Pod Insights a16z Podcast Why America Must Lead in AI Investment About this episode Senator Todd Young (R-IN) speaks with a16z General Partner Martin Casado about the importance of open innovation and American leadership in AI, and why we need to support AI research at all levels — from the classroom to the war room.
To learn more about the American Dynamism Summit, visit our website at a16z.com/ad-summit.
Visit the full playlist: https://bit.ly/3IqWn1W
Transcript 0:09 Speaker 1 Great. Well, we have a lot of stuff to cover on. We're going to cover, you know, in general, like how policy intersects with AI and how we're thinking about it, you know, and how do we keep the United States ahead. As part of that, you had this, you know, Senate AI Insight Forum. I don't think everybody knows what that is, so I thought maybe you could first just kind of describe what that was, then we can talk about kind of like what the follow-up has been.
0:30 Speaker 0 Sure. Well, uh Martin, thank you so much for having me. This is uh a great event. I understand your speakers uh you know shared some great information today. But uh for those who may not be familiar with how Congress typically does its business when we're uh when we're really following the normal processes, we work through the committee process. So if we had gone the traditional route and trying to address artificial intelligence issues, that would have meant that a number of committees uh of jurisdiction would have started to hold hearings on artificial intelligence in a very public way. Members of their staff would have uh prepared uh members uh you know hours before uh each given briefing, and uh we would have learned what we could from those experts who briefed us over the course of two or three hours.
1:20 Speaker 0 But um oftentimes members aren't able to dive deeply into uh the the essence of of various policy issues when we take that approach. So instead, Senator Schumer, working with three other senators decided to deviate from the process. Instead, uh we held first three briefings and then nine, as you talked uh mentioned AI insight forums. These were uh these were round table events, uh long tables full of of talented innovators, entrepreneurs, uh policy makers, and others. No cameras were present, as is typically the case uh uh in our in our hearings, so there's a lot of candid conversation on dedicated topics. We held uh insight forums on national security, uh alignment, innovation, and other topics, and uh members were invited to witness these guided conversations, uh, take extensive notes, and then after that, uh we have a lot of work product that we can deal with, recommendations, uh policy concerns and in and other things that will lead to actual committee hearings and legislation being proposed as is traditionally done.
2:23 Speaker 0 So those were really constructive. Members uh gave a lot of positive feedback about it, and we learned a lot in the process. What are your key takeaways having said through them? Well, two really. I would say uh the first is is uh just a general sense uh that this isn't a particularly partisan issue. I would say one could probably detect slight differences in approaches when it comes to regulation versus innovation uh as you go uh consult with members on either side of the aisle. But as of related to this topic, I think, and this gets into the second issue. There was a general embrace by my colleagues after they heard uh from some of the world's best minds that we needed to regulate, but it would have to be a light touch approach.
3:08 Speaker 0 We had to be very careful not to uh go overboard and constrain what is right now uh uh uh uh sort of leading edge industry and the United States is is in the lead. We want to keep it that way.
3:20 Speaker 1 Yeah. Uh do you think we can expect some legislation framework come out of this? Is there a time frame, or is this still kind of too early to
3:29 Speaker 0 Well, I think we can. I my my hope is that we can expect all kinds of legislative efforts, not just this year, but in coming years. I don't think we have to rush to do everything at once. Yeah, yeah. Uh we need to get clarity on on uh a lot more uh of people's concerns, but more importantly, uh their uh uh what they perceive to be their opportunities as the technology evolves as we learn more about the technology as as new versions are are released. So uh sort of a wait and see approach. I do, however, think that uh there will be some things we agree on now. We agree that there needs to be some resident expertise within uh the White House on an ongoing basis uh that other departments of government look to when it comes to uh uh issues surrounding artificial intelligence.
4:17 Speaker 0 I think we need to uh revamp our human resources uh approach so that we embed more expertise in every agency of government, and we can we can dialogue with businesses, uh uh uh consumers and and others uh about uh things within a department's field that touch on artificial intelligence, and then there'll be some things as you heard about earlier today that deal with national security where we're already behind.
4:43 Speaker 1 Yeah.
4:43 Speaker 0 We need to make some key investments in in people and platforms. And and so all of those things are are among the things that I think in the coming months will be addressed through the committees of jurisdiction in a bipartisan way, if politics doesn't get in the way. And uh as it relates to most of this, I don't I don't think it will.
