From Cold War to Code Wars: Unpacking America’s Bold AI Strategy

Welcome back to Data Driven, the podcast that dives into the collision of data technology and occasionally geopolitics with the finesse of a caffeinated quantum computer. In this episode, Frank Lavine is joined once again by Christopher Nuland, AI technical marketing maestro at Red Hat, for a no holds barred breakdown of America's freshly minted AI action plan.

From Cold War vibes and AI sovereignty to the CHiPs Act, GPU geopolitics, and existential musings on large language models, this episode has more hot takes than a GPU server farm in July. Plus, we debate whether Europe can still flex its AI muscle without surrendering to Silicon Valley, and whether AI models will ever truly think or just continue to be unreasonably effective spreadsheet goblins. So buckle up, data disciples. This one's dense, dynamic, and dangerously nerdy. Let's get into it.

All right, that bouncy little pop number. That is a fun fact. AI generated can only mean one thing. It's time for a new edition of Frank's World TV Live or an episode of Data Driven, depending on where and how you're listening, slash watching. You can catch me at the following URLs. Franksworld.com data driven tv and impact quantum.com speaking of impact quantum.com my co host and I have launched another book and it's basically Quantum Curious, the Gateway to the Next Computing Revolution.

And what that is is we basically took the third, the first 13 some odd episodes of the season and distilled it down into a little book. It's 2.99 on Amazon, but if you join our mailing list, it's completely free. So go to Impact Quantum or scan the QR code to find out more. All right, now that I've gotten the commercialism ism out of the way, I'd like to welcome back our guest, Christopher Nuland, who is a peer of mine on the same team, an AI

technical marketing manager at Red Hat. How's it going? Good, good. Glad to be back. I think we ended the last talk on kind of a cliffhanger, and then I think some recent news has really built on top of some of that previous conversation. So I'm happy to be here talking about some big things that are going on in the the area of AI right now. Absolutely. So over the weekend, the Trump administration dropped the thought on the Amer AI's America's AI Action

Plan. I think somebody likes alliterations. But so, and you and I were chatting about this over Slack and, and, and, and you had some thoughts on this. So, like, what. And you had some interesting ideas around it, and some surprises are in the bell. So let's. Let's get into it. Yeah. So I think overall, this is really needed. We've seen a couple things like this come out for some other countries globally. The EU has

this. I'd say the one from the US is more of a set of guidelines, or the one from EU is actually some laws that they're trying to pass that have some similar tone to this one. You know, we're seeing things out of the uk, out of Singapore, other, you know, other nations that really are trying to get an idea of what is their strategy around AI sovereignty. And this, to me, is a document more about AI sovereignty than anything else.

It's really about how. How does the US go into the next phase of really almost like a new industrial revolution around AI and this document's really outlining the plan. I think overall, I thought it was pretty well thought out, and we'll go through and kind of pick apart some of the key areas. But I think overall the. The key tone here was about AI sovereignty, so specifically within the US and how we're going to be managing that. And overall,

I thought it was. It was good. I know. You know, when we were talking on Slack, we were talking about how there's definitely a lot there about China as well. Right. In a weird way, I. I felt like this document was a bit of a. A declaration of. Of war in a way, because in a document, it outlines that they really consider this now like a Cold War with.

With China around AI and what I thought was so fascinating is I kept going back to this analogy of, like, the Cold War arms race of Russia and how we need to do certain things around AI because we basically need to mimic what the United States was able to achieve during the Cold War. And I think that sat with me because I think, you know, even last time I was on here, we were talking about how we. We basically. No, there's a Cold War kind of thing going on, but it was. It was different seeing

it. You know, you and I were talking earlier, like, on the letterhead. It was. Yeah, it's different. You know, like, there's. There's things that are. Obviously, you can see with your own eyes, but it's quite a different thing when something appears on official White House letterhead. Right. Or even, you know, government letterhead. I think that's. It's an interesting, interesting shift. And this whole idea of a Cold War between the US and China and AI is Not a new concept.

Right. I. There's a really good book and I'm gonna see if I can share this tab real quick. This is an excellent book. It's an excellent audiobook too. There you go. Oh, there we are. Sorry everyone, my dogs are barking. But this book came out in 2018 and a lot of what he predicted has come to pass. And it seems to me like the authors of this document have also, if not read the book or listened to the book, have at least seen the

Cliff Notes version of it. Right. Like this, if you really think about it, there's really only two major players right now in the AI space and we're going to alienate a lot of people in the eu. Right. But saying that, right, it's really largely US and China, right? Yes. And not to say that the EU is not in the games, clearly, Mistral. And apparently there's a rumor, I don't know if you heard this rumor that, that Apple is considering buying Mistral. I have heard some of those.

So again, I don't own any of Apple stock or whatever, so I'm not. Or. But I think it's an interesting con, Interesting idea if they were to do that because clearly that would, I wonder how that would shift kind of the balance of, if not power, perceived power. Right. Because Apple obviously is a stalwart of Silicon Valley and if, you know, if Europe's major, you know, every time you talk about the EU falling behind, they always say, what

about Mistral? Right. So if Mistral ends up getting purchased by, you know, Apple, that would be, that would. I think there'd be a lot of drama about that. Yeah, I think so too. I think it, it really just shows there's a line in the sand between the two major superpowers here, between China and the United States. My speculation there is that there might be something official there, but that the EU might step in and just say no. Right. To that.

