Scaling Laws: How To Use, Govern, And Lead On AI?

[Intro]

Alan Z. Rozenshtein: It is the Lawfare Podcast. I'm Alan Rozenshtein, associate professor of law at the University of Minnesota, and a senior editor and research director at Lawfare. Today, we're bringing you something a little different, an episode from our new podcast series, Scaling Laws. It's a creation of Lawfare and the University of Texas School of Law where we're tackling the most important AI and policy questions from new legislation on Capitol Hill to the latest breakthroughs that are happening in the labs.

We cut through the hype to get you up to speed on the rules, standards, and ideas shaping the future of this pivotal technology. If you enjoy this episode, you can find and subscribe to Scaling Laws wherever you get your podcasts. And follow us on X and Bluesky. Thanks for listening.

Alan Z. Rozenshtein: When the AI overlords takeover, what are you most excited about?

Kevin Frazier: It's not crazy, it's just smart.

Alan Z. Rozenshtein: I think just this year, in the first six months, there have been something like a thousand laws.

Kevin Frazier: Who's actually building the scaffolding around how it's gonna work, how everyday folks are gonna use it?

Alan Z. Rozenshtein: AI only works if society lets it work.

Kevin Frazier: There are so many questions have to be figured out,

Alan Z. Rozenshtein: And nobody came to my bonus class!

Kevin Frazier: Let's enforce the rules of the road.

Welcome back to Scaling Laws, the podcast brought to you by Lawfare and the University of Texas School of Law. I'm Kevin Frazier, the AI Innovation and Law Fellow at UT, and a senior editor at Lawfare, here to tackle the important intersection of AI policy.

And of course, today's guest is U.S. Representative Nick Begich, who is a proud voice for the people of Alaska. We dive into the latest twists and turns in AI policy, as well as outline a path to continued U.S. leadership in AI innovation. Rep. Begich is in a unique position to shape these conversations.

Before entering public service, he spent much of his career in the software industry. Today, as a member of the House Committee for Natural Resources, as well as the Committee on Science Space and Technology, he has a key perspective on how America can lead on the infrastructure and technology questions posed by AI development.

Giddy up for a great show, and as always, please leave us a review and follow us on X and Bluesky.

[Main Podcast]

Representative Begich, welcome to Scaling Laws.

Rep. Nick Begich: It's great to be here, Kevin. Thanks for having me.

Kevin Frazier: Yeah, so when folks talk about Congress, they rarely say, boy, it's just full of these tech wizards and these folks who really get technology, and yet. We've got you. And folks may not realize that in addition to Ted Lieu, who often gets cited as one of the more tech forward members, or Jay Obernolte who also has a background in technology, you yourself are no stranger to tech.

So tell us, what were you doing before you were making the trek from Anchorage to D.C. regularly?

Rep. Nick Begich: Yeah. Yeah. So my career has been predominantly spent in the technology sector. I founded a software development company that grew to about 150 folks. We built a custom software products mostly for startups, but also for the enterprise.

And we would often help coach our early stage companies in the development of their businesses. We'd take equity stakes sometimes in these businesses. And be a part of their growth story. But yeah, you know, my, my career started, on the tech side at Ford Motor Company in Dearborn, Michigan, and then later with the software company that I founded.

So, it's been amazing to see what's happened with respect to the software space as one of the first majorly disrupted industries by large language model, AI revolution. But that's been my background. So it's great. You know, when you come to Congress, you have a portfolio of expertise.

People come in, you've got physicians, you've got former service members, you've got people from the intelligence community. You have folks who spent time in state government or municipal government. And so you get a wide variety of experiences. But you're right, there's not a lot of folks that come from the technology industry in Congress.

Kevin Frazier: Well, and I think that part of the issue is we don't necessarily need all coders to become congressmen, right? That might actually be quite awkward, no offense to coders, but they're not always known for their eloquence or their capacity for deliberation. But I'll leave that for a different conversation.

But we do need to make sure that folks have awareness of what's going on in the technological domain. And I think most critically knowing what questions. To ask, because oftentimes we shouldn't expect that congressmen are in the weeds of what a world model is. Right? Right. But knowing to ask the question of what is a world model or what's coming down the pike is so important.

Yes. Have you found that to be the case?

Rep. Nick Begich: Well, it is. It's all about the questions. You know, there's a proverb that says he who asks the right questions cannot avoid the answers. And so, you know, this is something that I live by. Look, we don't know everything. There's no one person out there that knows it all.

And as I mentioned, people come in with different sets of experiences, but if you know how to drill down on core topics and you're not afraid to ask. Those questions you'll learn quickly. So one of the challenges of any member of Congress, and certainly this is true for me as being the only representative in the United States House from Alaska, right.

We're just a, we're a big state with a small population.

Kevin Frazier: One and done. Yeah. You only get one.

