portfolios when I have uncertainty, so I don't know how to answer your question because I see promising advances in all three of those topics, and I think we have to invest in all of the above in order to see where the most promising things go. Like, who would have predicted that the GLP-1s, you know, would... Actually, we saw a reduction in average body weight in this country the first time in, you know, decades last year because of a GLO monster molecule that somehow turns out to, you know, if you just do the right biology.

56:48There was a scientist knocking on some kind of door to make that happen, right? Yeah, I just... I mean, that's the only sad thing about science. it's hard to predict where the best things are going to happen. So you have to have a portfolio. But all of those areas, to me, look like they're very promising. And as I've gone around the country to talk to people, I'm excited about all of it. So I can't wait to see what we produce. Do either of you have a prediction to that question, or is it also... Well, this is the debate we have every week in terms of where we want to invest. Our answer is yes, yes, yes, too.

57:18Correct. All of the above. Well, it's a great place to close. Dr. Bhattacharya, thanks so much for coming on the podcast. Thank you. Thank you so much. Thanks for being here. Have a great day.

57:55Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see A16Z.com forward slash disclosures.