U.S. Congressman Beyer on AI challenges facing America and the World

0:01Welcome to the Practical AI Podcast, where we break down the real-world applications of artificial intelligence and how it's shaping the way we live, work, and create. Our goal is to help make AI technology practical, productive, and accessible to everyone. Whether you're a developer, business leader, or just curious about the tech behind the buzz, you're in the right place. Be sure to connect with us on LinkedIn, X, or Blue Sky to stay up to date with episode drops, behind-the-scenes content, and AI insights. You can learn more at practicalai.fm. Now, on to the show.

0:41Welcome to another episode of the Practical AI Podcast. I'm Chris Benson. I am a principal AI and autonomy engineer, and today we have a special guest who has been a previous guest from a couple of years ago. I want to introduce, if you haven't already seen the episode or recognize him up front, this is Congressman Don Beyer of Virginia, who is, in addition to being a congressman, has an incredible background in AI, which is obviously why we're having him on this particular show today. Welcome back to the show. It's great to have you. Chris, thank you. I'm flattered that you invited me back with Second Thought.

1:20Well, the first time was very inspirational. I know it's not the primary topic, but I remember one of the things that really had an effect on me was you were in a PhD program at George Mason University in AI. And I would bet that most members of Congress don't delve into such things. And so whether you like it or not, I think that makes you the coolest member of Congress, period, the fact that you're doing that. So thanks for coming on the show to talk a bit about the world of AI and how it touches you and your primary job. Yeah, thank you. It's really fun. I'm spoiled because I live so close to the capital.

2:01Northern Virginia is right across the river, so I don't have to be on an airplane for 8 or 10 hours a week like most of my fellow Congress people do. Fair enough. Fair enough. You got those extra few hours to work on that PhD program there. Um, so, uh, and we got a lot of feedback, uh, from that when we were on a couple of years ago, really positive. Uh, so, uh, anyway, welcome back. You know, the landscape, uh, of the world has changed dramatically since the last time we talked to you. Um, where we have a new, uh, a new administration that's, that's in versus, uh, uh, president Biden was in back when we talked, We now have President Trump.

2:44We were talking about a whole set of public policies that were that were in kind of being developed at the time. And I know that that has changed. This administration has thrown out a lot of the work that had been done prior and and kind of gone their own way. I was wondering if you could kind of start out by just kind of laying the landscape as you look at it as an AI expert who is in Congress. How has the world changed from your perspective? What's the same? What's different? And how has that changed how you're looking at things and acting upon them? Well, Chris, almost nothing's the same. Just because AI is accelerating so very, very quickly.

3:29In terms of the Trump administration itself, it's sort of a mixed bag. On the one hand, the new president throughout Joe Biden's executive order on AI, which was incidentally the largest executive order ever written by a president. But then he instituted his own, which was largely the same. The probably most important thing for me, and I think for most people on the planet, is that Donald Trump saved the Safety Institute out at NIST, the National Institute for Standards and Technology. He renamed it. It's called Casey. Kept a lot of the same people. It changed the leadership, but that's normal.

4:04So at least there is some safety perspective within the current administration. He brought in David Sachs to be the AI czar and Mike Kratios to be chief science advisor. That was interesting because it basically brought in two business people rather than scientists who made a lot of money, both of them in Silicon Valley doing that. That was different from the scientists that he had before. You know, for example, the head of OSTP was a distinguished scientist out of Stanford. There was a push in the first year or so of the Trump administration towards, you know, full-blow accelerationism. You know, no new laws, no restraint.

4:46From the defense perspective, the intel perspective, it's like, we have to beat China. But it was also mixed because then at the same time, Trump decided to sell a bunch of the H200 chips to China, which was opposite of what Biden did. Biden tried to restrict China's growth in AI by withholding the best NVIDIA chips, and Trump reversed that for other reasons. So it's a complicated scenario. And then with the anthropic battles over the use of anthropic in Iran, or even more so with Mythos' introduction a few weeks ago. It actually didn't come to me, but it was laid out what was happening with it.