Rep. Nick Begich: I only me. So I don't have the ability to share my portfolio of responsibilities with other members of. My state delegation in the house. 'cause I'm, it, so if it's healthcare, I gotta know about that. If it's social security, I need to know about that.

If it's military, I need to know about that. If it's tech, I need to know about that. You name the issue. I've gotta know a lot about a lot of things. And so, you've gotta be able to ask the right questions to drill down and get the information that you need in order to make effective policy.

Kevin Frazier: No pressure on you there having to be a, you know, nick of all trades.

Rep. Nick Begich: Yeah.

Kevin Frazier: And so, with that mentality. I wonder, having a background in entrepreneurial activity, do you think that also lends a hand, especially on tackling some of these frontier issues of not necessarily being okay to just say, well, you know, D.C.’s gonna D.C., the swamps gonna be a swamp, but instead really pressing and saying, what can we do to actually solve problems here?

Rep. Nick Begich: Yeah, a hundred percent. Look when we worked with entrepreneurs, you know, we're navigating them as sort of a sherpa through that valley of death that, that so often talked about. And it was very simple at a high level, right? You gotta have a product. You gotta have a product that customers have proven with their, with voting, with their dollars that they're willing to pay for.

And you put those together and get investment and grow and scale that business. And so we would do that across industries. People would often ask me, well, you know, what industry did you focus on? We did everything. Mm-hmm, anything but video games we would be involved in. And so, you know, it gave me a lot of diversity of perspective and.

Sort of trained myself and my team to drill down on things we weren't familiar with at the outset, but use sort of first principles and say, okay, there is a similar pathway. It's a different context, but how do we navigate that pathway to something that's gonna be successful for the entrepreneur that we're working with?

So I think that when you come in to Congress, taking that approach can be very good. Not just because of asking good questions and understanding sort of those first principle models. How do we move things through to fruition? But how do you get creative? How do you do things that people haven't tried before or, you know, try a different angle on an old problem.

When you can do that, maybe you can bust through and get to solutions that have been elusive for a long period of time.

Kevin Frazier: Yeah. Congress is not exactly known for its creativity and I find it's just so helpful to have that breath of fresh air and in your case, very cool air coming from Alaska. Right?

I promise, I only make Alaska jokes this whole podcast, although it's very tempting. So I'm very proud of myself restraint. But thinking about the sherpa metaphor, which you brought up, not me.

Rep. Nick Begich: Yep.

Kevin Frazier: If we think of shepherding Congress through this AI challenge mm-hmm. How's Congress doing right now?

Where do you think we are? Where is the estimation of their capacity to kind of mount this very tricky, complex challenge? Where do we stand right now?

Rep. Nick Begich: Well, I think we're behind the curve. I think we're probably further ahead than most other nations. But. You know, AI is moving at a geometric rate.

Mm-hmm. Right. We're experiencing incredible sort of J-curve growth right now, and we don't know when we're gonna hit the S-part of that curve. It's even the strongest researchers in the space couldn't tell you where this starts to plateau. And so, you know, Congress has to deal with this issue, but Congress is built as a deliberative body.

The Congress is built to help ensure that we're making smart, durable policy. And so there, there's effectively a legislative constitutionally, instantiated lever on the speed at which we can get things done. And when you encounter something that's moving as quickly as AI, it can put a lot of stress and strain on a body that's intended to really deeply think about what policies are appropriate for any emergent industry or technology.

And so, look there's a lack of information in the Congress, a lack of awareness about not just what is AI, but what are the implications of artificial intelligence across the domains of responsibility that each of us have on the committees to which we've been assigned.

Kevin Frazier: And that's been one of my chief frustrations is we're more than three years into this. And depending who you ask, we may be more than 50 years into this, right? If you wanna say, well, we, you know, defined AI back in 1955, so on and so forth, but, it seems to me like we're still having somewhat elementary conversations about AI and the perhaps best signal of that is the fact that we're still just referring to this as the AI conversation, right?

As if,

Rep. Nick Begich: Right.

Kevin Frazier: It's a umbrella term that actually is sufficient to describe all of these myriad circumstances in which we're deploying it in health. Yes. And so on and so forth.

Rep. Nick Begich: There's so many different use cases. I mean, you can think of a use case every 30 seconds and we could be here all day long still thinking of use cases.

But you know, it reminds me of, you know, sort of that turn of the century conversation around the worldwide web and what we used to call e-business. And I recall, you know, there were entire classes on e-business. Now it's just business, right? The “e” has sort of woven its way into every aspect of commerce whether that's purchasing things online or engaging in sort of back office operations.

It's hard to imagine a situation, but this is how it used to work, where people were faxing things and copying things, and now we've got email, and we've got all these things. We thought of that at the time as somehow different from the core workflows that we engaged in. It was not different from the core workflows, but it took time for everyday folks out there to integrate this in a way that made sense and ad now it's just, it's sort of in the background.