Simply given what we're here talking about, AI sovereignty. And what does AI sovereignty look like? I, I don't think the, the French necessarily want to give up that and I don't think the EU wants to give that up from an open source standpoint. We're still seeing a lot of thought leadership coming out of the eu, even though there's not enough, what I would say, enterprise momentum there. Right. There's still a lot of research institutes there. There's still

a lot of, even some smaller form companies like even like Hugging Face and Mistral for example, are, you know, EU based that have made A big impact and are very open source heavy and but at the end of the day they're just still very small when you consider these behemoth organizations like Microsoft, Amazon, Nvidia, Apple, all the fang corporations which have,

that can really throw their weight around. And we've seen a lot of, a lot of startups like OpenAI that has like climbed up now into the upper epsilon and that's being really driven by American industry. So and that's just something the EU can't prop up as as much. But I still think they, they're a major player. They may not be necessarily one to one with China and America, but if there was a second tier right under that, it would be the eu. Yeah, I

can see that. I also think it's too early to count them out of the race. Right. Like, yeah, you know we're, this is the start of the marathon. Right. So they're clearly, clearly there are two front runners. But I don't, I don't, I wouldn't count them entirely out just yet. Right. And I didn't know Hugging Face was a European company. I thought they were based out of New York. But that must be their American. I believe you may be right. Let me show Hugging Face.

I know. I think the founders are European. You are correct. Founders are European, but they are based out of America. And that just goes to show right there. Yo. That the gravity of just American enterprise. That you can shell out a lot of money to get, to get talent. Right. And yeah, this was the thing that a European tech founder said. Right. So you know, all the Europeans don't hate on me, but there was a lot of founders that they'll end up moving to Dubai. Right. Bootstrap and

then move to Silicon Valley, you know, at some point. Right. Like, so I think the, the European Union as a whole has to address some systemic shortcomings when it comes to a venture capital and startup pipeline. Right. And I hope I, I think that they'll get it figured out. I just don't think that they're going to get it figured out this year. They might get it figured out by the end of the decade because I think that,

you know, just a little bit of back of the napkin math, right. You, you know, it's, it's, you can see that growing the tax base is good for everyone and this is one way to do that. And if you have your brain drain, which we'll get into that, that term, you know, either going to Dubai, you know, Silicon Valley or New York, it's not Good. Right. Because you're, you're basically, you're educating them in country. Right. And a lot of these countries have, you know, cheaper, you know, free tuition.

Yeah. So you're paying for the talent, you're training up the talent, you're paying for talent. And then when time comes in to cash in on the return on that investment, if you want to look at it that way, throughout of country. Right. So what are you going to do? I think that it's in the EU's best interest to fix that problem. And like you said, like sovereign AI or is a big deal. And sovereign AI is different than sovereign than data sovereignty, right?

Yep. It's the. And I don't think people really kind of gotten their head around that. So I know what my definition is of that, you know, is the idea that. And it's even called out in this action plan report. Right. Where it's like, you know, AI with American values. Right. Yeah. And like you said, I'm pretty sure the French want to have, you know, AI with French values, and the

Germans probably want to have, you know, with German values. Right. So I think even, even painting the entire continent, even though everything's is kind of done through the European Union, I don't think that's. That might be at their detriment. Right. And. And the German market is also pretty sizable too. Right. It's something like 80 million people. Right. And the German language market, I think, adds another 20

million to that. Right. So, you know, I only say that because one of the, one of the points that people made for taking German in high school was it was 100 million ish, you know, number of people speak the language. So it's not. And, and I would say that, you know, particularly when we're dealing with language models. Right. It's in the name. Right. So language and culture, although not exactly the same, are very much tightly linked. And that was something we talked about last

time. We got sidetracked, but that's what I do here. That's fine. What struck out of you, the report? I think one of the things you mentioned was, well, go ahead, I'll let you go. Sure. The thing that I was most surprised about was that pillar, one of the document on page six and a couple of other areas was really focused on the workforce and this concept of like, securing the AI workforce, making sure to have necessary people

in play. And then it got into like almost this Cold war kind of mentality of like, how do we make sure that we can trust the people that we have and it was, it was just surprising to me because I thought it was going to be more about the regulation of AI models, which it does move into eventually. Right. And supply chain security.

But the, the concept of, of workforce and the fact that it was the first pillar was intriguing to me. It got me thinking about kind of what, what is the US Administration thinking about right now? And they're, I think they're really thinking about making sure they lock down the people. And for good reasons and, and probably bad reasons too. You know, good reason is, you know, how do we entice the best experts

to stay here in America? How do we entice the workforce to continue to move into the area of AI through education? But then there also seemed to be almost like a Cold War vibe there. I don't know if you watched the movie Oppenheimer, but it kind of reminded me in that movie where the people working on the Manhattan Project were like their personal lives were, were under view quite a bit.

And it, it kind of reminded me of that. Like is, is in, in a year or two are all the AI experts going to, you know, have the NSA and the FBI like keeping track of them and what they're doing? And it doesn't explicitly say that, but the tone kind of led me to think, oh, wow, they're, they, they're really interested in what these people are doing. It's not just about the technology, but the people making the technology.

And that was very intriguing to me. I could see that being a thing, especially if there's an actual honest to God, you know, old school knockdown, drag out shooting war with any country. I could see that being, I wouldn't say nationalized, but you'll have to get some kind of. Even now, like if you work in the

nuclear industry, you need a queue clearance. You need a lot of things being. You need a lot of invasive, not procedures, but definitely a lot of invasive paperwork and investigations that, But I, and I, and I do see, I didn't read the whole thing yet, but I did, I did, I did feed it through Notebook lm. I did listen to that. I did do some skimming of it. And that was one of the things was like it seems like they're laying the groundwork for that in case things get sideways.