5:29All of a sudden, there was this wake-up call within the administration that AI is progressing so quickly, it could endanger all of the cybersecurity measures that American companies and American government put in place over the decades, and that they really have to pay attention to the security and the safety sides of artificial intelligence. Yeah, it's Mythos in particular, you know, with kind of the shock. It was interesting that there was the, you know, the administration was very much in battle with Anthropoc for a little while. The Mythos came out. They seem to be backing away from that slowly.

6:08At least that's how it seems coming across in terms of the comments out of the White House. I'm kind of curious from more of the AI perspective on this as we have seen each advancing model and supporting technologies such as the various agentic harnesses and stuff like that come out once something is out you continue to have development from other companies and stuff like that how do you think with Mythos coming out and the scramble to address safety concerns that it brings, but the recognition that you're probably going to have other models that are able to have similar capability evolving over time from various companies, whether they're domestic or foreign.

6:56How is that, you know, are folks thinking about such things, about the evolution from mythos on? What is the safety picture there? Yeah, Chris, I think very much so. You know, I had been in the more comfortable position of thinking that all of our protections against cybersecurity inclusions, our password protection writ large with multiple layers of protection, that we're going to be safe until quantum computing came. But no, one of the things that Mythos did was to look deeply into how the protection efforts were created and began to unravel them, unfold them right away. And yeah, sure, if Anthropic can do it, you've got to figure OpenAI and Gemini and stuff are close behind.

7:40the Chinese tool. I thought that Anthropic was very responsible in giving the code to a handful of people to anticipate how we were going to have to strengthen our security devices ahead of any widespread dissemination of the METHOS software. Hopefully, we can get a running start before the other people catch up. But it's like any arms race. It's not going away. This is going to be step-by-step accelerating for the indefinite future. Yeah, I mean, just that, and there are so many topics to hit here, but with the mythos and as other competing capabilities, you know, come out, and some may just be released without any, you know, there's a point here where Anthropics talked about kind of holding it back for a while, but I think the presumption is that mythos will be generally available at some point and likely competing models.

8:37So it kind of seems to be changing. I know coming from the AI, you know, where AI merges with the software development world, it's definitely changed how everyone's looking at their cyber position. Oh, absolutely. And it would not surprise me, Chris, if there is a wholesale rethinking of how cybersecurity works. If we look and say that the tools we've used are not going to be effective anymore, we step back two or three or four steps and think about what do we need to protect? even the question of what do we need to protect? Do we choose to protect much less? And do we protect it in very different ways?

9:13It's a phase shift, really big picture. Absolutely. So as we step back to kind of the notion of government, we're seeing a lot of conversation around regulation. There is an ongoing debate about where regulation should occur, whether it's at the federal level or at the state level here in the United States, and a little bit of a struggle. Can you talk a little bit about how you see that landscape in terms of regulation? Who should be doing what at what levels of government? And what is a sensible approach to that? And are we doing that or not? Well, I think in the largest picture, the sensible approach is a new Geneva Convention where we get together with the Chinese and the Europeans and the people in the Middle East and anyone who's doing important work and trying to figure out guardrails that work for everyone.

10:07As people have fairly pointed out, if we have this beautiful regulatory system in the United States and China has none, that is not going to work in the long run. In the meantime, we have a more continental concern, and that's should the federal government do the regulation or can state and local governments do it? And I think I'm very sympathetic to the argument, probably best made by my friend Jay Obernolte, who's a Republican congressman from California, that we shouldn't have the Tower of Babel with Virginia's regulations be different from California's being different from Texas's. However, at the federal level right now, we've basically done just one bill, and that was Ted Cruz's Take It Down Act, which gave us the ability to, when someone puts evil sexual imagery of you, Chris, up on Facebook.

10:58Don't horrify the audience with that. Oh, my gosh. You can demand that Adam takes it down and maybe even have a cause of action to sue whoever put it up. That's a good thing. only one that we've done. And you're familiar with the bipartisan task force that Mike Johnson, Hakeem Jeffries had. We had 80 specific legislative recommendations, and so far we've done one of them. We look at Congress's total inability to do anything on social media over these last two and a half decades and say, yes, we need a national framework, but in the absence of one, we should not restrict state and local governments from doing the best they can.

