Acceleration & Balance The American AI Strategy presented by Kindred Ventures

All right, Good morning. It's nice to see you all. Thank you for being here. My name is Kanye Makubela. I'm one of the partners here at Kindred Ventures. We're in the midst of a technology transition that the world hasn't seen possibly ever. And so we're so lucky that we have these esteemed guests who I'll introduce in just a second who are going to be able to talk about what this means for the United States and what it means for our government. But without further ado, and I wrote it down because I want to make sure I get all the details right, including on my partner. I'm going to first introduce my partner, Steve Jang, who is going to be the moderator here. So Steve and I manage a venture firm called Kindred ventures. It's a $900 million venture firm. And we are investing in entrepreneurs at the earliest stages who are transforming the world with AI. We have been focused on mobility and logistics, robotics, aerospace, crypto, health, finance. So sort of important areas to rebuild and build the United States economy. We're typically first money in. Steve, I'm going to embarrass him. I hope he can hear this. Has had the opportunity to invest early and advise in companies that are leaders of markets, just like U Uber and Coinbase. And in the Crest, in the beginning of this AI wave, companies like Perplexity, which you may know, but is the innovator in AI search and deep research companies like val, which is the leading developer platform for AI, video, images and audio, both are just extraordinary companies that you should please be sure to check out if you don't already know them. And Steve has been named a couple of times actually into the Forbes Midas list for his efforts. And, you know, we are very lucky to have a portfolio, many of whom are in here, of 100 companies that are currently representing over $275 billion of enterprise value. So we've been so lucky to have partnered with some really wonderful teams. So our first esteemed guest here is Congressman Ted Lieu. So Congressman Ted Lieu represents California's 36th congressional district and is serving in his sixth term of Congress. So the 36th district is in the LA region. It's El Segundo and parts of Culver City, and so that part of la. Congressman Liu sits on the House Judiciary Committee and the Foreign Affairs Committee as Vice Chair of the House Democratic Caucus. He is the highest ranking Asian American ever to serve in House leadership. He's a former active duty Air Force officer. And Ted continued in the reserves. We retired in 2021 with the rank of colonel and he's emerged as a leading voice in Congress on artificial intelligence. In the last con, he served as co chair of the House's Bipartisan Task Force on AI alongside Congressman Nolte. So it's really special that we've got both of them here today. As a legislator, Ted has established himself as a leader in artificial intelligence, the environment, cybersecurity, civil liberties, foreign affairs and veterans. Time magazine named him one of the most 100 most influential people in AI, which is pretty cool. He was born in Taiwan, emigrated to the United States at age three and he's one of the only members of Congress or one of a few who has a degree in computer science. And so he graduated from Stanford with a bachelor's in Computer Science and a bachelor's of political science in 1991. We also have Congressman Jay Obernolte. Jay represents California's 23rd district. So if any of you are familiar with the Big Bear region of California, that's where Congressman J. Obernulti and his esteemed wife Heather, who's here with us today, are from. And he's serving in his fourth term in Congress. Congressman Obernolti serves on the House Committee on Energy and Commerce. He's a member of the Subcommittee on Communications and Technology, Subcommittee on Health, Subcommittee on Commerce, Manufacturing and Trade. And that committee actually oversees much of the Congress's work on technology policy, data privacy, digital commerce. He also serves on the House Committee on Science, Space and Technology and Chairman of the Subcommittee on Research and Technology and sits on the House Budget Committee. So he is not that busy. Last Congress he served as co chair of the House Bipartisan Task Force as I mentioned, alongside Congressman Ted Lieu. Before being elected to Congress, Jay was actually founded and ran a successful video game studio which is still operating and led by his son at the moment, and served as Mayor of Big Bear Lake and was a member of the California State legislature. So he championed policies that support innovation, reduced barriers for small businesses and expanded technology across stem. He's a flight instructor, certified flight instructor, martial arts teacher, lifelong technologist. He also holds a BS in Engineering and Applied Science from Caltech and an Ms. In AI from UCLA and a Doctorate in Public Administration, California Baptist University. So I don't think I'm overstating it when I say that these are probably the two smartest members of Congress. And so as a result, without any further ado, it's my pleasure to introduce Steve Jang, Congressman Lu and Congressman Obernolte. Come on up.

Thank you very much.

Thank you.