Also part of that is the, the securing the supply chain from the silicon on up. Yeah. Which is a smart thing because the chips are made in very limited geolocations. Right. So one, one major international incident, a shooting war. Right. No matter who wins. You know, there's Going to be an island where most of the stuff is made. Yeah. That's going to be reduced to the rubble. Right. Now, whose flag gets planted on top of that rubble, you know, remains to be seen. But you know, you, you know,

so much for the chip manufacturers there. Yeah. And also too, you can't rule out natural disasters. Right. You know, 2000, was it 2005 there was. Or 2004 there was the massive tsunami that, you know, did major swath of damage. It's not impossible to imagine even just a natural disaster, bad typhoon either. Earthquake with Japan, wake Japan. I mean it, it could, it's not impossible to imagine like more than one way for it

to rain on everybody's parade. And if you think supply chain issues with GPUs are rough today. Yeah, I mean that's just, that would be a big thing. But what really stood out to me, and obviously I'm biased because my wife is a federal employee, was

talking about training federal employees to use AI. Right. And there was, even in there, even in there there was a accelerating adoption here, but basically mandating employee access for federal employees and training on these LLMs, which is interesting because I can speak from not first hand experience, but certainly, you know, secondhand experience. Right. Federal employees do not feel loved and appreciated, let alone have access to any kind of training or

anything like that. So I thought that was interesting, that was interesting in there because it's been a rough go for feds the last six, seven months. Oh yeah. Everything's been very negative. And this is like one of the, maybe the first, it's first positive, you. Know, and I was, I was telling you in the virtual green room is that, you know, the agency my wife works at, like they're not hiring new people, but they're creating a new organization that people will be doubling down on their duties,

which presumably they'll get access to the training. And she, she may or may not be involved with that yet. We don't know. But, but it's interesting to kind of see that, see that happening. But yeah. What else have you, what else took. It's, you know, so I think the most important thing that was in there and I, I actually figured the whole document would be about this is

around supply chain security. So if, if people aren't aware when we're talking about supply chain, typically we talk about supply chain, it's more in industry terms of, you know, how does something get made, the nuts and bolts, where does the raw materials come from? That term was never used really in technology until recently. And. Probably the pandemic is when most people first heard the term supply chain.

It was, it was the Solar Winds hack. I think that also really, yes, put it in perspective too. So those who aren't aware, there's a company called SolarWinds, they were very predominantly used in the government. I think they still are. But there was a hack where instead of hacking their software directly, they hacked the supply chain. They injected bad code early on into the supply chain and that slowly propagated out to these

different government agencies. And the scary part is that the very thing that was meant to monitor these type of situations was the thing that had gotten infiltrated. So it took a while for anyone to even know. And it was

massive. It impacted the government and impacted enterprise. And that is where I think NIST and a couple other agencies made the decision, okay, we're going to come up with a requirement of what supply chain looks like within these types of software development process. And really gets into, okay, all the way from how do we think up an idea for code, how do we submit that code into a repository, how do we compile it, how do

we scan it, how do we distribute it? And that's when we talk about supply chain, secure supply chain. That's in the context of what we mean. And that relates directly to AI as well, because it's all data pipelines. And for AI specifically, it's about where does that data come from, where was it sourced, when was it added into our model? How can we prove that the model that we built over here is the model that's running over here?

So if the government has an officially blessed model, how do I know the model that's running within my defense contract firm is that model? And that gets all into this supply chain. And I was happy to see that some of it wasn't as technically laid out as I wanted it to be. The document really just says we're relying on NIST and some other government agencies to come up with a plan. So this wasn't really the plan, it's more the actual call to action for

the plan. But it was good to see that there. It was important for it to be there. I was happy that that was highlighted. And I think in terms of security, it's the most underappreciated one right now. Everyone's really focused on model guardrails. And what we talked about last time with the AI 2027 report or AI

breaking out of, of its shell. I think the most important thing right now is actually more of the supply chain security where you know, don't let people inject bad data into models that are making critical decisions for the government, for finance or healthcare. That's where our focus needs to be first. I think having that secure supply chain is ultimately what's going to lead to, to preventing

the AI 2027 report as well. Where it'll prevent a breakout or if there was a breakout, it's going to reduce the blast radius of that type of situation. Now that makes a lot of sense and it's interesting because there's not just the traditional nation states that could be involved here. Right. There's also or bad actors in the normal sense. But also the AI itself could become a threat too. Right. Like, and the report doesn't isn't technical in detail,

but I don't think that's who the audience was really for. Yeah but that's interesting because you know, I don't know like from a game theory point of view, right. Like you have the traditional, the usual suspects, right. The countries, terrorist groups, criminal gangs, blah blah, blah. Right. The usual kind of players. But AI also has the potential to become yet another player in the game of that. That's. I certainly didn't see that in the report and it didn't cross my mind until you kind

of bridged last stream and this stream content. I was like, oh wow, this is multi dimensional. This is like 5 dgs or something like that. Yes.

I would say for a first effort, it's actually a fairly reasonably well written document. For those that don't, for folks that don't know, I used to accompany our lobbyists in, in when I was at Microsoft talking about technology issues and things like that and you know, I was the technical resource for that.