5:02 Speaker 1 So I um this is a bit of a personal story. So like I did my undergrad in Flagstaff, Arizona, like small Arizona town, and at the university, which is like not a notable university, there was like a lot of investment from DOD in in supercomputing. And like they've like bought supercomputers and I worked on them as a researcher, and that led me to a job in Lawrence Livermore National Labs, and then at Lawrence Livermore National Labs, I worked in the weapons program. I worked on these like huge supercomputers with a bunch of other people that came from similar paths. And I remember thinking at the time, like, isn't it amazing that like we live in this country where people understand there's new technologies that are like you know, very powerful, and they invest like all the way down to Flagstaff Arizona to like stay ahead and you know, you know, find you know the people that are very interested in doing that. And it feels to me having gone through that, and then you know, having gone through um, you know, also the intelligence community, which also embraced like the internet and like a lot of the network technologies that like we're it feels to me like we're in a bit of a doctrine shift now that like when new technologies come, like we kind of are more afraid of them than wanting to get ahead of them.
6:04 Speaker 1 And so like I just wanted to ask you is my perception correct that there is had been a bit of a doctrine shift in the United States where we kind of worry about the implications of technologies before we're able to actually harness them and and become leaders in them, or is that you know maybe just like a perspective from the outside.
6:21 Speaker 0 I think it may be fair. I I I I think this is one of the reasons these AI insight forums were helpful uh to sort of keep certain colleagues honest uh that that might have been right ready to move forward aggressively and and with various regulatory actions. I can't promise that won't happen in certain discrete areas. But um I I I think there's a tendency uh to catastrophize different technology areas as
6:50 Speaker 1 opposed to different potential outcomes.
6:52 Speaker 0 You know, historically, the the reason uh uh one of the many reasons uh we've been a dynamic country and created a favorable regulatory atmosphere uh for uh our technology developing businesses is um we we have laws that reflect our values and uh we apply those laws to technologies only if they run afoul of certain prohibited behaviors that normally don't even mention the technology, right? And uh if we can take, at least for the most part, a tech agnostic approach as it relates to uh AI development and adoption, I I think that uh that would be more helpful. So no special carve-outs. Um special uh you know benefits uh except when explicitly those rare instances where uh there's a compelling argument that we need to.
7:45 Speaker 1 So I I I got a sense, you know, I so I I went to you know one of these kind of uh the the these AI meetings. Um and there's a lot of like kind of concern around kind of AI, and I'll listen to the concern. I'm like, wait, that's not AI, that's like all computers. Or it's like that's not that's not AI, that's the internet. Like, you know, someone
8:03 Speaker 0 that's
8:03 Speaker 1 been in these things for a very long time. And so I I've gotten this sense that like maybe there's a response to like the prior battle. Like people felt like we didn't get it right with the internet, and therefore they're trying to like, you know, use that consternation now, but for a different technology.
8:21 Speaker 0 Well, I think that's absolutely happening. Yeah, yeah. It's happening with respect to data right now. Well,
8:25 Speaker 1 let me tell you what the dangerous of the data, so so ha having been through the rise of the internet very closely. There were things that were very unique to the internet, right? Things like asymmetric attack, right? The more that you invested in the internet, the more vulnerable you were. And if you're dealing with a terrorist threat, that's a big deal because they didn't have infrastructure to take down. We did. So that was a very particular thing that impacted like you know, defense. Um there's other things like exponential growth, like, you know, things can get out of control and it's everywhere. There was these huge implications. I will say they actually don't exist with AI, but somehow people are imbuing AI with it. And so I'm just wondering if like, you know, there's maybe like a lack of like literacy, or maybe people feel like they've kind of, you know, didn't really do it on the previous war, and this is being applied now. And if that's the case, what can we do to educate or kind of get ahead of it?
9:10 Speaker 0 Well, this is helpful. You're in the right town, right? Uh but but to the extent some of you have not been visiting with uh members of the Senate, members of the House, who will soon be considering, you know, we'll we'll be playing with live rounds, as we said in the Marine Corps, right? We'll be considering actual bills. We'll be developing bills, and we'll be voting on them. And there will still be some members who haven't done extensive homework and and are uh uh carry uh uh superficial views of what constitutes artificial intelligence uh technology. And uh we need all of you to to disabuse them uh that of this being you know an extension of social media, this is very different. Um data, we're gonna have to think about data perhaps in a different way.