Okay. Thank you so much, Congressman, for joining us today. Please have a seat. As we know, there's a government shutdown right now and so really appreciate your time here. I know that you're both working hard to get everything back and running at the federal government level, so I really appreciate the time you're spending with us here. We have a lot of LPs, co investors in the VC community and founders here today. So we wanted to go through a little bit about some of your work. First, at the House AI Task Force that both of you co chaired. I wanted to start off by asking, how did it come about, how did that task force actually come into existence? Because I thought that was really interesting that it came up so quickly and very early on relative to other sort of government initiatives, sort of catching up to what's happening in Silicon Valley. I felt like that one was prescient and actually happened very quickly.

Well, we had one previously that was the AI working group that was a creature of the 117th Congress. And when Kevin McCarthy was actually the 118th, when Kevin McCarthy was speaker, and he was the one that asked me to head up the working group to try to get some idea generation going for federal regulation of AI. And then when Speaker Johnson assumed the speakership, he was the one that formed the task force. And I thought that the task force, it was very useful to have a defined goal. Our defined goal was to, by the end of last year, produce a report detailing a proposed federal regulatory framework for AI. And so I think it was really helpful to have that goal that kind of got everyone moving in the same direction.

And what month and year was that the task force came together?

Well, I was. Speaker Johnson and I were talking about it in December of 2023. And I think officially we were formed. I think it was in the spring of 24. It was early in 2024 that we were formed.

So first of all, thank you, Steve and Kanye, for hosting this and all of you for coming. I'm a recovering computer science major, so I'm very excited whenever I can talk about AI and technology. I happen to have been vice chair of the Democratic Caucus when Mike Johnson was vice chair of the Republican Conference. And we did things together, including we had a bipartisan dinner where we brought in Sam Altman. And I had this bill to create an AI commission of civilians and experts. And I remember, I want to talk to Mike Johnson about this one day, and he said something interesting. He said, he goes, I'm not opposed to concept But I think something along the lines, we should do it. And that just sort of stuck in my mind. And then when he became speaker, Jay and I went and lobbied him and said we should do it. And then he and work there, Keem Jeffries, our leader, and they created the task force.

That's great. One of the things that struck me about not only reading the report, but also hearing both of you speak in public and in media is that this bipartisan task force was not actually really in conflict or at odds, but that you're more aligned on everything than I would have expected. The alignment of both of your opinions on the findings and the recommendations was great to see from a technology perspective. In Silicon Valley, we're constantly worried about complication division, which we'll go into a little bit. But maybe if you could talk a little bit about some of the key principles that you found as part of this task force and this great report. If you haven't actually read the report, we won't do a TikTok of the entire report today. But I highly suggest downloading it and you'll probably upload it to Perplexity or chatgpt to get it summarized. That's okay. But I actually suggest reading it because there's a lot of findings that a lot of these LLMs and these search engines will gloss over. But it's really good reading. So please download it if you haven't read it yet. But I thought maybe we could talk. One of the key tenets of it is how do we promote AI innovation. Both of you are very pro AI innovation for the US economy and society. Can we talk a little bit about what are the things that you think are very great offense moves for the American AI strategy? Then I'd love to go into what are some of the balancing moves that we need to take to put guardrails on that. So maybe Congressman Lu, if you could kick off on that one.

Sure.

So the report is very bipartisan. It doesn't mean we didn't have disagreements during their process. And we would have these hearings where people would ask questions and we would have disagreements. But at the end of the day, 12 Democrats and 12 Republicans came together and we all signed this report. By the way, you mentioned AI and the report, it was interesting. I ran the report through an AI product. I'm so curious in the product, and it says something like, yeah, the report in the opening letter says something like, there's 85 recommendations, but actually there's 89 recommendations in the report. Like, huh, I guess you should have double checked this with AI before we actually launch it, release it publicly.

So perplexity says 89 and chatgpt says 85.

Oh, see, yeah. So there we go.

So.

So in terms of some of the highlights of the report and energy, I do think that as far as we can tell right now, there's just sort of a causation between the more compute power you have, the better the model is and that's going to require more and more energy. My view, it's just simple economics, is you just have to have lots of energy projects over the United States, whether they're green energy or not green energy. And I think it's not great for this administration to be cutting all these projects for energy just because they happen to be green. And I think that's basically going to reduce supply, which eventually is going to keep on increasing electricity costs for people across America. And we know Today inflation has hit 3%, the highest since January. And so my view is we just have to have more energy and all kinds of energy.