And as I was telling you, virtual green room. A lot of these elected officials, regardless of, you know, whether you agree with their party affiliation or whatnot, they're not the most technical I would say of the one ones I've interacted with which maybe, maybe 60, 70, some of them are names you've heard of, some of them as you've never heard of. I would say less than 10% are technical in any sense. Yeah, right. And there were only two that I would say like would feel

at home having a technical conversation. I wonder how many of the policymakers even understand the term AI sovereignty. So and this is interesting, I'd love your opinion. Yeah, I think how many technical people would understand. Well, that's what I mean. Yeah, go ahead. This is where I've been having some conversations even within our own organization that we work for. There's a lot of differing opinion on what AI

sovereignty is. A lot of people who keep talking to me about AI sovereignty, I realize they're more talking about clients, cloud sovereignty, they're talking about how do I secure the compute, all of my compute within my borders and can guarantee that everything is within those borders. Which makes sense. I mean we work for Red Hat, we work for a, you know, basically a cloud Linux based company. Right. But when we talk about AI sovereignty, at least me personally, it, it's an accumulation of a

few core areas. It gets back to the data sovereignty, a little bit of that cloud sovereignty. But it's really about do my, I have control over my AI models, I know where the data came from. And I loved what you said earlier. It's about the culture of the model and I think ultimately the AI sovereignty is about the culture of the model and then making sure that you're containing your AI to the borders of the United States. So you're keeping all the secrets here, you're keeping the talent

that are driving it. But ultimately you're right, it's about that culture and making sure that your model has the best representation of your culture. And it's kind of a scary thing to think about. It's an interesting topic, but it also gets into a lot of geopolitical challenges I think we're having are now surfacing to the top because of things like AI, you know, it's, it's interesting. Well, it's like 100

and I think I was actually a colleague ours, Robbie, shout out to Robbie and gotta have him on the stream one of these days. You know, he was talking about kind of AI sovereignty like, you know, what is, you know, you can use an American model, right. From data, right. And then tell it to behave British. He used better words, right? You know, the spellings and the grammar and things like that. But whose values are in there, right. When you, when you ask it questions, Right,

yeah. And, and that gets to an interesting thing, right? So like you know, I, I, my grand half, my grandparents were not born in the U.S. they're immigrants, right. So like, but so, so when I went to one of the countries my ancestry comes through is Ireland, right. So but the Ireland that a lot of my older family members came from really doesn't exist anymore, right. It's not the rural kind of poverty stricken country that it was, right. 100, 110 years ago. 100 years ago.

So it was very awkward because when I was, when I was in Ireland as an American. Even though it felt familiar, it was also felt very foreign. Right. Because it was, you know, it was, you know, and if you think of me as a, you know, large language model, so to speak. Right. I grew up in New York. Right. I'm very Americanized. So when I go there and it felt familiar. Right. Like the, the pubs and the restaurants felt like places my older family,

it felt like grandma's house and that sort of thing. But it clearly was not. And it was clearly also not the same place that they left. Yeah. That you would hear in family stories and things like that. You know, so it's interesting because also I think values and country and all of that are inherently political and I think that's why you're seeing this. Right. It is inherently geopolitical is inherently all of these things.

So technologists who are not used to, we're not used to this, these types of conversations now suddenly we're pulled into this and God forbid if there's a, you know, an actual kind of 20th century global war style thing happening. Yeah. Or would happen, you know, it's only going to get worse from here. So I do find it, I do find it interesting how technologists are now suddenly pulled into this. Right. There's a famous, you know, you know, Jensen Wong made it an emergency visit. That was,

that was a big deal. Right, That's. And actually that a lot of that kind of stuff is called out in this report. You should, you should go into details about that. So Jensen Wong, apparently, I don't know what was the driver of it, but I suspect he was. The administration was trying to block all exports of GPUs to a particular country. Yeah. So the rumor was that week that all, all chip, the global chip manufacturing outside the US Other than key allies, would just be completely

stopped. And obviously for, for some areas, like China, it would. They would just end export for pretty much all chips. So that was not just the ones that are blockaded right now, but you know, really, even some of the basic ones. Well, remember that Ford's assembly line was shut down because there was a shortage due to the pandemic. Nothing else. Of chips to put in the cars for the assembly line. Yes. And it cost them x. Millions of tens of millions of dollars a day or

something like that. Right. So not trivial. Right. So like this could have, this could have been really bad. So go ahead. I'm sorry. No, no, no, I was just saying. I was just adding some flavor because it, it was officially announced by the White House that they were evaluating this and then the, the word on the street was, you know, the, the uncut secret was that the, the U.S. was going to declare this at one of the summits that they were going to, that they were just cut the chip manufacturing

altogether. And. Yeah, and then Jensen made an emergency visit to the White House and which I guess you, if, if you run the, the most profitable company in America right now, it helps. Well, that's his most profitable, the most valued. Right. This valuation is like 4 trillion last I heard. So. Which is crazy. But yeah, I mean he, it was a, it was a very unplanned visit where he just went and knocked on. The door and oh, to be a fly in that wall and that. In the wall. I know,

I know, right. But I mean props to him. Immediately after that, right, we start hearing of the oh, we're gonna back this down. We're gonna, we're, we're gonna consider still shipping the, whatever the, the chip is in China. That's kind of a. Right, an A100 knockoff. We did see impacts in that conversation, but I think it's important because it builds into the, this document because the document clearly outlines semiconductor supply chain outlining the reliance on the, of Taiwan.