11:35And there are interesting things. HB 53 in California, you know, I believe that the governor vetoed it or amended it, but it became law, is an important first step in understanding how to regulate the artificial intelligence. There's a guy named Alex Boris in New York, a member of their Albany Assembly, who, again, is out there trying to think of really important ways where a state can make a difference. That shouldn't be the end game, but it's probably a good place to start. The freeway line is that state governments are the laboratories of democracy. They can move much more quickly. They don't have filibusters and things like that.

12:15And then maybe we can learn from, I guess, over 700 pieces of local, of state legislation on AR out there right now to build where we go. I hope it doesn't take 10 years. Maybe it should take two or three years. What it will take, Chris, is an administration that wants to press forward with meaningful light touch regulation at the federal level. That's not come from President Trump and this administration yet. They've not, you know, we do a lot of stuff on not taxing tips or overtime, but nothing on AI regulation. Yeah. So I'm kind of curious. It raises a question. I'm failing to see it. Is there something partisan inherently in AI?

12:59And this is where acknowledging I am not a political figure and that's not where I'm spending my thinking time. But is it perceived as a partisan topic in general and in the large? I think there's a danger that it does it that little way. You know, our little task force was completely bipartisan. We're trying to do everything as bipartisan as we can because this really affects every person. It should be much more like, you know, our defense posture, which has typically been very, very partisan, our foreign policy posture. The one thing that complicates it, Chris, is that typically Democrats have been more inclined to regulate and Republicans have been much more inclined to deregulate.

13:39And so when you use the word regulation or even the idea of putting restrictions around what artificial intelligence can do, can be, how it's used, it's going to stir a little bit of that D versus R. But we have to do what we can to overcome. Gotcha. It's, could you talk a little bit about it, Lee, when you talked about some of the, you know, we out here, you know, that are not in government and not, you know, we see it on the TV, we see it on the browser and stuff. But, you know, I think we hear so much divisive language. And you talked a little bit about the ability of Democrats and Republicans to kind of come together and get and try to, you know, try to agree on topics.

14:25You talked about foreign policy. Within AI, what, if any, are some of the strengths of, you know, you see both sides of the aisle working together. Are there any, first of all? And is that something that is working at any level? I recognize that the current administration kind of wants to go their own way in a variety of topics. But is there any silver lining there? I think so, Chris. I mean, I certainly want there to be. Let's just think, for example, one of the big concerns with AI is surveillance. You know, that, you know, and this came to the fore, especially when Doge came in and copied a bunch of Social Security and tax records, loaded them, we think loaded them into Grok, Elon Musk's thing.

15:14And I know my Republican friends well. I'm a Democrat for those who didn't know. That the last thing they want is a central government that knows everything about them. I mean, that's the reason my Republican colleagues have not wanted gun registration over all these years, because that means that the government then knows exactly who owns what gun and they can take it from it. I don't think they want the government to know everything about us, you know, our habits, what book we read, what time we go to bed, you know, tracking our location devices in our cars or our phones. So I think there's agreement there.

15:50Certainly there's agreement on the sexual imagery, the misuse of generative AI visually, et cetera. And then I think where there's probably the greatest concern right now is what about the job displacement? We're all familiar with Dario Amadei's predictions about 25%, 50 % white collar job displacement in the next two to five years. You know, the different numbers are out there. But we all know that this is not going to be at the speed of the agricultural revolution or the industrial revolution, which took place over decades or a century. This could be two to five years. I had a speech opportunity this morning with the enrolled agents of America, all the folks that do our taxes, CPAs and small accounting firms.

16:34And it's very relevant to them that all of a sudden, if all those functions like accounts payable and accounts receivable and payroll production are all agentic AI handled, what does that do to not necessarily to their jobs, but to the people who work for it? And just extend that through the entire 18 million white collar workers we have in America. and once again we did a terrible job of adapting to the job dislocation in the manufacturing sector that came both from trade and from even more from technology so you have all these wiped out former manufacturing towns especially in the midwest but around the country you know our so-called trade adjustment assistants didn't do a very good job at finding them new ways to be productive, to have the dignity of work.