So, you know, we in the task force approached the problem of promoting AI and American leadership in AI from the standpoint of our role as legislators. So that's what the task force report was focused on, is, you know, what are the appropriate legislative steps for Congress to take to be able to achieve this. And you brought up the word balance, which is the same word we used over and over again in the report. Because on the one hand we think that we need to have protections that protect Americans against the very real harmful consequences potentially of AI, which in our view the main one is malicious use of AI. But on the other hand, we need to promote, have a light touch regulatory scheme that promotes, allows American innovation and entrepreneurialism to flourish. So we spent a lot of the report talking about achieving that balance. And it was interesting when we first started meeting and having hearings on the task force, this was right when the EU had just passed their AI Act. And so, you know, they were evangelizing this centralized bureaucracy where the governance of AI was, was all centralized and they were creating a separate licensing requirement for anyone that was going to use AI in their sectors. And we very consciously took the opposite approach. And we said here in the United States, sectoral regulation is the way to go. And we had kind of the advantage in hindsight of the fact that NIST had just come out with their AI risk management framework. And if everyone's looking for some other bedside reading material, I'd highly recommend that to you. If you haven't read it, it's short, but basically what it says is the risks of AI deployment are highly contextual matters. Very much what you're going to do with the AI when you're making a list of the risks that you're trying to manage. And it's critically, we think it's critically important to understand what those risks are are before we try and adopt a regulatory framework. Because the purpose of the framework is to guard against the risk. I mean, we're not just regulating for the purpose of regulation. We're regulating to mitigate these risks. So if you don't understand the risks, you can't do the mitigation and you can't regulate. And our sectoral regulators are the ones who understand the most about the risks in their sectoral spaces. So you could have an algorithm that's unacceptably risky in one usage context, like implanted in a pacemaker, for example, that would be completely benign in another usage context. Right. So exactly the same algorithm, two different usages. You know, one is unacceptably risky, one is completely okay. And so the sectoral regulators are kind of the guardians of that here in the United States. And, you know, they're already doing this job. The FDA has already processed over a thousand applications for the use of AI and medical devices, which is just about the highest risk use context that you can think of. So, I mean, this is the main point we're making in the report, is that we have a framework that works here in the United States. And if you approach governance as the process of mitigating risk, and you recognize that most of the really harmful uses of AI are already illegal, so we're regulating outcomes, we're not regulating tools, then that really sets the stage for American success.

Yeah, that was a concern when this topic sort of rose up to the top of our conversations here in San Francisco, Silicon Valley, where there was a thought about, hey, maybe models should be regulated. And I think the majority of our opinions, especially at Kindred and with our founders, is we don't want that. We want to accelerate on development, training, models, making them as high performance as possible, and then that way on application layers on those outcomes, then critical industries, how do we keep that balanced? And guardrail? Because there are certainly concerns there. But recently SB53 has come about for California. If you could give us delving into regulating with some oversight and some controls on models. Right. And which is something that, you know, we're not sure. You know, there's always going to be some level of regulation and oversight, but it's a. It's a, it's a very fine line. If you could maybe talk about some of your opinions around SB53, which affects California, and then how that kind of, you know, how you see this other states playing out, because it's very concerning to see a state by state regulatory path. That's very complicated for a lot of our companies.

So I know Jay has a lot to say in this. I'll just be brief on this. The administration this year tried to basically put a 10 year moratorium on states regulating AI. That failed. And so my argument now is it's not whether you're going to have AI regulation, it's do you want Congress to do it or do you want 27 states to do it? And I know Jay, and I think it's much more reasonable to have Congress do it.

Right. We are in furious agreement on that. And you know, specifically SB53, I mean, far better than Senator Wiener's bill last year, which was much more invasive. And you can, we could have a, like a point by point discussion about the requirements of SB53 and whether they're good or bad for the industry. But one thing that is absolutely abundantly clear is that many of the provisions of SB 53 clearly impinge on Congress. Article 1, responsibility to regulate interstate commerce. You know, the Constitution takes the topic of interstate commerce and very clearly says in Article 1, Congress and only Congress should be the ones regulating this. And it's pretty clear that when you're talking about regulating the development of models, particularly the development of models that are going to be deployed interstate, that that is interstate commerce related and not something that should be subject to, to the state regulators. So this is a problem because we've been down this road before with digital data privacy. You know, Congress was slow to act. Europe got out ahead of us with gpdr and we had a bunch of states that passed regulation because they, you know, Congress wasn't doing anything. And now we're making an attempt to pass a unified federal framework for digital data privacy. And one of the chief opponents of that effort is the states because they feel this creative ownership and of what they've made. But what we've created now is this patchwork quilt of, I think it's currently 27 different state regulations on digital data privacy. Eventually it'll be 50. And the same thing is going to happen to AI if we're not careful. I mean, we have over 1,000 bills pending in state legislatures across the country on the topic of AI regulation so far this year, just this year. So this is going to happen to us again. And what people lose sight of is the fact that this is incredibly bad for entrepreneurialism. Because if you are, I mean, not to pick on them, but if you are, you know, a meta or an open AI, you have got a building full of lawyers, you've got the legal sophistication to deal with 50 different state regulations. The people that don't have that capability are two folks in a garage trying to start the next meta or the next open AI. And this is really dangerous for us because we don't want to get into a situation where there's that barrier to entry. You know, we want to be the country where entrepreneurialism flourishes and we get all the, the great AI startups. We want that here, we don't want that overseas. And so, I mean, this is really what we need to prevent.

