"China Could Win a War in 72 Hours" • Sean Gourley (PrimerAI)

Physicist and Rhodes Scholar, Sean Gourley, founded Primer, a start-up building mission-critical AI capabilities for the United States and its closest allies. He joins Niki in the studio to discuss Silicon Valley’s ambivalence about building and selling defense capabilities necessary to protect democracies. He explains how artificial intelligence is a “third offset” – following nuclear weapons and precision munitions – that could end a war in just 72 hours. And China is hyper-focused on overtaking the United States in this arms race.

“AI is a technology so powerful that the country that wields it will quickly defeat any opponent who does not.” -Sean Gourley

• Learn more about Primer and AI in Warfare
• Follow Sean on Twitter
• Follow Niki on Twitter

• Learn More at www.techedup.com
• Check out video on YouTube
• Follow Niki on LinkedIn

Transcript

Copy Transcript Copied

Intro

[music plays] 

Niki:  I’m Niki Christoff, and welcome to Tech’ed Up.

In this episode, I’m joined in the studio by physicist and founder, Sean Gourley. His start-up is building AI military tools for the Five Eyes alliance. That’s the U.S., U.K., Canada, Australia, and New Zealand.

We talk about Silicon Valley’s hemming, hawing, and general avoidance of building or selling defense capabilities necessary to protect our democracies.

If you care about a free press, freedom of speech, freedom of assembly, the freedom to practice your religion, and the right to elect your own government officials (as shitty as the choices on the ballot may be), then this episode is for you. 

Transcript

Niki: Today in the studio. I have Sean Gourley. Thank you for coming in [Sean: and thank you for having me. Yeah] So we crossed paths last week. You're in Washington, DC. You were testifying at an artificial intelligence hearing on national security, an issue near and dear to my heart. So thanks for popping by while you're here.

Sean: No, no, it's, it's, it's easy, and it's wonderful to get a chance to sit down and talk.

Niki:  So, I briefly want to talk about your bio. So you have a Ph.D. in Physics, you're a Rhodes Sholar studying the mathematics of war. You're currently the founder of your second startup. Tell me a little bit about Primer.

Sean: Yeah, look, so Primer is, is a company focused on bringing artificial intelligence to mission-critical defense applications. And for us, we've really specialized in a lot of unstructured data. So whether that's text, audio, images, or video, bringing that in and providing structure using artificial intelligence to identify what's going on inside of those pieces. And this becomes really critical for applications, ranging from ISR all the way through to information operations.

Niki: What's ISR?

Sean: Intelligence surveillance reconnaissance. And so this is particularly important if you take something like Ukraine at the moment where you're getting a huge fusion of information from both classified sources,  satellite imagery, but you're also getting on-the-ground recordings posted on Telegram that are being overlaid with unclassified Russian radio comms.

And so if you're an analyst in the, in, in that environment, trying to bring all that information together is, is just a nightmare. And so what we do with our tools and software is provide a lot of structure. So the analyst can get a clearer picture of what's unfolding.

Niki: So they're taking in huge amounts of data, and you guys are…computing it? Sorry. I work in tech. It's unbelievable that I'm, I don't even know what that means!  But you're turning it into something more usable for them. 

Sean: Yeah. So it's, it's being able to recognize enemy tanks. It's being able to automatically transcribe, translate, and pull out key locations from Russian radio, intercepted radio comms. All of that stuff, which humans are currently doing manually, machines can now be trained to perform.

And I think the important thing here is, is that, y’know, humans are able to kind of sit down and take these machines and start to use them to understand better, y’know, what's unfolding. 

Niki: I feel like the listeners need to know that you are not Australian.

Sean: They do!  Although they, they would be forgiven for thinking that I was Australian because my, my accent, I think, has moved to be firmly somewhere in Melbourne. 

[both laugh] 

But it's, it's originally from New Zealand. Yeah. 

Niki: So…New Zealand! But my, my hunch is that if you're working for U.S., D.O.D, or Intel, you can't work for other, even allied, countries. Is that true?

Sean: No, look-

Niki: [interrupts] Oh, I'm wrong! 

Sean: I would characterize the, the Five Eyes alliance as one of the strongest alliances that, that we have.

And it's got a, a very long history and I think a very deep set of intelligence sharing capabilities. And what I think is particularly interesting in this is of course, the Five Eyes have data sharing. I think we're now moving to a place where they increasingly have model-sharing. And,  y’know, for me, I think this is really, really important as we think about our allies and, and the intelligence environment is, y’know, being able to kind of train a model in Australia, and have it contribute too in America. 