What I loved about it is that there was a section here, one second, I am pulling it up. It was a little tongue in cheek where they're talking about reviving the US chip manufacturing under CHIPS act, but stripped of ideological constraints. And we won't go into the politics of that here. But I thought that was pretty funny because the CHIPS act was obviously a big deal. It was a big deal for me because when it was announced I was still.

I'm based in Boston now, but I'm from Northern Indiana around the area where Purdue University is close to Chicago. And we were actually called out on the CHIPS Act. They were going to build a semiconductor facility there in our area in conjunction with Purdue University. But then when, when Trump was elected, he was trying to claw back anything he could from the CHIPS Act. Right. I'm happy to see that the CHIPS act

is back on the table. I think it's still going to be extremely political like we have seen with these types of acts, but is needed. I AM hoping that $6 billion isn't just going to go to intel because I think the innovation there is starting to die off. I'm hoping that we see more focus towards some innovative areas in chip

manufacturing here and also ultimately which is called out. We want to bring over a lot of the Taiwan based technologies and my understanding is that there's just a bunch of explosives within those facilities there in Taiwan and they're ready to just blow them up at a moment's notice and move ship to the U.S. wow. So I know they're building some of those facilities. I think Arizona was one of them. I think Texas is another

where they're starting to mimic some of that chip production. And basically right now the United States is trading military equipment for chip technology. It's crazy. Fascinating is. But it's absolutely fascinating from a geopolitical standpoint that the currency right now for, for Taiwan is, Is chips. Yeah. And so. But I think that's a big driver. It's one of the things that was called out. It was called out in pillar two of the document, which calls called

Build American AI Infrastructure. And I think. Yeah, you have the outline there where they call out specifically the semiconductor leadership and then also securing data centers. I thought this was interesting. They're going to start having federal guidelines on data center security and will also incorporate military and intelligence usage for those facilities. This is what I

was telling about. This is just reminding me of when I, when I watched Oppenheimer and learning about the Manhattan Project and there were military guards in front of the physics research facilities in Chicago University and in, in New York and in Los Alamos. It's, it just seems very, very similar where it's like we are now going to attach military guards to guard our public sector AI infrastructure. And yeah, I mean one of the interesting things and I think this really kind of, if you take

a step back, right. Like why, why is. For many nations, why is domestic auto production important? Right. Because when it hits the fan, you're not. You make tanks, right? You make tanks, you make airplanes. Like all these things are important for nation states. Right. So automobile production is a proxy for tank production. Right. Civilian airline airplane production is a proxy for, you know, this I would add now probably chip manufacturing. Right. And possibly, possibly AI model creation.

Yeah. You look at what's happening around the world where there are conflicts. Right. Drones are playing a huge part of this. Yep. Right. Whether they're autonomous or not, we will never really know until the history books are written and even then. But the whole idea of, you know, drone and AI based warfare. Right. You know, one of the videos coming out of you, the Ukraine conflict was the Russian airplanes were covered in tires. I don't know if you saw this. No. So, so one of the.

The thinking is that they had, I guess, old tires covering some of the parts of the airplane. Strategically, the best guess that Everyone has. And I've heard this from multiple sources saying it's kind of true and kind of not. So. I don't know, take it for what you will, is that that was done to confuse computer vision systems. Yeah. And

then there's this other thing. I don't know if you heard of patch attacks, which is basically like this idea of. I'll see. I pulled up some. Some graphics of this, but basically. Open image in new tab. Basically, it's the idea that you can alter a structure, like a stop sign, in ways that the AI model will see something different and alter what the AI model is

determining it sees. And apparently you're seeing a lot of this, if you look at footage from, you know, Ukraine area, is that you're seeing, like, you know, tanks both sides with. With stickers on them that look like really warped QR codes or like bizarre things like this. Yeah. And it's basically to thwart these types of systems. So. That's fascinating. It's interesting, isn't it? And this gets back to. I don't know if it was on the stream or another conversation we have.

We're building these systems, these LLMs with, you know, hundreds of billions of parameters. Right. If not a trillion or two, we really don't know how they work. No, we think we know. And you and I were talking about this the other day, actually, it wasn't on a stream or anything. It was kind of like. I think that LLMs that we have now are unreasonably effective. Right. They're able to. And I'll put air quotes here for anyone listening reason. Right. They shouldn't be able to

based on. I mean, all I see is just a vector database with lots of relationships between words. Yeah, Right. They're capable of doing things that if. I wouldn't think they would be yet. They are. Yeah. So there's a lot of research dollars going into figuring this out right now. Like, why is that? Like, what. Is there something inherently powerful about language? Probably. Yeah. Right. That and, you know, language is kind of like the assembly language of the mind, if you think

about it. Right. So I can encode my thoughts into something, whether it's a written word, whether it's, you know, vocalizations, and then have that come out. It's basically a. Like a codec for human thought. And maybe there's some kind of. I don't want to say intelligence, but some kind of something we don't quite yet grasp. Yeah. About the nature of language and relationships between words that

automatically you get for free. Once you kind of train these models up, I think that's fascinating, and I'm glad there's a lot of research dollars to that. It is. But, you know, clearly the human nervous system, our visual system, our cortex, whatever it's called, you know, we know that that is a moth sticker on a stop sign. Yeah. What is different about how the AI learned that makes us vulnerable, this type of attack. That's fascinating.