Let me just add real quick, so for any of you who have more direct contact with administration or if you're in the industry, just know, if you choose not to engage with Congress and you just let you know, sort of this mentality that Congress should do nothing on AI, then you're going to have the consequences that Jay just described. So hopefully people will engage.

And this is a, you know, what we've discovered is that this is a bipartisan view that everyone is very aligned on because of the complexity for constituents like us, is that we then have to wait for models to have a different version of the model in each state on state by state data centers with different regulations and different terms of service, which is incredibly damaging for entrepreneurship. So applaud you on that. Hopefully we can get some federal acts together so that we can sort of smooth out some of the state by state complexity. I thought maybe we could also talk about on, on sort of a national level how you think about infrastructure investment. You know, this year alone will be about 2, 2 and a half billion of infrastructure investments in AI. There's projections that call it at 15 to 20 trillion based upon modeling out just the recent announcements of some of these interesting financing and commercial deals that you've seen in the news, and that's worldwide. The US has always been a stalwart in technology leadership, but there are many things happening in China. There are many things happening in East Asia outside of China, Middle East, Europe. How do we, how do we think about infrastructure investments paving a more efficient path for growth? What are some of the measures that you'd like to see? And since the report came out, you made some recommendations and there's been some time so if you could maybe talk about what you've seen actually play out well and what you've been disappointed by, that would be great.

Sure. So, I mean, when I think about infrastructure for AI, I think on three different axes, right. I think we've got silicon, we've got energy, and we've got workforce. And those are the three things that really need investment to be able to keep America on top. In some ways, we're making good progress. You know, I think silicon is one that we've made a lot of progress on. And you've seen not only the Chips and Sciences act and renewed investment in having foundries here in the United States, but also kind of a recognition that to provide American leadership overseas, we need to make sure that our allies are equipped with the ability to have the US AI stack in data centers in their countries, which is a big change from when, the year we wrote the report, when the, the administration's attitude was the opposite. We had the AI diffusion rules that basically said, no, we want all of that capability in the US and nowhere else. So, you know, we've made progress there. Energy. Congressman Lu hit on it. I'm completely aligned with him in my belief that that is where we probably are failing the most right now. We were already failing before it became clear that data centers were going to strain our grid and our capacity. And the consequence of failing is that energy continues to get more expensive in the United States. And if you look back in the entirety of human history, every major expansion of human prosperity has been heralded by a reduction in the cost of energy. You know, however you define that. And so that's a lesson from history that we can't ignore because we can't allow this, the cost of energy to keep going up and the data center demand is just going to exacerbate that. So there are lots of reasons why we need to make a bigger investment in generation here in the United States. And you know, I think I agree with Congressman Lu. I think we need an all of the above approach to energy generation because if you look at each individual sector, none of it is going to be sufficient by itself. I'm a big fusion energy fan. I'm co chair of the Fusion Energy Caucus in Congress. Fusion is going to be great. It's not going to get here soon enough and the ramp up isn't going to be fast enough to save us. So, you know, we really have to make those investments and then, you know, workforce, that third axis, we're still behind. You know, we look at the fact that China will probably graduate twice as many PhDs in the computer sciences this year as the United States. You know, look at the fact that they'll. Last year they awarded over four times as many bachelor's degrees in the STEM fields as we did. That should be very concerning to everyone.