Niki: Got it!  So I, I guess I just assumed you'd have to sort of pick a country, but it sounds like they're letting you cooperate as a startup or as a private enterprise [Sean: yeah] with anyone who is an ally in that group of countries. 

Sean: Yeah. So look, we've, we've got a, a federal entity here. That's a U.S. one that holds our employees that have U.S. clearances. We've got an entity in the U.K that, that holds clearances for those that are in the U.K. We've got operations in Australia, and we've also got operations in Singapore as well. 

So y’know, where we would look at this, and I think it's, it's, it's relatively straightforward its like, y’ know, if, if you're, y’know, on the U.S., y’know, side of the, of the ledger,  y’know, that's certainly a, a set of customers that we're very happy to work with.

Niki: Okay. So let's talk the stakes of all of this. So anybody who listens to this podcast knows that I'm… I'm just constantly talking, harping about the People's Republic of China. Not the Chinese people- the government. And what I, what you testified on last week, one of the things you mentioned is that we're in this really high-stakes moment when it comes specifically to artificial intelligence and the relationship between China and the U.S. and our allies.

Sean:  For me, it, it frames us up as what I, y’know, what we call a third offset. And for those that aren't sort of, kind of, deep in sort of military theory, there's this concept in, in defense of, of offsets, whereby you create a technology that is such an offset or such a jump or a phase transition from what was previously available. That if you have this technology, it, effectively,  renders conflict useless because you would just win, y’know, in, in very short order.

So the first, y’know, offset was nuclear weapons. The second offset was precision munitions and stealth weaponry, and the third offset now is artificial intelligence. And so, y’know, if you look at this, nuclear weapons ended the second world war with precision-guided munitions and stealth weaponry.

The U.S. was able to defeat Iraq in 72 hours. And, y’know, I'd posit that artificial intelligence will have at least as big an impact on conflict as those two previous offsets. Now, what this means is if you don't win the AI arms race, you are liable to be beaten in 72 hours. Right?  Which is something to sort of internalize as the stakes of what we're actually playing for.

Niki: So I've worked in tech for 15, 16 years. And, I tell me if you think that this is right, but I have a theory, which is when I started working at Google, we would have conversations about; well, we pulled out of China while I was there over censorship concerns snd we were constantly being hacked by our joint venture partners. And it was just, like, untenable.

But I believe that the company really perceived ourselves to be an American company. And one of the things that I've noticed is I believe that some of the big players have shifted to thinking of themselves first and foremost as a multinational company because a lot of them are refusing to sell - or having a lot of handwringing - about selling to the U.S. Department of Defense or specifically to our Intel agencies.

Sean: There's this notion in the Valley that, that you can just ignore defense, right? We'll just keep the Silicon Valley bubble around us. And we'll just, y’know, assume that the world doesn't, y’know, y’know, want defense. And I think the reality is if you're the superpower of the world as America is, at some point, somebody's gonna wanna come and take that from you, and you have to have defense. If you want to defend the things that we, we know, love, and hold very dear in this country. And so I think, y’know, as you look at that, defense is obviously tied very, very closely to technology.

In fact, you, I would argue that technology is gonna, y’know, drive the, the vector of defense more than anything else in the coming decade. So, technology has to engage in this. I just think a lot of people in Silicon Valley haven't really made that, that jump. And so as a result of this,  I would say a lot of technology companies have, have just, y’know, put their head in the sand a little bit and said, “We're just gonna not do defense. And I hope someone else does for us. And that's great, but we're just gonna, y’know, go on our way and sell ads.” 

And, y’know, I, I think you can see, practically speaking, why this might be the case, right? Because you have huge kind of footprints into y’know, different regions of the world, they become customers, and you don't, y’know, want to kind of pick sides.

Niki: There was a, a real willingness at these companies in, in Silicon Valley to work with the department of defense and NSA, and then two things happened. One, Edward Snowden and those revelations. 

So I think that was one major knock against the Intel community that was problematic and kind of fractured that cooperative relationship. And then the second was the election of Donald Trump. And so I'd be sitting in meetings, and it wasn't just employees, although employees do have a bottom-up pressure on executives not to sell to the U.S. military. But you would have senior executives saying, “I'm uncomfortable selling to the Department of Defense or to the Intel communities, under this President or just in general.”  Right? And I'm like, “Okay, great. So,”  and I would say this out loud, “Download Duolingo. Start working on your Mandarin.” [Sean: Yeah] Like, I think that's the actual outcome.  If you are gonna be sad about democracy falling apart, you have to start supporting democratic nations militarily. That's my opinion! 