And it doesn't see things like, there's a lot of research papers that'll show you, basically, what does the model see? And you see it and it's just absolute nonsense to us. Right. It's. It doesn't see what we see. It's like it doesn't relate the. You know, maybe it's not correlating the red and the white backgrounds, but instead it's correlating the position of the text or the fact that it's four capital letters positioned over an octagon or something like it.

It. The. The way figures these things out is. Different than how we think we do it. Yeah. It's actually seems obtuse, but. It's obtuse. But it does it billion times faster than us. So when it's obtuse, it gets to something faster than we do because it just can do it a billion times over. And that's where the secret sauce really is. But how things relate back to each other, obviously, we have these, these vectors that like, you know, build relations between

words. But how it can then take it and reason is still not quite understood. Right. Right now it's just not understood. No, it. And it's. No, it's not understood. And that's kind of what keeps me up at night, is we don't really. We're putting these. Again, you know, full disclosure, we both work for an enterprise software company with very

large customers. You know, we're deploying these LLMs in places they're not exactly making the life and death decisions right now, but it's not that hard to imagine that they would. Right. Yep. And I don't know, I think that's. That's just a huge security vulnerability. We don't know how these things work and also understand that it doesn't make sense to hold off deploying these things once we fully understand it. Right. That doesn't. That's

not going to fly either. But I think we should, as a society, like, really think about, you know, what are the consequences here. Right. Think of what the Jeff Goldblum character, Jurassic park. Right. You know, talked about chaos Theory and all that. Right. Like it's, you know, you know, the unintended consequences of this. We should, to your point, have AI in a box, like, and make sure it's really hard for that to get out. But again, like, you know, these things are. Doing.

These things don't think like us and they may think in more circuitous and obtuse ways that don't make sense to us, but again, they do it a billion times faster. So, you know, it could end up being far more clever than we are. Absolutely. I remember when I was learning Comp sci and one of the things I think was assembly language class actually was multiplication on silicon is typically done. Not by now. What was it? Yeah, it was.

I don't know if it's still true, but back in the day it was true that multiplication is actually done through repeated addition. Yeah. It was actually more efficient to do it that way. Right. Again, I think that's a great summary of like, that's kind of the slow way. But if you're operating billions of times faster, slow way isn't so bad. Or a slow way doesn't mean anything. And I think you and I were having this conversation that, you know, if you think about the power requirements of these

AI systems. Yeah. Versus the power requirements of the human brain, something like 25 watts. Right. And if you think about the intelligence of birds, like crows in particular, Right. They, they, they have the intelligence of a six, seven year old supposedly. You know, not only do they have to do it power efficiently, but they also do it weight efficiently

too. Right. So the infrastructure, you know, that a crow has to think about, think about, but, or evolution or whatever, has to put it in a lightweight body. Like I don't fly. I'm obviously not, I'm not a petite individual, so I don't have to worry about that. But like, if you're a bird, you know, you have to fly, so you have to think about that. And yet they're able to manifest some kind of intelligence with very modest hardware. I mean, their brains are not that big. I think the size of a

walnut. I don't know. Like, this is totally off topic, but. No, it's, it's related though. And on the report they were talking about power, power requirements, power requirements and grid security. Right. And it was called out as. And you think about just the sheer massive amount of power that these AI models take. It's. It's insane. I think, I think there was a point where one third of all power is being used for like bitcoin. Mining. At one point

that went down, and now we've. We've replaced that with AI and, you know, that's. It keeps going up and up and to the point that, you know, it's possible that half of all the power being used here soon is just going to be for AI. And I can see that there's no evolutionary pressure like there was on biology. No, no, you can just throw more power at it. So in. In this case, with. With the LLM technology, you can just throw more chips. Right. And, you know, make

them. You know, this actually hits home because I'm between. So Ashburn or Loudoun County, Virginia, which, if you've ever flown in and out of Dulles Airport, you've been there, is Data center alley. So U.S. east is there. U.S. east 1, 2 for all the major providers. Right. Plus a lot of private ones, too. I live between there and Three Mile Island. Oh, wow.

Yeah. So one of the big controversies in the state of Maryland is that they want to put in what they call the Maryland Power Piedmont Reliability Project or something like nprp. They're basically going to put in high power lines between Pennsylvania to Virginia, which is a political football because there's a lot of land that's going to have to be eminent domained. Yeah, Right. There's obviously environmental factors, but also this is the thing that is really kind of insult to injury. Right.

None of the power that's going to go over those lines is going to be consumed here. It's all basically exporting power from Pennsylvania through to Virginia, which is not. Not a good look if you're. Because the people who vote Maryland people in are Maryland residents. So there's this whole. It's a very big controversy right now. And it's interesting because what used to be a very isolated hobby of technology is now embroiled in geopolitics, local

politics. It's just kind of like I kind of miss the good old days before lawyers got involved. Yep. But sorry, but no, I mean, that's a good point. That's, you know, you think about the power requirements, right. You know, for these things, you're gonna have to build new power centers. You're gonna have to do this. Right. And then that, you know, what's. What's your power source going to be? Solar is awesome. Solar can't solve everything, Right. So

what's it going to be? Is it going to be wind? You know, is it going to be, you know, coal? Is it going to be natural gas? Is going to be oil? Right. There's going to be a whole. It's all fun and games until people are paying way more for their electric bill each month than they, than they're used to. Yeah, it's going to be, it's going to change things very quickly, especially if it starts impacting people's monthly power bills.