I think AI may follow a similar path as people become more familiar with how this tool set is going to augment their existing workflows and automate in many cases their existing workflows. I think what's sort of different this time is the speed at which this integration is occurring.

And the disruption with which the integration may take place.

Kevin Frazier: Mm-hmm. Well, and I want to go back briefly to your point about Congress being somewhat slow on this, and some people will say, by golly, what the heck is wrong with Congress? They've had three years to figure this out. They should have passed a law within six months of GPT-3.5 being released.

But as you pointed out. Yeah. In some ways this is Congress working as intended, working as designed.

Rep. Nick Begich: Yeah. It, it is. And here's the thing too. You know, the members drive Congress, but you know who really drives Congress, the staff, right? And so, the Hill staff is adopting these tools faster for sure than the members of Congress.

Okay. And that's not a criticism to my colleagues, but it's a function of the work that needs to get done. You know, the staff within the Senate, the staff within the house, they are, we have limited budgets. Right? So we have a certain headcount that we can have and that headcount gets dedicated to certain things.

We're always finding that there's more work to do than there are people to do it. And so the staff has logically turned to AI to support workflows that will help to accelerate them in their daily work. And certainly, we've adopted that in our office. We help other offices understand that. But I think that becomes a driver for adoption and awareness inside of the membership structure, right?

Where the staff ultimately ends up driving awareness for the member.

Kevin Frazier: And this is one of my favorite parts about going and talking about AI. Whoever invites me to go to random corners of the country, including Alaska, I got to go more Anchorage and hang out with AG Stephen Cox, which was a welcome, fantastic invitation.

When you go and demo these products and just help folks realize that it's not going to blow up your computer or cause the end of humanity just seeing how boring AI is. That's what I love. I go and I talk to judges, and I often say, all right, here's a recent state Supreme Court opinion.

Rep. Nick Begich: Right?

Kevin Frazier: You have five minutes to read the entire opinion and give me a 500-word summary, and I say, go. And some of them don't even try to pull out their phone, right? Meanwhile, I just demo a deep research project and show how quickly I can get that analysis, and then I translate it into Spanish, and the scales just fall from their eyes, right?

And they realize how boring AI can be, and yet transformative.

Rep. Nick Begich: Transformative that is the right word for it.

Kevin Frazier: And so I wonder, for your own office—

Rep. Nick Begich: Mm-hmm.

Kevin Frazier: Because I don't think enough folks are hearing about AI actually being adopted in Congress. What does this look like for you all? How are you all using, using AI?

Rep. Nick Begich: Well, we've actually built a training program for our colleagues.

Kevin Frazier: Wow.

Rep. Nick Begich: Yeah. So we're establishing some thought leadership in the space. You know, and it goes from sort of beginner to maybe high level intermediate where you say, okay, what's a large language model? What are some of the tools out there that you can use?

Believe it or not, there's still people that don't know what these things are. Don't integrate them. Then how do you use how do you integrate this as a team? How do you build projects? How do you deploy agentics? How do you define core workflows in your daily tasks that could be accelerated or the quality improved, or both.

Right? How do you—How do you leverage the tool set as a force multiplier for you and and the member that you're working for? And I think that once people realize, similar to the situation that you just described to a judge, right, where you're saying, Hey, do this large volume of work in a short period of time and then translate it into a language you may not know well or at all.

Once we walk people through those use cases, adoption becomes the natural progression.

Kevin Frazier: And I think that a central challenge is just being transparent and helping, for example, your constituents understand that force multiplier effect.

Rep. Nick Begich: Yes.

Kevin Frazier: Because if the question is, do you want staff alone going through hundreds, if not thousands of constituent inquiries and perhaps getting to that response, weeks, if not months later. Or being augmented with AI, being able to triage, Hey, this request actually needs attention right now. And so on and so forth. That to me, helps folks see, okay, this is the true value add.

We're not using this to try to dumb down our work or give you a bad response. Right. We wanna reach out and prioritize certain responses.

Rep. Nick Begich: That's correct. And, you know, I'll tell you on the flip side, it can jam up in office. Right, because a lot of folks who spend time sending communications to offices they can pump out a lot more communication as well.

And so that it works both ways. One of the things that I think is really interesting that we've observed in our office and other offices on the Hill, and I think may be common outside of politics is the idea that a, that the staff can spin up a lot of work for a physical individual who has to make decisions, right?

And so you find the number of decisions that you make in a given day can dramatically increase. Now the human in the middle, kind of bottleneck, becomes very apparent when your team has the tools to be five times more productive and provides you with five times more of the decision making that needs to be done.

So that I think becomes sort of an, a hidden governing factor to the AI adoption curve, is ultimately there will be humans in some of these nodal decision points that have to make a call up, down, left, right? Yes, no. And their cognitive load can sort of act as a governor on adoption.