I want to add another aspect of infrastructure for you to think about, and I'll start with this analogy. If you're a cutting edge physicist and you want to do your cutting edge experiments, what if it turned out that every particle accelerator was owned by a private sector company? Would we have a problem with that? I think we might. And through no fault of anyone in AI, this is what it is. If you want to work on frontier AI models, they're all owned by private sector companies. So if you're in academia, if you're in government, you have literally no access unless the private sector company chooses to give you access. And then this causes all sorts of various different consequences. You have professors who may want to do certain research in certain areas because they will know that, hey, they'll get access because they work on this issue and not this other one that the private sector companies don't care about. It hampers our national security or entire government ability to access these models. So J does have legislation on there creating a national AI research resource that essentially provides some of this compute power to academia and government and so on. I think that's an important component of infrastructure.

Are open open models part of that as well? Is that something that the congressmen have an opinion on?

I support open source. They're just not for many things. They do the same thing, but they're not as advanced as the most advanced private sector AI models.

Yeah, we had a whole chapter in the task force reports, as you know, on this whole closed source versus open source debate. And what we concluded, and I still think remains the case, is that both models have a role to play in US AI dominance because both of them are suited for different things. When Congressman Lu was talking about the AI research resource, a big part of that it's not just compute, it's also data which lends itself to being used in conjunction with open source models. Because you can get access to an open source model of high quality pretty easily right now. But getting access to specialized training data that you need to adapt that model to a specific use is much more difficult. So that's part of the effort with Nair, is to make sure that people have access to not just the compute, but the data sets that they need to succeed with open source on this program.

Is there a good resource on the task force website or where can people find out more about these research resources and the plan there?

Well, there's a pilot program right now for nair and so if you go on, if you do, you know, just ask your favorite search engine, you know, for about nair and it'll, it'll take you to the pilot program. But what we were, what Congressman Lu and I are trying to do is to codify that and to make it a, a permanent asset with dedicated resources. And that's what needs to be accomplished at this point.

Yeah, I think with open source models, the great part about open source models, it allows access to academia, small companies, it really diffuses potential. You can fine tune models, you can make them safe if they're coming from another country and there's some concerns as long as the weights are open and you can host them on US data centers. So it allows for a level playing field and accessibility, democratizes all of that capability and doesn't kind of collect it at the top with a couple closed source platforms. The final part of that equation that we're always concerned about. I just got back from the Computer Vision International Conference for Computer Vision and on these research posters, all the PhD students sponsored by many companies like Nvidia or Meta are up there presenting post cutting edge research, state of the art models. And if you look at it, University of Illinois, Texas A and M, all these great American public and private universities, all of those names, like 80% of those names are Chinese immigrant students who want to become Americans. They want to work at, they want to eventually start companies or be a part of an amazing American startup and they want to move here. And I think recent wins on immigration have been really problematic to think about stifling that. So to your point about all this great talent that's getting trained and educated in other countries, I think that we want to have an open door. And that's part of it is, that's the other part of that equation, which is if we want to really bring these research resources and accelerate US capabilities there, it's really got to be an open door to this really high value immigration, this talent that wants to be part of the American system. So I want to add that as another component to the work that you're doing. If we can have a good strong plug for that, I think that's really important. If you go to any of these AI conferences with all of these PhD students from American universities, you can see their names. And so a lot of them unfortunately had notes said I couldn't get into the country. I couldn't get a visa. So they had a proxy presenting for them. So we're seeing this already happening right now.

Donald Trump in the past had some along the lines of Anyone getting a PhD here should have a green card stable, something like that.

It's still, it's challenging to see it, though, at the front line. It doesn't exactly meet out. So I think something more expansive and efficient would be great.

Yeah, I think in time, you will see the administration get there. I've been vocal in my belief that every PhD awarded in a technical field by a United States university should come with a green card, because I think it's crazy that we have the best system of education in the world here. People come from all over the world to get an education, and then we send them back to where they came from to compete with us. That makes no sense. And then just recently, you know, this hundred thousand dollar fee for H1B visas, I think that that's. That's pretty misguided because what it does is it creates this uneven playing field between entrepreneurs who can't afford that and establish companies that can. And that's exactly the opposite of what we want. We want to catalyze entrepreneurialism here. So Congressman Licardo and I just sent last week a letter to the administration, you know, bipartisan letter, urging them to reconsider that $100,000 fee for that reason.

Thank you for doing that. A lot of our startup founders and CEOs immediately there's a lot of concern, and as you noted, small companies just can't afford that. Right. And so again, do you want to create an oligopoly or do you want to create a democracy of startups competing? I wanted to switch gears a little bit and maybe talk about what you're seeing out there in terms of what has played out over the last year since the task force put together the report. There's been a lot of individual efforts now to see things through. If you could maybe talk about some of the things that you're most excited about or most concerned about, that would be great.