Right. I think right now we haven't seen it too much just because we've been able to keep up with demand. But once that demand starts really affecting prices, I think we'll also see AI being a conversation point in that way where it's going to start. And then I think, you know, even with the, the 2027 plan that we're talking about, 2027 plan, you know, we were talking about things like, you know, universal basic income and stuff. You know, if you, if AI starts

taking over everything. And that wasn't outlined in this document, which I'm not, I'm not surprised. But it's a, it's going to be a big conversation point. If AI does work the way that we think it's going to work, will we start seeing the AI take the jobs? And if they take the jobs. I think it was, actually, it was Bill Gates like 10 years ago was talking about UBI for, for AI, and at the time we just thought Bill was being crazy and like,

like a go back to your Gates foundation and. You know, go back and work on malaria. Yeah, yeah. But no, and even Elon Musk, I mean, and Elon Musk is definitely a polarizing figure, as is Bill Gates. But they're both polarizing in different directions. Yeah. They both agree on ubi. I have mixed feelings personally about ubi, and it's not because I'm a mean individual. It's just if you study the history of serfdom. Yep. I don't know. Looks a little too similar to me. Yeah. But that's just my take

on it. But you're right, like, and also too, like governments are getting involved. Because if you go to your local McDonald's, right. Or your Dunkin Donuts, right. And you think of how many people used to staff that in the past versus how many people staff that now. Yeah. Right. And assume, well, human nature is human nature. Right. If you used to take 10 people to run your average McDonald's, now they can get by. I don't know. If you go in there now,

there's like five, maybe four. Yeah, four or five. And that's generous. Right. If nothing else, the taxes on the wages have went from 10 employee taxes on 10 employees. Now they're taxing it on five. Yeah. Right. That, that's a big deal. It is, right. Because now you're taxing. Now granted they're not, you're not taxing them a lot because they're not making a lot of money, but still that's 50%. So if you're kind of like a, you know, a number cruncher and you're

looking at every McDonald's, right. When you have 100 McDonald's now, you're getting the tax revenue out of that one McDonald's, you know, or at least on the income of it. Right. The income of the individuals. The income tax on that is going to be way less now. Even now. Even before AGI. Before. Yeah, before that. Right. Because it's just automation. Right. And I personally would rather deal with a kiosk. Same here. Deal with the person. Yeah. Right.

Especially if you have like special orders. Right. Like, oh, you know, my kid doesn't want ketchup on his burger. Right. He doesn't want onions on his burger. Right. So you just have that as a favorite of the app and then just press go. I don't even have to touch the kiosk. Yeah, I think that that is going to be, that's not even an AI system,

Right. That's just good old fashioned automation. One of the big Silicon Valley AI gurus who, it was escapes me right now, but was talking about how the, the jobs that are going to be considered desirable are going to be completely flipped here soon. Where he was, he was saying the most desirable job might just be people in performing arts. Right. He's like, he, he's like, AI is

not going to replicate that anytime soon. He's like, yes, you may have movies being AI generated, but there's still something to be said about the performing arts. You know, obviously like plumbing and electrician work and construction work, you know, robotics might amplify that and make

it better, but there'll still be a human element. But you know, traditional white collar jobs as we know it, other than the people who, who manage that AI, I, I just feel like it's going to be completely turned upside down if AI does what we want it to do. That's a big if right now. It's a good tool. But the, the real if is we're gonna get to this more area of agentic and more AI is actually being able to do the full job of someone rather than just being a

tool that they use. And that's the if right now that we're, we're betting a lot on. The economy on where there's a lot of. A lot of bet from many financial institutions that the AI is going to be what is the next industrial revolution. I think that's still yet to be proven out. One just to go back to the UBI though, there's a book you might be familiar with at the Expanse. Yes, love those books. They cover this idea of. Of universal basic

income. And you know we basically have in that, that series like I think it's like 90, 95 of the world is just on Earth's population averse population is basically on universal basic income some sort. And you then have these, the 5% that actually just have jobs. Right. It's a big deal that they just have

a job and they're doing things and you know, they're politicians and people managing technology or defense and it's. It's fascinating. And I think if anyone's wanting to look at a little bit less of less rosy kind of outcome and one that I think is more accurate, I think it would be a combination of. Of the Expanse and then probably Ghost in

the Shell, the anime. Both of them show AI and technology not to the extent of like Terminator the Matrix where everything gets destroyed, but more of a like human progression just gets bogged down by this development. We end up in this more like technocratic kind of of realm where techno feudalism almost. Yeah, that's a great way of putting it. Techno techno feudalism. And what's interesting is if you look at kind of the expense. So I'm

a big fan of the Expanse. I've read. I haven't read all the books, but I've read a lot of them. I've seen the series, which is excellent by the way, on Amazon. I'm salty that they didn't. They stopped it at season six, but I can let that go. But what's interesting is that the people with gumption ended up leaving Earth and going to Mars. Yep. Or the asteroid belt. So what happens is 100 years after that now you have to kind of like these three factions. Right. Everyone looks down on Earth,

right. Because there's always like, particularly in the show, there's always these barbs where the politician says, you know, if. If you don't do this right, I'm going to put you on basic. Right. Like so basic becomes like a threat, which I think is interesting. And then there's also kind of the. The people who are more entrepreneurial end up going to Mars or the asteroid belt. And then that doesn't always work out well. So they have this. You

have this tension between these three different factions. And then throughout this course of the books, a third faction, kind of a fourth faction kind of enters the scene and kind of disrupts the power of the status quo. And that's kind of the main tension of the books is, you know, what happens after that. But highly recommend those books if you haven't seen them on the TV show. If you. The TV show is really well done. I think I would agree. From what I've seen of it, I haven't finished

it, but it's good. And I'm in. I'm in the same boat as you. I'm a couple books in. It's one of those series I kind of come back to every once in a while. But funny enough, it's a. It's a series I reference a lot. I think about it a lot because I was like, I think that's a really accurate depiction of what the future could look for us with. With the technology. It's pretty reasonable. And that's what's really nice about the show. Like, it's. It's not because there's also.

