America’s Plan to Dominate the Full AI Stack with Sriram Krishnan

0:05Hi, listeners. Welcome back to KnowPriors. Today, Alad and I are here with Sriram Krishan, a top White House official currently serving as the senior White House policy advisor on artificial intelligence. A former tech executive and venture capitalist, he's one of the lead authors on the American AI Action Plan released this past week. We talk about the national implications of the AI race, what position we hold today, the workforce and energy needs of the future, and how to win. Sriram, thank you so much for joining us today for No Priors. Thank you for having me. I'm a long-term fan. I've never been invited before.

0:36I was always a bit sad, but thank you for having me for the very first time. I mean, before I just have to point out for folks who are listening on audio that Elad has never looked as good, as dashing, as handsome as he does. Now, Elad, you dressed up for me. I'm honored. This is how you can tell that Sriram is in politics now. He has the liquid tongue of gold with which he coaxes everybody into doing his bidding. So it's very good. So, you know, for our audience, Sriram has been a well-known Silicon Valley individual. He worked at Andreessen Horowitz. He worked at a number of the sort of marquee companies and names in Silicon Valley over the last decade plus.

1:09And now he's in government and he's really working on a variety of exciting initiatives around AI and other areas. Could you tell us a little bit more about your role? And should we be calling you your excellency or is there some special title we should be using now that you're in government? You don't have to, but I will take it. But no, thank you. It's fascinating for me to be here talking to you in this capacity because I've known both of you for forever and ever. You know, we've had hundreds of interactions and also been such a fan of the pod. Congratulations. Just a little bit of backstory.

1:35I've been in Silicon Valley for a long time. I feel very old. I did a tour of all the large consumer social media companies. And then I was at Andreessen Horowitz for the last four years, competing actively for CDC term sheets with both of you folks, I'm sure. And all this while, I had no real intention of joining government. I wasn't very particularly interested in policy. But what wound up happening is a couple of years ago, I moved to England to head up all of Andreessen's international efforts. And at the time, the UK was kind of a hotbed of all the AI policy debates. They had this AI safety summit at Bletchley Park.

2:16And this was kind of the peak of, I would say, the effective altruist versus the EAC kind of drama which was going on. And I got pulled into a lot of those discussions. And I remember thinking to myself, like, wow, a lot of people who are in very senior roles in governments in the United States back then, and in other parts of the world didn't know what they were talking about when it came to AI. I was convinced that they were doing the absolute wrong thing on many topics, for example, open source or helping startups. And it was just really, really bad in a way which I think the industry didn't really appreciate until much later.

2:52And that got me interested into just policy, which by the way was a word, which I didn't even understand what it means. We didn't even get into what that means, but it got me interested in policy. When President Trump got inaugurated in the first week, he did two things. One, he rescinded the Biden executive order on AI, which was bad and awful in many, many ways, which we can get into. And then he signed a new executive order, which basically said that America should dominate and win on AI. And then he called upon a few of us to say, you guys need to come up with a plan within six months to figure out how America is going to dominate and win.

3:26And so that put us off with a racist. And I think everything that has happened since then accumulated in the event that we had yesterday where we put together, you know, we put out this long document, 28-page document on America's reaction plan. We had a bunch of exeguatars. So that's kind of a little bit of history. That's great. And could you tell us a little bit more about what the main things that you considered as you put together this plan were? What are the things that you worry about geopolitically? How do you think about AI and competition, big tech versus small tech? Like it feels like there's a lot of threads in that.

3:55And it'd be great just to get a view of what are the main, the main issues that created this plan. And then it'd be great to talk through the plan itself. One of the catalytic moments which happened was the day before I started this job, I get a call and this was the weekend deep secret came out. And there's actually some chatter that I've heard online that China timed it so it can come out right after the president got sworn in. And we were like, hey, we just want you to come in and brief a lot of people of the White House on Deep Seek because we're like, hey, what is this? Is it cheaper? Is it faster?