So there is a line in the report that says something along the lines of, we don't know what we. We don't know. And I had really wanted to call this report version 1.0 because of how quickly AI moves. When we were doing the report, for example, AI agents weren't really a big thing. Now they're sort of a thing. Right. And so now you've got different companies coming up with various AI agents. They're not great, but this is the stupidest they're ever going to be.

Right.

Every passing month they get smarter. And that's something we sort of should think about because it's one thing. Right. If, you know. So all these larger models do testing and. Right. OpenAI has disclosed that they got the models to engage in strategic deception, which is the model will take actions not aligned with user intent. Anthropic has disclosed they've got their models to cases where they'll try to blackmail the user or report the user authorities, which is all sort of interesting. Okay. If it's writing an essay for you on why the Battle of Midway was so important in World War II that you can turn into a history professor. But another thing, if there's an agent in the real world doing stuff and let's say it goes rogue, it could empty your bank account or it could make stuff up about you and send it to law enforcement. So those are things that we might just want to think about. It's not going to happen now, but maybe a year later they might get these sorts of capabilities. And. And again, we just don't know what we don't know.

You know, I think when you talk about what's happened since the task force report came out in December, I mean, the folks in this room, you guys are the experts on what's happened in the field of AI. But if you ask us what's happened in the world of AI governance, I think that two big things come to mind. The first thing is that we have seen an acceleration of state efforts to regulate AI. And that's. Congressman Liu was talking about the moratorium, which was originally going to be part of HR1, but. But ended up getting stripped out in the Senate. And that was kind of a reaction to. To that, you know, this idea that, yes, states are going to have a lane in regulating AI, but the federal government has to go first. And so this is what I think has become critically urgent for us, is to create some guardrails that make it clear where what is interstate commerce related and therefore clearly set aside for regulation by Congress. And then outside of those preemptive guardrails, where the states can be the laboratories of democracies that they are. Right. And our system of federalism kind of makes it clear what those guardrails ought to be. But we need to get busy and codify that. You know, I think is one thing that's become, you know, very urgently apparent. But then the other thing, and this is kind of A positive development. So, you know, we saw a transition between the Biden administration last year year and the Trump administration this year. And one of the big fears was that you would see a complete change in the way that successive administrations approach AI regulation. And I think people were a little alarmed when one of Donald Trump's executive action executive orders overturned the Biden EO on artificial intelligence. You know, which, you know, there was stuff in there I agreed with. There's stuff in there I didn't agree with, but I didn't. We were urging them not to completely repeal everything. But then when you looked at the, the action plan that the administration put out this summer, I think that should give everyone some comfort because there are a lot, there's a lot of continuity in the approach of the Biden administration, the Trump administration there. I mean, for example, the AI Safety Institute, I guess we're not allowed to call it the Safety Institute anymore because that's a Biden thing and therefore must be bad. But you know, now, now it's Casey, which, I mean it's still housed under nist, which is appropriate because NIST is a non regulatory body. You know, they're about standard SETT is what Casey's all about. And the Safety Institute was all about. So these approaches are very, very much aligned. And I think that that ought to give people some comfort that the rules aren't going to radically change. And so this is another thing, another urgent task for us in Congress is to codify this approach, codify the adoption of sectoral regulation, codify Casey, the old AI Safety Institute as kind of the, the hub and the hub and spoke approach on this and to create these preemptive guardrails. And you know, this is an urgent task for us in Congress. If we can figure out a way to get the federal government back open, you know, then we need to go back in Washington D.C. and work on this.

It is, it is extremely poor timing for many reasons, but to have this state by state patchwork quilt. It works for many other aspects of life, but not so much for having a consistent AI strategy, especially for entrepreneurship. The little tech sector, the big tech sector, they have the resources to deal with that. Startups, they don't. They have very little cash. It's like two years of Runway at any given moment. And so being able to build one product domestically, it's concerning to feel like you can't even do that in terms of like forward growth. We touched upon it a little bit with this AI stack, the national US AI stack, which, you know, Nvidia is obviously a big part of that. But there's also compute, right? Bringing those resources out, making it global, having US Companies be in the leadership position there. When we look at Europe right now, Asia, we look at every other continental region. COMPUTE will permeate everything. Every application, every website, every mobile app will be using inference at some level. And so what we counted on for CPU based clouds and hyperscalers, with NEO clouds now entering the picture and many different chip manufacturers, many different models, agents, as you mentioned, proliferating. Is there a sense in coming in your thinking in upcoming congressional work to not codify, but maybe think about how do we create measures for international relationships with other countries to say, hey, the AI stack that we have, we want to make a bridge, we want to allow our companies to be global because I think it's very concerning again to have our companies not be able to enter into certain regions because of politics or because of inter country sort of relationships. Is there something like that in the framework of how you're thinking?