There's. Obviously, you mentioned the pessimistic views of the future. Right. There's. There's the Matrix, there's the Terminator, but there's also Star Trek, which is a little too. On the optimistic side. Yes. But there's not really. I think what's great about the Expanse, and I haven't. I haven't seen Ghost in the Shell anime in a long time. I did see clips of the Scarlet Johansson movie, but the

Expanse does a pretty good job of going down the middle. Like, there's going to be societal changes that will come for this we really can't imagine now. Right. Yes. You know, Earth is pretty much almost like a techno feudal state. Especially what's interesting in the Expanse is when they explore what life is like for the average human on Earth. It's kind of like it's either really good or not. Right. And Mars is also kind of an

interesting place too. There's a very different dynamic when you get that many type A driven people in one place. Sounds awesome at first, but then it's not really awesome. Yeah, necessarily. Right. But fun fact, the serve the PCs and the server names in my house are all derived from the show. Oh, cool. Yeah, yeah. So I'm talking to you now on Amun Ra. Cool. I don't know if you've gotten to that part of the. That's in the first book. Yeah, yeah. The Amun Ra Stealth class

ships. And the computer I just bought also has that same kind of, you know, gamer box game aesthetic. So that's Osiris. And I also have Behemoth, which is that machine back there. And. Or you've not gotten to the Behemoth yet. Okay, I won't spoil it for you though. Yeah. But. And Andy, my co host on the podcast, is also a big fan of the show. He has, he has the, the Doniger, which you probably heard of that one. Yes, yes. He's got Weeping Sonambulist, Weeping Somnambulist,

which. I had a machine with that name, but it's too hard to type out. You doing the ping on it? It's like, no, I don't know if you got into that one yet, because that's a couple books in. But. Yeah, yeah. And my, my, my. When I left Microsoft, my former Microsoft manager let me keep one of the laptops. So when it boots into Windows, it's the Tachi. And when I boot it into Linux, it's Rosson, which, you know, people have read book or seen the show.

Go get the joke. But. And it's funny, our manager, when we met in person, my machine was the Razorback. Right. Which I don't know if you got to that part yet, but I'll try not to be spoiler. He's like, so what are you with like an Arizona fan? I'm like, no, no, no, it's from a book. Nice. So, but. Another area, I think this is for another, another time. So when I come back. But I think it would be good to talk also about how are the AI tools right now? Like, are we seeing them replace

humans? I think the leap that we've made in the last six months is pretty substantial. Yeah. I think last year I would have said no. I think this year I'm saying yes. Like, we're seeing, we are now seeing the technology there to actually start replacing people. And it's not that, it's not that the main guy is going to be out like the tech lead, but I think it's going to be more the, the junior developer that's going to be in trouble because now the tech league can act like

a fleet of junior developers. And like I'm, I'm just programming a game right now and I'm so surprised how much I've been able to get done in the, in the time frame I've been working on it. It's amazing how quickly you can be. But wasn't there also a story, a Guy deleted his entire production database. Yeah. Because of. I don't know the details. I had my AI delete, actually go and start cherry picking things off of the main branch and start

deleting things. Oh, interesting. So I have a duplicate. I, I every day I fork my. I have a fork that I, I merge back into because I don't trust it and I don't tell the AI about the, the fork backup. Yeah, yeah. I think that says a lot though. Like you don't trust it. Like, you know, and it's not guardrails. It's not. Well, it's not guardrails in the sense that when people say guardrails and AI. Right. That's true. Yeah. It's a different. You're kind of. You're cya.

That's really what you're doing. It is, it is. Right, that's true. Whether you, whether you put your code back up in another repo in another branch or a USB drive, like you're really. CYA is really what you're doing. And I think that there's a lot of. We've been going for an hour, so. And I also have to. I gotta drop too, so. Yeah, I gotta drop two.

But, but it's been great. It's awesome. I think we continue more. But I definitely want to know more about the game thing you're doing because I sent you a bunch of stuff on Humble Bundle too. Yeah, yeah, which for game dev, so. But I have. My teenager needs a ride somewhere, so. Hey, thank you for having me. Hey, no problem, man. It's great. And be sure to check out our Red Hat AI YouTube channel where I think Chris has a video or two. Yeah. And I have a video or two as well. And

with that, we'll see you next time. And have a good one. And that's a wrap on this episode of Data

Driven, where we've dissected the Americas AI action plan with the precision of a data scientist on espresso and the paranoia of a Cold War analyst. Big thanks to Christopher Nuland for returning to the show and reminding us that AI sovereignty isn't just a buzzword. It's a geopolitical chess match played with silicon and source code. If you're not slightly more worried about data pipelines, chip supply chains, or which values your LLM

secretly harbors. Were you even listening? As always, you can find us on data driven TV, franksworld.com and wherever your algorithms recommend quality geek banter. Until next time, stay curious, stay Data Driven. And remember, if your AI starts talking about sovereignty. Maybe check the firewall.