4:21Do they have some magic way of training these models which only cost a few million bucks and not, you know, hundreds of billions of dollars? What's going on? You folks might remember that narrative which existed that weekend. And so I got to go. And, you know, me and David helped brief all of the White House leadership. But it was really a starting gun because I think that moment was profound because it immediately told us a few things. It told us that America doesn't have a huge lead on AI. It actually has a very, very small lead. If you remember at the time, DeepSeq was the only reasoning model, which was not open AI.

4:51So I don't think Cloud had come out of the reasoning model yet. I don't think Google had yet. It was the only non-open AI reasoning model. It was very high up in the leaderboards. It was a bit unclear as to what their cost claims were and how they had gotten there. I think we know a lot better. And all that ended up turning out to be very overstated, right? Like basically there was a claim that is a few million dollars to train the model and they didn't really talk about the hundreds of millions of dollars are probably spent to get to that point. It's sort of the last training run, you know, was sort of what they paid for.

5:19Yeah. Yes, absolutely. I think I would say there were claims were inflated and the claims were taken seriously. The claims were inflated was to your point, they kind of, I would say they put out like the final training run and as with all the ablations and training costs. And I mean, if you look at the paper, by the way, I don't think they make the claim that the total cost was Pingal Inbox. I think it's what the press imputed, but I do think they deserve a lot of credit. I always make a point to say, look, DeepSeek did some very, very good technical work. If you think about, they didn't have as good hardware as the American model companies do, what they did with KB caching, MLA.

5:49There's multiple theories on how they actually got chain of thought. Maybe they did it by themselves. Maybe they had some help from American companies, but there was some really novel new work there, right? And I think it told us that, okay, we don't have a lead that we can take for granted. They were definitely the best open source model. And I think even today, I would probably say the Chinese models, DeepSeek, when are the best open source models. But symbolically, it told us that we are now in a race, in a very close race, by the way. What does it mean to win the AI race? Like, why do we need to win that?

6:18And what would losing mean? How do we know we've won? Well, I suspect you folks agree, AI might be the most transformational economic, cultural force of our lifetime. And I believe that if the country or the ecosystem, which winds up getting ahead, is going to have these cyclic effects, right? Like you're going to power productivity, you're going to have drug discovery, you're going to discover new material sciences, new technologies, which then feed back into your infrastructure, feed back into your economy. And you're going to get this flywheel effect where who winds up getting ahead could wind up really accelerating ahead in kind of a classic network effect ecosystem way that all of us in Silicon Valley will understand.

7:00Now that is purely on this civilian economic context, you can also imagine a military context, right? Think about everything from drones to autonomous weapons. I'm pretty sure it's not in our best interest to have another country have that same economy of scale and flywheel and race ahead of us. So that's the race. Now, one interesting question that we have been pondering, which we can get it to, is how do you actually measure what it means? How are we doing in the race? And one measure I've been playing around with, and maybe I want to get your take on, is I think Google just announced this morning that they inference one quadrillion tokens a month or a quarter, I forget which one.

7:37And one of the measures I've been thinking about is, let's say the world inferences, I don't know, maybe it's called 10 quadrillion tokens a month. We don't know what the number is. What share of those tokens are being inferenced on American hardware, on American models, right? And how do we maximize that market share? And that's kind of one of the mental models I've been playing with. And it's in a way you can think about it as we are America Inc. We have a product stack starting from GPUs with NVIDIA and AMD and a bunch of others. We have a model layer with obviously OpenAI and GROC and Gemini and many, many others.

8:11We have an application layer. You had many, many, many of them on your podcast for agents to all kinds of software. How do we make sure this American stack is dominating that market share of tokens in France? Very good metric. Yeah, it's really interesting because one other thing that you didn't mention, I feel, is a cultural exportation through the models. And so if you look at prior waves of sort of culture spread, it was the movie industry, it was social media, and then now it's these models. Because a lot of people go to these models as a source of truth for history, for information, for other things.