9:59 Speaker 0 So I I I I do think that some are either confused. There are others who are just probably using the opportunity of this legislative effort, which is ahead of us to lever to try and pass bills that they had prepared two decades ago. Yeah, yeah, yeah.
10:15 Speaker 1 So I was like, now's my time. I've done it in other contexts. Yeah,
10:18 Speaker 0 yeah. But it if we do it in this context, it could be quite damaging.
10:24 Speaker 1 And and how much is the economic reality hit the calculus when it comes to these discussions? Because listen, um none of us really know the future. I will tell you when economics change this much, right? When the marginal cost of creation goes to zero, this is going to be another 10x expansion in in value for sure. And like historically the U.S. captured that, right? That's like why, you know, the Internet age was so great for us, right? And so, like, is that part of the calculus, or is the focus really on like how do we keep this stuff from like like hurting us?
10:52 Speaker 0 Well, we tried to make this uh part of the calculus. In fact, we spent at least half of the time, perhaps we should have spent even more on the uh AI insight forums discussing the upsides. Yeah, yeah. It's not as though we had to address the upsides. The upsides are going to be a result of innovators and entrepreneurs uh and and investors. We understand that, but I thought it was very important for us to focus on that, to provide some balance to the staff members and and my colleagues who were uh present. So constantly reminding people of that I think is going to be very uh important. There are, listen, um there are uh I think what you may have characterized in the past as some f fantastical, some fantastical um doomsday scenarios, and there are a lot of those. Yeah, and there may even be a couple that are real, and we need to hedge against, right?
11:40 Speaker 0 And we need to take those seriously. But we cannot become so fixated on those that we lose the force from the trees. And um, you know, we can do this. We can do this. It'll be complicated, it'll be challenging. We won't get it all right in the beginning. Uh uh, and we shouldn't try and tackle everything in the beginning, which is a bit of a concern. Uh but uh I think we'll get it.
12:02 Speaker 1 It'd be great. It would be great for you to talk about like on the positive side, your takeaways from the discussions on how AI can benefit like America that's not just strictly economic, like you know, from like an innovation or a tech or uh, you know, daily life or whatever.
12:18 Speaker 0 Um we we can rapidly, and it's it's very difficult for us to quantify some of these things. I could sound very specific and learned by just throwing out figures. Within X years we'll we'll solve cancer, but no, we're we're gonna rapidly compress uh the treatment and drug development uh timeline, something I've heard from uh our pharmaceutical uh makers. We will come up with drought-resistant, uh climate change resistant crops to help feed the world, become more productive, drive down the cost of food. Uh we'll develop the ability to uh uh uh have tailored, personalized uh tutor services and mental health services, uh, the vastly expanding uh the workforce in those areas, leveraging technology so that we'll have mental health health providers in areas uh that we don't.
13:09 Speaker 0 We can clean up the environment in all sorts of creative ways. Uh ways that our artificial intelligence technologies will uh, you know, they'll they'll illuminate uh some of the solutions uh to hard uh challenges. We'll have ways of of reading health records uh so that we can uh discern probabilities of of getting certain types of infections, uh certain uh uh prophylactic measures we can take to extend uh improve and and save people's lives. Uh we don't have the benefit of those uh AI sort of uh insights. We can decrease the rate at which mistakes are made within a uh medical healthcare context. So I mean it goes on and on. And uh this is what we should be talking about.
13:54 Speaker 0 This is frankly, um I'm I'm sitting on a uh commission right now pertaining to synthetic biology, and so much of advanced biology these days really involves the use of of artificial intelligence technologies, and and and so I have charged the commission with educating me and members on specific things, innovations from material science to medicine uh to environmental health that I can I can start touting because uh that's gonna be very, very powerful for my constituents and and also for colleagues.
14:28 Speaker 1 You know, as an outsider, it feels to me like the general discourse is getting more sane. And so I've actually gotten a lot of faith in the project.
14:37 Speaker 0 Well, I I tend to make people feel that way.