17:23And that's a big challenge that both Dems and Republicans are facing. And what's the thinking? I mean, that is a topic, and we talk about it on this podcast all the time, in terms of concerns over jobs. If I have, you know, we have holidays and, you know, come Thanksgiving, Christmas time, when we're having extended family around, and none of which in my extended family are AI people other than myself. But that is certainly the topic that everyone is worried about. And within our family, we have an array of different jobs that people are in. Some are blue-collar, some are white-collar. is, I guess, one of the things, kind of channeling some of the questions that I get that I cannot answer from my own family is, is, you know, is Congress, is government at large, is it thinking much about these problems?

18:17And where does regulation fit into this or things other than explicit regulation? You know, how are we considering that there is a worry at some level of this being a major issue for many families going forward. Kind of where are we at on that? Where does Congress at on that? Could you share any of your thinking or your perceptions about that? First of all, there are initiatives in Congress. Mark Warner and a fellow Republican, I'm not sure who, maybe Tom Tillis, in the Senate. I'm the co-lead with, I can't remember who, in the house on a bill for a commission on the future of the economy, specifically based on this one question.

19:04What do we do if AI displaces massive amounts of federal workers? And I don't think that the tendency is towards regulation. You hear people say, well, we should just say you can't use this AI technology to eliminate this job. That's probably not even plausible. Instead, what are the investments we have to make to make sure that people still have, first of all, a means to live? And then second of all, and not unimportantly, something meaningful to do with their days. The great optimist, and I am a major A optimist, can foresee a world with extraordinary abundance. Nick Bostrom's latest book on an AI utopia is worth reading.

19:47that basically if economics is a science of the allocation of scarce resources, we have a lot of things all of a sudden that may not be scarce. First of all, just look at clothing. This is as unscarce as it's ever been. You go back on two centuries and everyone wore the same set of clothes year-round. With all the energy, the fusion plants that are being built in America right now by Helion, by Commonwealth Fusion, the 44 companies that are racing to be the first. You know, in our, while we're still young, energy can be abundant, ubiquitous and low price. So what's going to be scarce? And that's where that scarcity is, is where humanity will probably go.

20:31Obviously, you know, it could be teachers, it could be care workers, it could be everything much more bespoke than it is right now. But then again, it may be that there's only a subset of the American people that fit into those new high-touch human relationship type jobs. And then what do you do? As you know, some of the AI czars talk about UBI, which works well in Alaska, but in general, I think most people don't want to be paid to do nothing. An interesting conversation with Jeff Hinton last week who suggested, If there really is that much abundance, let's just start with universal health care, that free universal health care.

21:14That takes one worry off of most people's plates and still leaves lots of room for them to work and be productive in other ways. It would. I think I get the sense along this topic. Certainly, I know a lot of our audience does software development and other other AI tangential jobs. And I think, you know, when Opus came out from Anthropic late last year at the 4.5 level, I guess, in late November, and Cloud Code was out at the same time through last year and gaining steam, is I know that there was a perception that we were kind of experiencing and living. And that the 2025 way of writing software, which was very human centric, may have may have AI assistance, you know, through various agents that were there.

22:06But writing software in 2026 has been a different experience. I think most people have have been recognizing that that kind of that almost a pair part, you know, a pair programming paradigm with your AI model. And I think aside from the technical aspect of that, I think there was a really strong psychological impact of this thing that we have been worried about or that we've talked about for some time has actually arrived. And we're having to change our behaviors and change how we approach our own careers to accommodate. And that's obviously only one white collar job out of many, many that can be affected and stuff.

22:46But I know there was in software development circles, there has been quite a lot of conversation around that in terms of upskilling. I think and ironically, I think that that's an area where people can upskill fairly easily. If they were willing to get into software development, they probably can upskill well. What about when we when we talk about jobs where where people may struggle a little bit for various reasons with upskilling? Maybe it's the level of education they currently have or whatever, and they need to step up into that. Is there any thoughts around, you know, we talked about universal basic income and stuff like that, but just the idea of being able to change something that you've been set in, in your ways for a long time, and get to a new reality as people are adjusting to this rapid AI innovation that's occurring.