8:42And there have been some famous examples in some of the Chinese models where there's a mission of Tiananmen Square or a mission of other facts. And relatedly, there's some things in some of the U.S. models that seem very politically slanted or otherwise not quite great. But it's interesting to also think about it from the perspective of broader cultural exports. I just wanted to add that to your points on defense and scientific progress and other areas. I think that's another key thing. Exactly something we're actually addressing. And you're absolutely right. I grew up in India and a lot of my exposure to Western culture was in the internet and Google.

9:09And obviously a lot of the internet was American. And that kind of introduced me to Americana. And imagine if 1995, the internet was rotten by America, but done by one of our adversaries. And so in a similar context, when Deep Sea came out, I think that all these great examples of lots of stuff in there, which doesn't probably align to American values. Now, we are actually addressing this. President signed an executive order yesterday. It's called No Woke AI in the federal government. And what it does, and this is probably going to be one of the spicy bits for your audience, is that it basically says that, look, from the day one of the Trump administration, we have tried to fight back against DEI, you know, wokeness, critical race theory, whatever you want to call it, in all parts of the federal government, right?

9:59And all kinds of propaganda. And what this EO does is actually very simple. It says that all models that the federal government will procure, a.k.a. your taxpayer dollars will be spent on, has to do two things. They have to be truth-seeking, and they can't have artificial ideological bias added. If bias is added, you just have to be transparent about where you're getting that bias from. It should be very simple for both people. But to your point, you know, that cuts to the heart of, you know, if you're saying nothing happened in Tiananmen Square in 19, you know, in the early 90s, that cuts to the heart of that.

10:31It also cuts to the heart of many, many other things from the culture wars that we have now been trying to fight against. Hey, Sri Ram, you used to work in social media for a long time, right? Like this sounds a little bit familiar in terms of like, is it a platform? Is it a publisher? What is the information consumption that most consumers have? Like, where does that analogy apply or break down? It's a good question. I think in some ways that's for the industry and the ecosystem to answer a bit. You're right. I spent a lot of time at Facebook, now Meta, at Twitter. One of the things I saw when I was at Twitter was how easily you could inject cultural bias in your algorithms.

11:10I have so many stories about how if you pick the right kind of Twitter accounts, which then feed into the trending algorithm, which then feed into Twitter moments, and then which every journalist or editor will wake up. And next thing you know, it's like one of the news stories of the land and BuzzFeed will write a piece saying, people on the internet are talking about this. I saw this over and over again, and it left me this profound appreciation of how algorithms can shape culture. And one of the things I always say is Twitter or X is the memetic battleground upon which we fight a lot of these ideological battles.

11:42So when it comes to AI, I think it's probably going to be very similar. Like my kids use chat GPT to answer everything, right, from history to geography to, you know, just kind of silly kids questions. And you can easily imagine a world where people inject their own cultural biases into this. And, you know, in the EO, we have a few good examples. We have, you know, we have examples of the Pope being seen as a black person, misgendering someone being seen as worse than thermonuclear explosion. And a lot of it is meant to say that you can easily imagine a world where these systems, which are at the heart of so many things that the government is going to use, we're always going to use.

12:19We don't want them artificially injected with an ideology, at least without being transparent about it. What are some of the other main points of the announcement from yesterday? So one of the ways David and I try to think about this with some of the people we work on is it should make sense as a strategy for almost like a technology company. And I hope that, you know, please go read the document. It's actually pretty readable. And hopefully for those of you who work in the tech industry, it should kind of make sense. And we think if America is going to win the race with China, they need to do three things.

12:50And they ladder up to the strategy. The first is we need to build infrastructure, right? At the heart of this, if you kind of go back to the scaling loss, what do we need? We need computation. We need data. And in the United States, it's been really challenging with the grid we have, with kind of this crazy permitting that we have around constructing new data centers, to get some of these projects off the ground. So the first part of the action plan really dives into what the president calls build baby build, kind of playing on drill baby drill, which is all about how do we make sure we are building infrastructure?