Kevin Frazier: That's one of the more exciting questions to me, is this kind of future of governance question of how do we reimagine how constituent services are done, how legislative drafting occurs.

Even crazy ideas like using digital twins to simulate how legislation is going to actually be implemented in the real world. All these crazy 20, 30, 20, 35 scenarios that we can think about. Maybe not that far off.

Rep. Nick Begich: Maybe not that far off.

Kevin Frazier: Maybe not that far off. Yeah. To ground this in some more mundane question,

Rep. Nick Begich: Sure.

Kevin Frazier: We're seeing a tremendous amount of activity at the state level right now, when it comes to regulating AI. And one of the difficulties I think has been sort of having all states see that this is indeed a whole of nation approach, right? Where this is an all hands-on deck moment where we need data centers across the country. We need data from individuals across the country.

And we need AI adoption to be taking place across the country. And yet in your own backyard, snowy. Very far away. We're seeing some debates about when and how Alaska, for example, should be building out data centers. Can you detail what's going on in your own backyard?

Rep. Nick Begich: You know, it's interesting because I think a lot of times folks, policy makers in some parts of our state, I mean we're a big state, different perspectives throughout the state, but in some parts of our state policy makers look at what's happening in the debates here in the lower 48, right?

And, and they say, Hey, if that's an issue there, do we need to think about it as an issue here? Now we've heard about the question of social license as it relates to data centers, right? People have an expressed concern in many areas of the country about the impact on their utility bills.

And I think data centers done right, actually lower the costs for consumers. And I think that the industry as a whole probably needs to do a better job communicating where that happens, and make sure that happens at a minimum, that those rates are not rising appreciably for consumers of energy.

What's unique about Alaska perhaps among the other states is that we have hundreds of trillions of cubic feet of natural gas stranded right now on Alaska's north slope. We've got great connectivity to the state, right? We've got nice fiber lines that go down to the lower 48, connect us to the rest of the world, great satellite coverage up there as well. For compute intensive data centers, Alaska's a perfect spot. It's cold, right?

Kevin Frazier: We know that.

Rep. Nick Begich: We know that. But it's great for data centers. Lowers your power consumption costs. We got a, the largest pool of energy that's untapped in North America, sitting right there waiting for some folks to come up and tap it. What's great about Alaska's model is that this, the state of Alaska and its citizens share in the wealth of the resources.

So unlike other states in the country, and this is what a lot of people don't realize, like down in Texas, all of the resource wealth is privatized in. In Alaska that resource wealth is shared. So we have a sovereign wealth fund called the Permanent Fund in Alaska, and the Permanent Fund issues, what we call the Permanent Fund dividend, which is a payment to every resident of Alaska based on the performance of that fund.

Well, the fund is initially funded by oil and gas and mineral revenues, right? So a portion of that goes into the fund and then is invested in a diversified portfolio of private equity, VC stocks, bonds, et cetera. So, we are aligned because if we have a mechanism to monetize those energy resources, it benefits everyone in the state.

And I think that's a unique model. You don't see it, to my knowledge, anywhere else in the country where you get a direct benefit from the resource development that's happening in the state. So the state, generally, very supportive of responsible resource development. And I think there's a real opportunity in Alaska for data centers, particularly low resource requirement data centers where it's commodity equipment, you know, blade servers pop one in, pop one out. It doesn't require huge staffing. It's not multi-tenant generally.

That's perfect for us and I think there are a lot of opportunities for people to explore in Alaska. We're certainly ready and willing to facilitate that.

Kevin Frazier: Yeah, and I think this speaks to the fact that a lot of these conversations occur without any nuance.

Right? It's data center, good or bad.

Rep. Nick Begich: Yep.

Kevin Frazier: There's no sort of consideration that there are different kinds of data centers. Data centers serve different purposes for different communities, require different amounts of energy, so on and so forth.

Rep. Nick Begich: That's right.

Kevin Frazier: And so just having that more nuanced conversation and getting to that basic literacy can really help have a more, for lack of a better phrase, adult conversation.

Rep. Nick Begich: What, you're 100% right. I mean, a multi-tenant data center has different requirements than single tenant data centers. When you're talking about commodity equipment, that's different. When you're talking about the fact that, you know, AI and crypto are compute intensive. Right. They're not really data intensive, right?

They're, once the data's in the data center, if you're running a training model, it's there, right? And the queries that go into an AI model, the LLM right, are very small and the packets that come out are very small. So, you know, connectivity is less of a constraint than it perhaps used to be or would be under a multi-tenant data center, traditional web data center, model.

Kevin Frazier: So where's the log jam in your opinion? Because even though we're having this highly sophisticated conversation about data centers and getting into the weeds of it. We still see these headlines about data center, good data center bad. And we still see some fellow members of Congress proposing things like moratorium on data center development. And so is this a matter of just political incentives? Is this a matter of a lack of literacy? Is it both?

Rep. Nick Begich: Yes.