14:39 Speaker 1 With you, but I mean in general, I just meant the national discussion. Yes.
14:43 Speaker 0 No,
14:43 Speaker 1 but the natural discount, it does feel like I saw you
14:45 Speaker 0 meant in this conversation.
14:47 Speaker 1 Also this one too, yeah. Yeah, yeah, yeah. Um but I know I do I think like people are becoming much more reasonable about it and so forth. And like I I just feel like you're one of the people that's really been advocating for AI research. I think it's the right thing to do. Thank you very much for doing that. Um is your job getting easier, in my right, or is it, you know, uh has it has it shifted noticeably since it started, say, two years ago.
15:08 Speaker 0 You know, I I would say we actually for we were able to foresee the need for more research dollars as it relates to artificial intelligence. Myself, Senator Schumer, uh, who helped get signed into law of the Chips and Science Act. We made a massive uh allocation for additional federal research through places like the National Science Foundation and Department of Energy. It wasn't easy to get that included with the CHIPS piece, uh, but arguably that will even be more consequential towards our economic security and our national security in the longer run than will the the CHIPS uh piece. So that authorization will allow us to now appropriate money if members of Congress can be persuaded for uh AI.
15:53 Speaker 0 And and and so I feel like half of that hard work is is is done. And um we're just uh we've we've got a lot more work to do, but I think uh what was the second part of your question? I'm sorry.
16:06 Speaker 1 Well, I was just wondering if your job was getting easier, but I'm so glad you brought the CHIPS Act. So I would like I'd love for you to answer like, you know, is your job getting easier? Because it seems to me that like that there is a lot of rationality in the discourse. I I would love to get back to CHIPSAC because I think it's probably the most significant piece of legislation in the last 50 years for innovation. So we'll do that next.
16:22 Speaker 0 Actually, that's where I was headed on the point. Thank you for centering me uh again. But the CHIPS Act. Yeah, so the job's getting easier because we pass the CHIPS Act. It's not just the authorization of research, but we've already made the argument very recently that something that falls outside of the direct DOD context, uh, as most of the Chips and Science Act does, is indeed a national security investment. So we got the institutional muscle memory, used to that notion. Uh that is a historical and modern history. It's not uh my colleagues typically don't think of an investment in the National Science Foundation as a national security investment, uh, but it is. And then we also were able to make the case uh to my colleagues that by making these sorts of critical investments in in research uh and in next generation technologies as we did through the Chips and Science Act, uh, it's gonna lead to a lot of economic growth.
17:17 Speaker 0 And we were able to personalize that argument state by state. So um that's that's the very same argument uh that uh we have for uh uh this situation, artificial intelligence.
17:27 Speaker 1 So this is gonna be repeating a little bit, and I apologize, I just think this is so important. Like um so many of the DC insiders here understand the CHIPSAC and why it's important. A lot of people here that come from the investment committee or founders of whatever probably don't understand the implication. My view it is literally one of the most significant pieces of legislation in in 50 years on on innovation. It's a huge, huge, huge benefit. It was the right thing to do. And so if you could just take just a couple minutes to describe at the highest level for those that are not insiders, what it is, I think it's so important to get this message out.
17:55 Speaker 0 So the Chips and Science Act uh was at once it was a $53 billion investment uh in in both research and incentives for the semiconductor industry. Uh the incentives would be to reshore some of our manufacturing capacity so that our supply chains uh would would not be as vulnerable to interruption as they of course were during the global pandemic, and they could be, um, God forbid if if if uh there was a geopolitical effort to interrupt those supply chains, say, by the Chinese Communist Party uh making uh aggressive actions towards Taiwan. Um it was also a national security effort, uh, because of course we need microprocessors and we need our own radiation hardened, domestically manufactured microprocessors to go into nuclear weapons, uh, to go into our radar systems and all manner of other things.
18:49 Speaker 0 Um so that was the microprocessor piece. Um, and we have been in the process now, where's some of that money's starting to flow, over 200 billion dollars of private capital uh is has been invested, and uh we're not even at a billion dollars uh even close to it of federal monies that is has been released. So it's it's paying handsome di dividends, the market is responding, and we're becoming less risky in our supply chains. The idea was not to become independent of other countries. And then there's this whole other and science piece, uh, which for the purposes of this conversation, big investments in research, which I mentioned earlier, not just in artificial intelligence, but that research can also flow to hypersonics, quantum computing, synthetic biology, uh autonomous systems, and other areas far upstream, of course, but a bit a bit more applied research than um than the curiosity driven research that most people uh tend to associate with the NSF.