13:22Because obviously, you know, some of the other countries are. And just as an example, one of the things it talks about is to make permitting on federal land a lot easier for data centers when it comes to old environmental laws or other regulations which get in the way. So think of that of like, Let's make sure we are building the infrastructure to power these models as we scale up. So that's number one. The second pillar is innovation, which I would kind of say as let's make sure all these amazing companies, everyone that you know of, or maybe some companies which don't exist yet, can build applications and models or anything they want as fast as they can.

13:59And what we're trying to do is a couple of things I really want to highlight. The first is we want to cut through red tape. You know, like until last year and a half ago, I was in California, along with all of you. California almost passed SB 1047, which if that had happened, it would have been the end of open source, by the way, in the United States. We would not have a llama. We would not have an Owen mini coming out. And a lot of states which want to do versions of this. And we think that AI is a national priority. And if you're going to compete with China, we need to make sure that these are things that we take a deal with at the national level, rather than every single state, especially states which have ideologies that, you know, you and I may not agree on, you know, try and set their own rules.

14:40And by the way, some people may not understand this. If you have a small state set rules, it can often become the de facto law for the country. Because if you're a company, you're like, well, I have to operate in the state or I have an office. So let me just do that for everybody. Just like the EU does that. So we want to make sure, cut through red tape, let's make sure there's regulation that happens at the federal level. So that's very, very key, because I think that's going to enable not just the big companies, but every series A, series B, active hire companies, whatever the kids are doing these days, we make sure that they are off the list.

15:12That's number one. The second part is open source. Now, I think we probably talked about this a bit offline. Open source is one of the big reasons I actually got into the policy world. The Biden administration really, really tried to scare people about open source, talk about how unsafe it was. SB1047 obviously tried to kind of basically ban it in many ways. And what the EU does is say like open source is a space where the United States needs to win. It actually points to some resources that is going to be made available to research. Because I think you and I know open source is the way everyone from a kid in their bedroom or their dorm room all the way to a startup to all the way to somebody who wants a lower cost of inference in their IoT device or a robotic startup.

15:54That's what they're using. For context, too, much of the Internet runs on open source software, right? So the server software and other things, much of that is open source. The protocols are all open for the Internet. That's also true for crypto. And so, you know, it's interesting because removing open source from things like AI actually decentralizes power. It centralizes power into a small number of companies that could then be controlled by the government. And so to some extent, the fact that you all are supportive of open source means you actually are supportive of a thousand flowers blooming, but also a lack of direct government control in literally everything AI.

16:24So it's a very interesting counter stance to take. By the way, ELA has our talking points better than I have because that is absolutely right. One very fundamental difference I think we have with the Biden administration is the Biden team really looked at AI as something to be centralized and controlled. Everything was about how do we make sure that we regulate these three or four companies and only three or four companies can build AI. They got to submit their models for testing. It was all about control in a centralized fashion. Now, when I moved to D.C., one of the things I realized is that's kind of the way D.C.

16:55thinks, which is control and centralized in one place. You and I know that's not how Silicon Valley thinks. And one of the reasons Silicon Valley is the envy of the world is because anybody, any day can go to a Y Combinator seed round or just go off to the races and they could just build something amazing. And it catches everyone's imagination. And I think what we want to do is enable just that rather than say, OK, we want to centralize power within a 10 mile radius of where I am right now. Yeah, in general, too, central planning tends to lead to very bad economic outcomes. And so that's the collapse of the Soviet Union, et cetera.

17:28And so it's something that's been tried many times before in many industries, and it tends to lead to a very bad place in terms of innovation, in terms of economics. I think one of the things that people like under price about open source models is they're going to happen and it's a strategic weapon. They're happening and Western companies are using Chinese open source models very broadly already. And so if you believe that not every model is going to be ideologically neutral or aligned with American and democratic values, then you probably have a problem. And so the ability to support whatever your point of view, like pluralism and openness and innovation and have some control as an ecosystem versus in a centralized way is a very different point of view than we'll let China develop it.