Kevin Frazier: And feel free to name names. Yes. Or be as vicious as you'd like.

Rep. Nick Begich: Well, yeah. It's all the above. I mean, a lot of this is political opportunism. People thinking I'll jump on this message and, you know, it will help me politically with a constituency back home.

It can be someone coming in a, a lobbyist or some, somebody bending a member's here and changing their perspective on something without providing full facts. It can also be outta genuine concern where someone says, you know, I, this is my perception right now. And here again, you know, the AI leadership needs to be on the Hill—and they are.

But they, they need to continue to work individual members, to provide them with accurate information about how this helps communities, how this is good for the United States, how it's imperative, strategically all those things are important considerations before people start putting legislation in motion.

Kevin Frazier: We know that a lie travels at least twice as fast as the truth. Right?

Rep. Nick Begich: That's right.

Kevin Frazier: We'll attribute that to Mark Twain or whoever he would've liked us to attribute it to. And I think we've seen some of those early hot takes on data center, energy usage and water usage, for example have become just so entrenched in people's minds.

And yet I very much encourage any listener to of the pod to follow Andy Masley. He's just got so much great information on this front and helps really show that we need to be more in depth about what it means for data center development to occur, and that perhaps it isn't as onerous and as destructive as some people would.

Rep. Nick Begich: It's really not. And I think one of the things that we're tracking certainly in my office is look at the—Look at the compute curve inside of crypto as a forecaster of where things go in AI, right? So, if you're talking about Bitcoin mining as an example, SHA-256 you know, hashing has become pretty much a dead on commodity, right?

So no longer is it how much compute can I fit onto a chip or sell you? Right? It's how much compute per watt, right? And so the innovation curve has gone to power efficiency. And I think AI will follow ultimately a similar curve because it's so, compute intensive and it's so energy intensive.

I think that we're gonna continue to see investments in that part of the curve. Okay, we know what compute looks like, how do we make it more energy efficient? 'cause that's the fastest pathway to increasing the amount of ai power that we deploy. Because right now it's the energy that's becoming the limiting factor for adoption speed, not the ability to produce chip sets.

And so, as that continues to magnify as a constraint, I think the industry's R and D budgets will align to efficiency, right? And that energy efficiency is going to ultimately be what dictates the curve for AI progress on a chip set.

Kevin Frazier: It's telling that a lot of the major labs are making these sorts of incredible investments in clean, renewable power, for example, to drive these data centers forward because they don't want high operational costs, right? They want to be operating for as long and as cleanly as possible.

Rep. Nick Begich: That's right. But there's also—if we're going to decentralize AI. Right, because right now it's a, data centers are effectively a centralized AI conversation.

But if you want to push AI into the edge of the network and have basically semi-autonomous AI devices, efficiency's the name of the game,

Kevin Frazier: And that's where those sort of smaller data centers you were mentioning could really play a role, especially in places like Alaska or other rural locations.

But, I know we've been framing this in a fairly positive sense. I do wanna acknowledge that data centers, of course, cause disruption to communities. There's new transmission lines that are gonna be developed, new cars, trucks construction that's going on, so on and so forth.

But this is where I think the framing of this AI build out as a national prerogative really has to be made clear, because surely no one wants electricity lines being built across their farm or a new interstate highway coming through their community. But for the development of these critical infrastructures that have unleashed and created new markets, it is something that everyone ultimately has to contribute to.

Rep. Nick Begich: They're gonna benefit from it. And this is where, you know, I just challenged the assumption that folks don't want these transmission lines, don't want these roads coming through their communities. There'll always be folks that say, look you know, this is impacting me. We have to work with that.

There's gonna be folks who are always gonna complain. There are folks that are actually anti-growth. Don't wanna see development at all. But I think when we're talking about infrastructure, roads, highways, bridges, when you're talking about transmission lines, a lot of this infrastructure hasn't been refreshed in decades and is aging, it's crumbling out and it needs to be replaced. And this gives us a true business case to drive that forward.

So I think there's gonna be a lot of benefits for communities. Yes, there will be some impacts, but at the end of the day, in the aggregate, I think that it's gonna bring a lot of benefits to folks.

Kevin Frazier: And it's really that long-term perspective that we often need in these infrastructure conversations, because of course, if you ask someone, would you like a jackhammer going off in your backyard from eight to five?

No one's gonna be, no one's gonna be like sign me up. I can't wait to have this Zoom meeting interrupted or so on and so forth. But that long-term perspective of now we've got this new market, we've got these new opportunities, this new airport, so on and so forth. Those are the things that drive progress, that create new jobs, and so on and so forth.

Rep. Nick Begich: That's right. And look, if you're not growing, you're dying. Right? There's no such thing as stasis, right? I mean, even when things look like it's stasis, you've got depreciation on assets, things that are hidden, you don't see, we need growth in this country. It's good for us, and Lord knows other nations are gonna pick up that ball if we don't.