So let me talk a little bit about export controls. I generally supported the Biden administration's export controls on highly advanced AI chips going to China. I disagree with the administration on the export controls to, for example, Greece. I think it's better that we focus on who we think really are our main competitors and adversaries. But I didn't think we should just make it more difficult for US Companies who are trying to export, for example, chips all over the world. I generally support US Companies exporting chips all over the world because I'd rather have people around the world using US chips, US programs, US stacks, and so on. But I do make a differentiation for places like Russia and China.

I completely agree. And I think that you've seen a complete sea change in the last nine months in the way that federal government is thinking about export controls. Because, you know, before we had the diffusion rules, it was clear that we wanted US technology to be US centric. We wanted the data centers to be here onshore. We didn't want anyone else to be able to get access to that, particularly not in places where an H100 could fall off a truck and wind up in a data center in Shanghai. But now you've seen this recognition that if we want to have the US AI stack adopted by our allies, that we need to relax those rules and recognize the fact that we have allies all over the world like Greece who want data centers and need data centers in their countries. They need the ability to train models, frontier models on their languages with you know, with access to their culture and to put it in a position where they can deploy it in their country and use it to enhance their own productivity. So if we don't make the USAI stack available to them, we are going to force them into the arms of our geopolitical competitors, and that's the new recognition. You know, that's absolutely true, and I think that's why you've seen the administration change its stance on that.

We have just a few minutes. We're going to try to do a couple questions and Connie can jump in with that. That would be great.

Okay. Thank you so much for joining us here. It's just as a quick first comment, it's interesting how often you're saying, I completely agree with each other, but what did you expect from other sides, honestly? Yeah, I want to just acknowledge that and say thank you, because we hear a lot about the disagreements. That's what makes it into Politico. But the agreement is helpful to understand.

You know, can I interject, please, just for 10 seconds here? People lose sight of the fact that when we put out that task force report, I mean, first of all, we. We eschewed the usual practice of having more members of the majority party than the minority party on the task force. We had 12 Democrats, we had 12 Republicans. And what Congressman Lu said, you know, bears repeating. That task force report, which is 270 some odd pages, was unanimously approved by all 24 task force members and the speaker and his staff, and the minority leader and his staff. So you should have seen what was on the cutting room floor. I mean, the fact that we can take 80 plus recommendations and have them be substantive things that both parties can agree on I think is amazing. And really, I mean, it should improve people's confidence in the fact that Congress is capable of agreeing on this and acting.

Absolutely, absolutely. To follow up with that exact point. Actually, what are the thorny points that we actually can't figure out how to agree on? Maybe specifically relating to the concern of. I think there was a graph that showed the, you know, the expansion of the stock market and the contraction of the jobs market. It seems like there are some real challenging implications to some of how this is being rolled out, and particularly that one seems like it is, you know, a cause for some anxiety. I'd love to just hear a reflection on what happens to the jobs and how do we think about that? Just as a really quick first question.

Here, I guess I'll go first. You know, I think that most of the task force members are optimistic that ultimately AI will create a bunch more jobs for humans than it displaces. It will be disruptive, as technological innovations always are. I mean, most recently the Internet, but going all the way back to the printing press, you know. But it's always been true that more jobs are created than are displaced. So that does not ignore the necessity of making sure that we have a robust social safety net to catch people who fall through the cracks. We're going to have to do reskilling, retraining. We're going to have to absolutely make a renewed commitment to equipping our young people who are coming out of school with the education that they need to be able to use AI to make themselves more productive. We had a whole chapter on that in our task force report because we were so concerned about it. But I mean, none of that should distract from the overall optimism that, you know, it's going to probably create more jobs for human humans than it displaces. And I continue to believe that's true.