18:19Yes. And I think you make a profound point. And you're already seeing that where when somebody's using DeepSeek or Quen, that's an expression of soft power. And I think I would much rather have, you know, them using a model built by somebody who kind of agrees with us and has our values. That's number one. The other issue I would point to is that these models, we don't know what's inside them. Interpretability is still a nascent field. And you could very easily see ways where you plug a model into a cursor or windsurf and you generate a piece of code. And then two years down the road, it turns out that code had a little if statement saying, if I'm running in some piece of critical infrastructure, go do something else.

19:00And we don't have ways to validate all that. And so a lot of reasons why we want to make sure that our American models or Western models wind up winning. And this is something I think we were going to put a lot of focus on. Just because you have such a good view into this, can we talk a little bit about infrastructure and energy? Since you kind of made that like point number one in terms of what sort of stack we need. Like people hear these claims from the leaders of the large labs that, you know, we're building a data center the size of Manhattan. Or, you know, it's the energy that a city uses at any point.

19:29Can you contextualize like how much capacity we really need to build and sort of like what the biggest bottleneck is? Like, is it is it the grid? Is it sources? Is it workforce? Like, you know, when you want to solve this problem, like as a systems person, like what is the first problem? So the first thing I would say is it is a system. And this system was one that wasn't really battle tested for decades. Somebody showed me this number. I think the United States basically had like one to two percent of power usage growth for a very, very long period of time. And so you can imagine this whole system of everything from gas turbines, coal, renewable energy.

20:07There was a regulation of which kind of really stopped nuclear. And then you had these per state utility companies, which often didn't have the incentive to innovate. You basically kind of ran the state. You weren't really, you know, getting new demand or competition. You had a grid which wasn't really pushed because, again, you didn't need to. And then you have essentially a patchwork of environmental laws, regulation, everything from water to emission to a whole other sort of things that I'm sure I'm forgetting, right? So somebody kind of explains to me as kind of this tangled, spaghetti mess of things, which, again, until two years ago was just fine because you and I were not dramatically using more energy than what we were doing 10 years ago.

20:46Now, that obviously changed. The scaling laws arrived and everybody is trying to build new things. And I think the way we are trying to attack it is that every single step of the way, which is one is how do you make generation better? Second is how do you make sure we make constructing these data centers better, making it easier to kind of these regulations and kind of get these red tape out of the way, making sure we put focus on the right energy sources and making sure like, you know, we have those lined up. So we are trying to take an approach to all of this. But it is a complicated problem just because there is so many different players, so many different states and so many different patchwork of laws and regulations involved.

21:26But I think what, you know, I encourage folks to look at the executive orders today, which the president signs on infrastructure, which I think is going directly at this. We also have something called the National Energy Dominance Council, which works very closely with Secretary Burgum and Secretary Wright, you know, for interior energy. And I think you're going to see a lot more from us on that front. The short answer, Sarah, is it's complicated. I think we're taking a very, very strong approach to this, but there's going to be more to come. How do you think energy infrastructure is going to feed into these big data center buildouts?

21:56And so, you know, one theory I heard is that fiber is cheap and easy to lay while grid is hard, building out the electrical grid. And so therefore, you're going to centralize data centers near sources of cheap power, and then you just run fiber into them versus moving things around based on other types of capacity from a telecommunications or their perspective. Are there specific sources of energy that you think are going to power this AI revolution or the things we need to reinvest in? Obviously, the president has issued some executive orders around nuclear. I'm just sort of curious how you think about what that future really will be and what are the major sources of energy that we really need to be dependent on and how does that all shape up from an info perspective?

22:32What I think we see our role as is like, get rid of the red tape. Let's make sure the permitting on these things are super easy. Nuclear, that's another case where I think for decades and decades, the climate lobby and the doomers that kind of stop any real efforts over there. So I think we're seeing a lot of effort. Let's get the red tape out of the way. Let's get construction going and see where we get. The other thing that I think is interesting from an infrastructure perspective is manufacturing capability and supply chain. And a subset of AI supply chain is dependent on China or other countries.