Kevin Frazier: Well, and this is where I'm so excited that we're talking during the 250th anniversary, because if you look back through all of our history, it's the fact that we had the Erie Canal. It's the fact that we led on the railroad. It's the fact that we built out the whole internet infrastructure that have been at the core of all of our future economic growth.

But turning to your own agenda.

Rep. Nick Begich: Yeah.

Kevin Frazier: You're a freshman, you're diving into the weeds from the get-go.

Rep. Nick Begich: Yes.

Kevin Frazier: What are some of your priorities from a AI perspective in terms of the bills you're supporting or the bills you're advocating for and advancing yourself?

Rep. Nick Begich: Sure. Yeah. So, I'm proud to report, you know, we have passed more legislation than in anyone in Congress so far in the 119th Congress as a freshman office,

Kevin Frazier: Rookie of the year.

Rep. Nick Begich: Well, That, we'll see.

Kevin Frazier: You've got my nomination.

Rep. Nick Begich: Okay. So we've been effective at getting things done. Now some of those things are Alaska. Specific, but some of the Alaska specific legislation that we have carried has national implications. Okay. So, what we did in the budget reconciliation bill last year, and this of course didn't get a whole lot of press nationally, a lot of press back home, we mandated 30 million acres of oil and gas lease sales over the next 10 years.

Now, this is important because if we're gonna unlock the natural gas resources of Alaska, we've gotta, we've gotta be able to have that in law. It one, one of the things that happens, and we all see it in our, in various ways in our communities, when one president comes in, they've got one perspective.

When another president comes in, that person has another perspective, and that pendulum of activity can really disrupt long-term, particularly infrastructure investment and certainly AI investment. And so what we're looking for is durable law that allows us to build predictability because if you've got an investor that's gonna make an investment in a 20- or 30-year asset, well the payback period on that's probably gonna be at minimum five years, probably more like seven to 10 years, and you can't have a pendulum swinging around during that period of time because that's where you're gonna be able to justify the investment.

And so we've looked for opportunities in Alaska to make us attractive as a state and to provide those energy resources to whoever may need it, AI being right, right up there on the list. Now I'm also getting ready to introduce the House version of the DATA Act—and the DATA Act, for those who haven't been tracking what's happening on this bill, it was introduced by Tom Cotton in the Senate and the bill’s essential premise is to reduce the amount of regulatory burden required for new AI data center capacity.

So, traditionally if you're connecting into the grid, you've got a huge regulatory set of action items that you've gotta check off before you can deploy new power into the grid. And a lot of requirements as it relates to what happens in, in down states where you know you've got some power coming offline, you've got a, you have a responsibility to feed your power into the grid.

You can't necessarily be the sole consumer of that power. In fact, you can't be the sole consumer of that power when you're connecting into the rest of the grid. Well, this would allow folks to build out completely behind the meter. In fact, there wouldn't be a meter. It would just be their own independent power that they, if they're the sole consumer of that power, they would not be subject to those regulations.

And this would allow the data center community, if passed, to be able to say, look, incredibly so our power draw does not impact the consumer whatsoever because we're not connecting into the grid. We're not drawing power from the grid. We're utilizing our own power source.

And this is consistent with what we've seen from large players like Meta, Anthropic, and others who are looking seriously or are moving forward in building out their own separate capacity. This would help to accelerate that.

Kevin Frazier: Mm-hmm. And I think that longer term infrastructure vision is so important because when you talk to folks about AI, everyone seems to be asking what's gonna happen next week? What's gonna happen next month? And when you have that sort of rush to regulate, because you hear everyone talking about what's coming right around the corner,

Rep. Nick Begich: Right.

Kevin Frazier: You forget that we're gonna be at this for decades. This isn't going away. This is going to be something we really have to build out and sustain again if we're going to compete with our competitors and with our adversaries.

Rep. Nick Begich: That's right. That's correct. It's, this is a long term play.

This is a long-term layer that is being put into place and we have to prepare for not just what we know of today, but prepare the infrastructure for what may be around the corner that we don't necessarily have a way to envision, but we need to create a space for.

Kevin Frazier: And can you speak to, from the perspective of an entrepreneur, why this sort of patchwork problem is actually a problem?

Because some folks will say, look, you know, startups, they've been carved out from these regulations, or perhaps they're operating at such a small scale that 50 state requirements, they don't really care.

But from your perspective, earlier you mentioned the importance of stability from an investor standpoint.

Rep. Nick Begich: That's right.

Kevin Frazier: And this is where I think folks kind of black out and forget the fact that if you are an investor, which is the only reason we have a startup economy, people willing to gamble on the small guy is you need predictability. You're not gonna say—

Rep. Nick Begich: It's an indispensable component.

Kevin Frazier: Right. Yes.

Rep. Nick Begich: There are many components you need, but you can't do it without that component. For sure.