23:40Any thoughts around that? And I say that as someone who jumped into a PhD program yourself, you're upskilling your own skills. Any guidance, any suggestions that people might take? Well, a couple of threads, Chris. One is I think that as a society, we will be much richer. I really do believe in the abundance. I think one of the challenges we have, and I don't have an easy solution, is what we're already seeing as the concentration of resources in some segments. You know, the rich are getting very, very rich. You know, all the billionaires. And then we have a lot of people left behind. I don't want to project in this podcast how we redistribute the income.

24:26Any income redistribution comes with enormous social problems. We can't leave two-thirds of the people behind in this. Everyone has to be able to share in the abundance that's created by artificial intelligence. Then beyond that, the fact that the scarce resource, again, is going to be services rather than things. We're all going to have enough things. Is that we don't have to be burdened down by roots and location. If you're in Johnstown, Ohio, and the steel plant closes or the automobile plant closes, it's tough to move. You live there all your life. You own your home there. Your family is there.

25:09You can't just pick up and say, okay, I'm moving to Charlotte to get a new job, which is why the left-behind places have suffered so much. When you don't have to move because things are much more relational and even information-based, it may well be that we can see growth in non-urban America, suburban, small town, middle-sized town America, based on this, which I think would be a very good thing. In fact, even in Virginia, we're finding most of our growth in population is happening in rural Virginia, and that's being made possible by electronic communication, by all the communication systems that are out there.

25:46A lot of this post-COVID. Look, 20 % of Americans work from home now, Chris. That's very different. Yeah, case in point for me, certainly, most of the time. So, you know, it's a very optimistic and I like hearing that there's so much doom and gloom around this topic. I really appreciate you kind of sharing kind of what might, what hopefully will be path forward on that. With some of the other concerns that people have beyond just the jobs arena, if you will, and kind of going into things like misuse of AI, and I guess that there will be people on both sides of this equation, but we mentioned surveillance in passing a little while ago.

26:33Could you talk a little bit about kind of, you know, where, you know, what you're thinking is around, you know, we already have, we've had mass surveillance for a number of years in different capacities, you know, going all the way back to the Snowden revelations where many Americans became aware of different levels of surveillance that maybe they hadn't before. At this point, as we're looking at AI enhanced surveillance and how that relates to civil liberties and how that relates to law enforcement and other tangential topics, it touches on so many things. Can you share a little bit of your thinking around that and what you're concerned about and what you're not concerned about maybe on AI enhancement of surveillance and the step up of what's possible?

27:21Yeah, it's interesting, Chris. You open the whole door of abuse or just downsides. Surveillance is clearly one of them. We already know because of various things happening in Congress, they set us all up with delete me accounts. And you look at the number of data brokers out there that have enormous amounts of information about you and about me. and the whole notion that all this information, every time you accept cookies or anything else, you're creating an ever greater profile of who you are that can be purchased by many, many other people. Noticing that we are private people now is going to be more and more of a fiction.

27:57And I don't think that's good for our citizenship, for our own security. I know my wife hates it. And how we reverse it is not clear. I keep searching the internet for people doing interesting things. Tim Bernier-Scott, the guy that created the World Wide Web, has a really interesting project he's working on trying to create. He calls it a pod, which would have all of Chris Benson's information in it. And you'd have to have permission to access it or even maybe pay to get into it. And we begin to monetize our own personal information that everybody else is so used to using. When you look at people like Meta, where the whole business model is getting information about you and me and then selling.

28:36Yeah. But that's one. Just thinking of other abuses, we already see how very sophisticated the fraudsters are right now. The amount of trillies of dollars that old people especially lose to the people who scam them. And AI makes those scams ever more sophisticated, ever harder to determine. I seem to get one or two evites a day from good friends that I make sure never to open because I notice that I'm on a PCC. But I know that once I open it, I have once again opened my system to somebody who wants to try to get into it. Finally, well, not finally, but we talk a lot about what Anthropic's debate with the Pentagon, with Pete Hegseth, over the use of AI in autonomous weapon systems.