Sean: No, I mean, look, the, the dynamic of this is that, if you like what you've got, you're gonna have to defend it. Right. And I think the U.S. has enjoyed for, for most of our sort of adult lives being the sole superpower in the world, and you could sort of brush defense across to the corner. Right. But if you go back to kind of, when we were, y'know, really kids who you grew up watching Red Dawn. [Niki: Yes!] And, and you're like, “Oh, this is like, y'know, that that's a story of high school kids picking up arms to fight, y'know, the Soviets, literally on, on a battle for the U.S. soil.”

We haven't grown up with any of these things. Right? And so, it's so far from our mind, we're like, “Well, how bad could it be?” [Niki: I think]  And you and you, and you, you look at this, and you're like, “Oh, no, like it's, it's a different world!” 

And I think the dynamic here is that we're sitting here debating, y'know, the, the ethical kind of stance of engaging with defense. And if we keep debating that, y'know, we, we, we have that privilege, but pretty soon China's gonna make the decision about how we use AI for defense, for us. 

Niki: That's right! And to your point, if it becomes, which I think this is your argument, and I honestly hadn't thought about it [chuckles] until I heard you say it last week, you could end a conflict in 72 hours. If they have the, the better technology by, y'know, leaps and bounds, and people…people, in my opinion, especially young people, really lack imagination for how quickly states can fail and how quickly the world [ interrupts self] I mean, as a student of history, the world order can change so quickly. 

That's alarmist for this podcast. Sometimes we talk lower key things. What do we do next? What are you guys doing? What are the steps? What are the like solutions people should be focused on? 

Sean: This is really, as you say, it's about political control structures and whether or not the, the constitution we have here and the democratic ideals and the values of freedom that we've got here, whether or not we value them enough to fight for them. Right? And, and really fight for them against an ideology that we see represented by the, y'know, the CCP, which is diametrically opposed. Right? And these are very different, y'know, ideologies and, and structures. And so that that's one. Right? Is this worth fighting for, and, and, y'know, if you say no, then I, I, I think we're probably done on the conversation. Right?

And the second thing, well, how do you fight for this? Right? And you say, well, look, I think what we're seeing from China is, is investing massively in artificial intelligence in order to leapfrog the U.S.’s military dominance. Right? So this is what you get with an offset. Right?  It’s that if you invest in it and win it, you can render all the previous technologies obsolete.

What does that mean practically? It means you can take an F-35 fighter, and you say, “Well, I'm not gonna create a better, faster, y'know, machine, but I'm gonna create a swarm of 10,000 unmanned drones that are gonna swarm this thing, destroy all of the targeting abilities on that and gonna take this thing down by being sucked into the engines.” 

Niki:  And we have tech companies refusing to sell drones. We have American companies refusing to sell drones to the U.S. government.

Sean: And, and, and what you end up with, is, is this, I, I think, and this we, we will go through is, is a huge win, right? A huge win for the Chinese defense and government complex is that they are very, very closely tied and enmeshed, right?

Every single major technology company in China has a member of the CCP on their board. They are if they're told to jump into defense, they, there's no question about it. 

Niki: They say,” How high?”

Sean: Exactly!  We are sitting here and saying, y'know, “No, it’s quite, I'm not sure. I don't know. Maybe we don't wanna do defense. Like, y'know, I don't think that's what we do.”  And then you're like, well, if technology is not gonna engage with the defense space in the U.S., then China is gonna take that higher ground. And that again, now, as we, we get on that line, that takes us to that place. So as we look through this, I think we need to really kind of embrace the idea that we're in an arms race at the moment, an artificial intelligence arms race.

And we need to also kind of acknowledge that China is, is, is hell-bent on winning this, right? And they have been hell-bent on winning this for many years now. And we are slowly getting ahead around it now. We've got some advantages as we're going into this, but it's gonna be the speed at which this thing moves that is actually gonna determine the winner.

And that window is closing very, very quickly for us. So look, we've gotta get the Hill here to kind of like put the right policies in place. We've gotta get the right adoption from defense and intelligence agencies and importantly, technology companies out in Silicon Valley need to get their head wrapped around the fact that this is a very, very urgent thing as something that they need and is indeed their duty to participate in.