Kevin Frazier: You do not escape the garage if you can't open the garage door. And so finding ways for investors to feel like, okay, I can put a sizable amount of money in here because I see the legal path forward.

Rep. Nick Begich: Right.

Kevin Frazier: That's actually imperative. Can you talk it's imperative to why that's a critical view?

Rep. Nick Begich: Well, it's imperative because look for one, your total addressable market—It shrinks every time that one state or other jurisdiction decides that they wanna do things differently or not at all. And so, that becomes a challenge, right? Hitting critical mass and scale.

 Because remember, it's not just what we're building here, but these products will be utilized around the world, right? And so if anytime that you're reducing that market size, you're diminishing the investor case, for one.

Two, you diminish the innovation envelope, right? And what we've seen in what the federal government did, and I think wisely at the dawn of the worldwide web was, sort of assure that there would be a level playing field that platforms wouldn't be heavily responsible for the content as long as they weren't editing that content. That's something that's come up over the last few years, right?

We've seen platforms decide that they want to start managing, curating content, elevating some over others, and that kind of crosses over into a different space, but I think there has to be room for platforms to innovate, to make sure there's a, an addressable market and to make sure that the larger states in the country don't drive the results for everybody else.

And we see this in auto manufacturing. I mentioned earlier I used to work at Ford Motor Company when California passes a law with respect to auto manufacturers, it can forcibly change what happens in the rest of the country because of scale economics, because of shipping regulations for a whole host of reasons.

And we don't want to see that approach taken in a space that's rapidly evolving and rapidly growing and cause the nation to miss out on a generational opportunity. There, there's been a lot of thought leadership dedicated to what happens in the J-curve if you fall behind and if you're six months behind by many metrics you might as well be in the Stone Age.

That's how fast this is evolving, and so we can't afford to wait around and see where things land. If we do this wrong, we don't get, we only get one shot at this

Kevin Frazier: Well, and it is important to emphasize too that the current AI trajectory we're on isn't necessarily the end all be all right. We may see new branches, new technologies we mentioned earlier, world models, for example.

Rep. Nick Begich: Absolutely. That's right.

Kevin Frazier: And if we foreclose discovering those new branches while other countries are racing ahead on that critical R and D, for example. Then to your point, we may become way far behind, way faster than we could ever have imagined.

Rep. Nick Begich: That's 100% right. I mean, you've got the, most people are familiar with the large language models because that's what they see and touch.

But as these tools become more specialized to specific industries and models get, get tweaked and constructed around specific use case sets or arrays, then you really start to see a lot of snowball innovation in frontier science. Right?

And so a lot of what you see right now is corollary. It's okay, you know, this model's been trained on this dataset. I'm not fully familiar with the dataset as an individual. It dives in, it finds information that I wouldn't have found quickly, right? So a lot of what we see with LLM is an acceleration of what a human would have been able to do with enough time, right? So it's, you can think of some of this, some of these use cases as a search engine on steroids.

But when you start to talk about what we're seeing in mathematics, in physics, in chemistry, right, in pharmaceuticals, these are specific models that may use some of the base technology, but they've been tweaked to really specialize in pushing science forward.

You don't wanna miss out on those opportunities. That's where real transformation comes in.

Kevin Frazier: And those are the boring AI use cases that don't make the New York Times headlines, right, of 2% marginal improvement in material sciences just doesn't sell papers. But it is the thing that lays the foundation for so much progress ahead.

Rep. Nick Begich: Absolutely. Yeah, you think about rare diseases, you think about agricultural yield, some of these things, they're exciting to me. Probably not exciting to many readers, but I think this is where, you know, we can see sort of that abundance perspective on AI take shape.

You know, there's been a lot of discussion around which way does the curve ultimately go with respect to its impact on humanity. I'm one who believes that, you know, there is a real strong and exciting prospect for broad based abundance that carries forward the abundance we saw in the last 150, 200 years.

Where you've seen huge portion of the world's population lifted out of poverty, I think we can have some real transformational abundance that can come with the innovations that are discovered with AI over the next decade or so.

Kevin Frazier: Right. And that progress didn't happen by accident, right? It was investment in critical infrastructure.

Rep. Nick Begich: That's right.

Kevin Frazier: And that occurred over decades and with a lot of risks. And a lot of costs. I know you've got plenty of things to be doing, plenty of things to go back to on the Hill, before I let you go a couple rapid questions. First: What's your favorite model or favorite use case of AI?

Rep. Nick Begich: That's a good question, and I will, I'm gonna cheat a little bit. So, you know, I started out using Chat-GPT. We still use it in the office. I think it's got some, it's got some good qualities. It's fast, it's responsive.

When I want, the one I use the most is Grok, when I want unfiltered information, quick information, and something that doesn't have a lot of sort of social friction between me and the truth, I'm looking for Grok.