19:53 Speaker 0 Yeah, that's great. Yeah.
19:55 Speaker 1 So um a number of investors are kind of like these kind of free market. Could solve everything and like kind of this kind of almost libertarian event. I am not that. I, you know, I I worked at Livermore, I worked for D DOE, I worked for two war efforts. Like I'm a huge believer in the government, I'm a huge believer in actually national institutions and involvement. That said, there's been an ongoing dialogue with like what is the right roles between kind of private and public partnerships for things like AI. Like what is the right balance? And so I would love your view on how we can work together on this where the government stops, you know, where the free markets pick up, kind of like how do you think about that as we go forward?
20:35 Speaker 0 Well, I look to history. Um I I look at what we did in the space race, I look at the innovations that have occurred through uh our DOE labs, like the one you worked at, uh like the one that developed fracking technology before the frackers claimed it their own. Uh right. I mean, I we so many innovations uh have been uh uh earned off of of of the the toil of our researchers and in in our DOE labs uh through our our land grant colleges and and and the other constellation of research agencies. So we need to keep making those investments. Over the years, the federal government started to uh pull back from research that wasn't curiosity driven, uh wasn't basic uh research, uh theoretical sort of research as opposed to uh more applied.
21:22 Speaker 0 We need some more applied research. I think that's one of the lessons uh we've learned in in in recent history. Beyond that, uh we need clear rules of the road, clear regulations. Those regulations ought to typically be, and I use my words very carefully here because they're exceptions to everything, but they typically should be technology agnostic so that we don't favor or constrain different types of technologies within the market. That's investors, that's business people, that's consumers. Um then we're gonna have to work with our partners and allies on development of standards for some of these what I'll call platform technologies. There aren't many. Artificial intelligence and and synthetic biology are the ones that really come to mind.
22:10 Speaker 0 So there's a diplomatic component that if we want our values embedded in this and and cut and follow-on generation AI technologies, we better develop and design those technologies here, but we also need to embed the standards uh that we have as it relates to privacy, consumer protection, uh, and other things into those technologies rather than leaving it to I'll pick on the Chinese Communist Party again.
22:37 Speaker 1 Yep, perfect. Okay, so we have the last minute. So I only thought it was fair because when I was at the Insight Forum, you asked me this question. So I'm gonna ask you two questions that you asked me. The first one, what would it
22:46 Speaker 0 may be a hard question?
22:48 Speaker 1 It's not what would you give the chances of PDOOM?
22:52 Speaker 0 Oh, the the probability of P doom? The probability of P doom. This is colored uh by my my optimism, right? Um but I uh low. You didn't you didn't ask me to make it quantitative, qualitative. Yeah, I low but not non existent. Yeah. Uh which is how most of the you know doomsday scenarios are. Uh they're low probability, high cost, and we need to hedge against them. But I think uh let me say this. The first step in hedging against them is to really seek to understand them, to study them a lot better so that before you take more costly uh action, uh constraining the opportunity cost for humanity ahead, uh we really know what we're talking about. I love that.
23:38 Speaker 1 One last very quick question with my flair. Do you think that um PDOOM or PDOOM is a probability that we humanity has a catastrophic event? Do you think it's greater with or without AI?
23:53 Speaker 0 I think in the long run, I think I think in the short run, uh you you'll see I'm not pandering here. I love it. This is great. I'm the I think here's what we're probably gonna have. We're gonna have in the short run some unsteadiness, yeah, right? Yeah. Because we're trying to learn countermeasures. We're trying to figure out how systems work, we're trying to figure out all sorts of things. Um then in the long run, I think we can actually push the risk down. Yeah. Awesome.
24:20 Speaker 1 Thank you so much. You are wonderful. Thank you so much, everybody. We appreciate that.
Previous What is American Dynamism? Next From Silicon Valley to the Pentagon: The Future of Defense Innovation Page updated August 25, 2026 · transcript and summary are machine-generated