What's interesting to me about this technology is it is so easily accessible, right? So in the past, let's say you had a company to get ahead, you needed some super expensive laser cutting tool. Well, maybe not so easy for everyone to get that, but now, right, you could be in Kentucky or Zimbabwe or Belgium, and you can get access to a very highly advanced AI model for free, and you can pay a little bit more and get an even more advanced model. And so all of a sudden you've got billions of people who now have access to some of the most powerful technology in the world all at the same time. And what does that mean? I don't know. We're still sort of seeing what the effects of this, but I think there'll be disruption and people are going to lose jobs, People will gain jobs. How it all shakes out two or three years from now, I don't think anyone really has much of an idea.

One thing that's interesting to your point and both of your points, is South Korean government. Recently the new president there did something really kind of wild said in their campaign promise, said Mike. One of my campaign promises is that every single citizen is going to have a ChatGPT Pro subscription or something of equal strength, right? And so how they're affecting, I think they had a meeting, they sort of did the math afterwards and realized, oh, that's really expensive. So what they did was they did something really interesting saying, you have a year and a half. I think it's a year and a half. I might be Wrong. But you have a year and a half. We're going to set up five teams. Five companies will create alliances with their own AI stack models, chips, data centers, compute, energy, and you will all compete. It's like Hunger Games set forth by the federal government there. And so it's like lg SK Telecom, like Naver, like so. And they all create their alliances for their own stack. And so they're competing every quarter. Someone gets removed. It's like a reality show, they get removed. The government reviews it and they get removed. So one there, one will be at the top of the mountain at the end of the day. And then the promises it should be a full AI stack where everyone has a chat GPT Pro like subscription. Now obviously we've suggested that perplexity should be part of that, but they said it's an American company, it has to all be all Korean. Right. And so maybe one thing that can be done is like you can have Congress, maybe the administration, the White House put down a Hunger Games style competition.

I think the President would probably like that.

Very appropriate.

Can I push back on that just for a second? Because I think it loses sight of the bigger picture. The problem that we have is not that people lack access to high quality models. Because I mean, you could go, you talk about ChatGPT Pro subscription. ChatGPT offers a free subscription that's quite capable. I mean, this is not our problem. Our problem is we need people equipped with hardware that can access it, broadband infrastructure to be able to give them access to it and most importantly, the education of how to use it effectively. And that's what we're failing at. We're not failing at the production of model side. If we fail, it's going to be because we failed at the, you know, the basic education and infrastructure side.

Totally agree. They actually contemplate that whole issue with going all the way down to bare metal and beyond energy. But I do agree we actually have the best access, the best economics, and we can't run. You know, it's actually the digital divide problem that still exists. Right? It's the old problem. After the 1990s and the Internet and web came about, we still have that problem, right? People of low income, certain regions of the United States, certain demographics, demographic groups, they just don't have access. Right. Pure like physical access to all these tools, let alone the pro subscription or the free subscription, one thing. And the Hunger Games wouldn't work anyway because Nvidia would probably win just because they have the most important, most important chunk of it. I Wanted to just say in closing, number one, how do we get more computer scientists in elected office? So please also have a House AI recruitment task force for more congressmen and maybe senators to pull from the PhD ranks.

Oh, please. Do you realize that between Ted and I here on the stage, we are fully half of the computer scientists in Congress?

I also looked that up on Perplexity. And this is 50% of the brain trust in AI in Congress.

So Jay and I and other members of Congress, we do this thing called the Congressional App Challenge for people in high school. And literally high school gets right apps. And then members of Congress, we sort of judge them and the winners go to D.C. and present and so on. And I spoke at this last Congressional App Challenge and I basically told the kids and the parents, I said, it is great you guys won this and are here. And I said, don't do this anymore because AI is going to replace your programming. Go, go find something else to do.

So, well, thank you so much. And just fun fact, I grew up as an immigrant child in your district, and I learned what snow is and how to ski in your district. So I really appreciate both of you coming and sharing a little bit of your busy schedule with us today. And if people want to contact the congressman, I won't give out their cell phone numbers, but there's an email address with their office and you can contact them if there's any issues that you want to bring to the forefront or you'd like to have a dialogue with their offices. So thank you so much again.

Thank you. One quick thank you. I just want to say another thank you. This is an ecosystem play here, which means that we can't do these types of convenings without the support of our partners. So I just wanted to make a quick thank you to oric, who is our eminently capable counsel, who is one of our supporters here. So thank you, Louis and Shannon and team. And thank you very much to Silicon Valley bank, who, you know, stalwart is still the backbone of so much of the financial parts of this industry. So thank you and thank you again, Steve, for doing excellent moderating as usual. And Congressman Overnulty, Congressman Liu, we appreciate you both. Okay. Go America.

Thank you.