On complex projects, Claude has been great. So, if I need to do web development, believe it or not, I'm a member of Congress who still writes code.

Kevin Frazier: There we go.

Rep. Nick Begich: Yeah, there's one, we know, one. I'm sure there's a couple others at least, but I'll use Claude for that.

It's a tremendous accelerator for what I'm trying to do, it builds beautifully attractive static HTML React applications on the fly—I wish I had a little bit more token allocation on my subscription, but,

Kevin Frazier: Dario, if we, if we can help a congressman out here, please.

Rep. Nick Begich: But I think it's just, I use a different tool for different use cases.

They all have—the ones that I use anyway, all have great features and they're continuously innovating. Love to see what's happening both at Grok and on, at Claude with respect to projects and your ability to upload contextual files and information and kind of keep context, retain context over time.

To me, that's very exciting and it allows you to sort of build more than just one session can provide.

Kevin Frazier: And now I'll let you play professor.

Rep. Nick Begich: Yeah.

Kevin Frazier: And you get to assign homework to every member of Congress. It can be a podcast, something to read or an activity, perhaps building something with Claude.

What would your AI homework be for Congress?

Rep. Nick Begich: With Claude?

Kevin Frazier: Well, anything, or anything, any AI related?

Rep. Nick Begich: Well, for folks who've never touched it, I'd say start with Chat GPT or Grok. Just start asking questions. You know, as we started at the beginning of this discussion, questions lead you to the answers and I mean, even the most just curious persons got questions, so, so for things that are burning on your mind. Go to AI and start asking questions and drill down.

And when we meet with constituents constantly all day long, I'm meeting with trade groups, I'm meeting with individuals who come from my home state or may come from another state, people who have specific issues, questions. I don't know the answer to all those questions.

And so oftentimes, you know, we will dig in, what can you tell us? What do we know? And so I would say start using it in your daily activities and build from there.

Kevin Frazier: Build from there. Yeah. Okay. Alright. Not the hardest homework assignment.

Rep. Nick Begich: No, it's pretty. Pretty straightforward. Yeah. And that's the beauty of these tools, they've been built to be accessible.

I remember when I sat in front of Chat GPT the first time, okay, how does this work? And I asked one question, I said, well, that answers that. There we go. It's very straightforward.

Kevin Frazier: Easy-sneezy.

Rep. Nick Begich: Yeah.

Kevin Frazier: Well, here may be a toughie. My wife is a very patient woman, and I've yet to take her on a honeymoon, and Alaska is near the top of our list.

What is one spot that should be at the top of our list, if we do indeed end up going to Alaska?

Rep. Nick Begich: Oh my gosh. If I tell you one spot I'm gonna be in trouble.

Kevin Frazier: One? I know. Okay. I'll give you, I'll give you, I'll say pick one out of your top 10. I'll give you, now you have some coverage.

Rep. Nick Begich: I'll give you a few, I'll give you some great spots.

I just returned from southeast Alaska. A lot of folks take a cruise through southeast Alaska. Little, start usually in Seattle, and they'll come up to Ketchikan, they'll hit Sitka, Skagway, Juneau, of course. Sometimes they'll go all the way up to Anchorage. There's a lot of great places on the coast, what we call coastal Alaska and southeast, and that Alaskan peninsula that comes down and borders Canada.

I have to remind people, you know, for me, that's my southern border: Canada. But any—

Kevin Frazier: Right. Yeah. The other southern border.

Rep. Nick Begich: The other southern border, yes. But there's other great places in the state as well. The people go up to Fairbanks and they, in the wintertime, they'll watch the northern lights, which can be incredible.

Anchorage is a jumping off point to so many great places on the Kenai Peninsula where you can go fishing, you can go deep sea fishing, you can go salmon fishing on the Kenai River. There, there's not a bad place to pick. And if you're really up for an adventure, you can go to Utqiagvik, or what's commonly known as Barrow, the northern-most settlement on the North American continent and check that out.

And so there's some amazing places. It's a big state, two and a half times the size of Texas, with about 750,000 people in it.

Kevin Frazier: Well, I usually say everything's bigger in Texas, but I guess the exception is Alaska.

Rep. Nick Begich: The exception is Alaska.

Kevin Frazier: Representative Begich, thank you so much for coming on Scaling Laws.

Rep. Nick Begich: Thanks for having me. It's been a great conversation.

[Outro]

Kevin Frazier: Scaling Laws is a joint production of Lawfare and the University of Texas School of Law. You can get an ad-free version of this and other Lawfare podcasts by becoming a material subscriber at our website, lawfaremedia.org/support. You'll also get access to special events and other content available only to our supporters.

Please rate and review us wherever you get your podcasts. Check out our written work at lawfaremedia.org. You can also follow us on X and Bluesky.

This podcast was edited by Noam Osband of Goat Rodeo. Our music is from ALIBI.

As always, thanks for listening.