Niki: You had talked about something last week that I thought was really interesting was not just the offensive part of, of the accelerated adoption of artificial intelligence in China but also how they're using what they know to jam up our surveillance. So you had talked about tanks being painted, like in a different color so that it sort of, [Sean: Yeah] I'm paraphrasing something you said, but I'll let you say it. 

Sean: Yeah, So, so look in the world of artificial intelligence, you, there's these, these foundational models. Right? Which are just trained on huge volumes of images or huge volumes of language. And they, they have incredible capabilities. Right? They can outcompete humans in computer vision tasks. They can out-compete us in translation tasks. It, it's, it's really incredible what they can do. But if you’re an adversary attacking them, they have these weird sort of sensitivities where you can exploit them. 

And so, in computer vision, one of the known computer vision attacks is you paint anything yellow and black stripes and it, it basically, every computer vision algorithm recognizes it as a school bus, which is a weird sort of, um, attack space because the system was never really trained with an adversarial environment in mind. And so, it's been interesting, y'know, watching some of the experiments that China's been doing where you look at sort of weird painting schemes. And you look at that, and I think the only conclusion you can draw is it's designed to avoid computer vision algorithms. And so, why is that important? Well, if you've got drones up there with computer vision for targeting, whether that's for surveillance or kinetic, if you can avoid detection on that, you have an advantage.

So what you've got, y'know, from China is,  an, an acceptance that the AI arms race on, a, an acknowledgment of some of the sensitivities of the current state of the art technologies and actually a kind of, a set of behaviors that are designed to kind of test out what that's unfolding. And we are sitting here kind of, y'know, with first with Silicon Valley, but also,  y'know, here in D.C. kind of blind to the fact that it's an AI arms race actually unfolding as we speak. And if we are not even acknowledging that it's happening, we're sure as hell not in a position, right now, where we can be confident that we're gonna win.

Niki: This is a moment in time that's really demoralizing when you think about our democracy. 

And I often feel like, so I'm not on Facebook. I'm not on a lot of social media because I feel like it's just a Russian psy ops campaign that I'm checking out. I mean, they've totally radicalized my mom, like, effectively, [chuckles] using Facebook. And so, part of me also feels like we're in this moment where we're fighting with each other, which is a huge distraction from paying attention to the larger geopolitical space.

And so to your point, in addition to D.C. kind of getting on top of it, which I think some people are…House Intelligence and certainly the agencies understand, and we were last week in a room full of people who get it. But also, just somehow, we need to get the message out specifically to the tech companies that they do have this obligation.

Yes, there are things we are frustrated with and are frankly really sad happening in this country right now, politically. But a lot of that is, there's gas poured on that fire by our adversaries. So like, if we can just step back from it and focus on, as you say, defense, it at least gives us a shot at defending democracy because I think we're really on the brink. And I think it's less about polarization in the United States than, like, a distraction among us.

Sean: Yeah. And look, the first AI arms battle is gonna be an information space, right? It's gonna be in the world of information operations, narrative conflict, and narrative war. And if you can convince the opponent that there's no need to fight. I, if, if you can convince the opponent that the fight is internal, then you gain a huge advantage. Right? 

And, y'know, we can see that already unfolding in, in Ukraine. Right? If Zelensky didn't, y'know, manage to convince the Ukraine population to come back, pick up arms, and fight, Ukraine wouldn't be here today. And so, the importance of winning the hearts and minds and the narrative is incredibly,  as incredibly, uh, y’know, urgent. And, and that's where we're gonna see artificial intelligence come on. 

Now, you talk about some of these psy ops campaigns and things unfolding, and that, y'know, if you don't have the superior kinetic military, you're gonna engage in information operations because y'know, it, it allows you, y'know, effectively to render whatever military advantage you've got, y'know, render it useless. Because if you're not willing to use it, you're not willing to engage, it doesn't really matter what the military dynamics you have. It's the same with technology. If you can render the technology industry disinterested, ambivalent at best but perhaps actively against engaging with technology, you gain a huge advantage.

And so if I'm running a campaign here as, as a Chinese government, I'm gonna focus massively on information operations, and I'm gonna hammer this kind of internal division and I'm gonna hammer this kind of sense of like,  “No. Technology companies don't do defense.”

There's plenty of evidence that China has become a lot more sophisticated and its information operations. And if you look at, y'know, the Intelligence Authorization Act that just came through specifically calling out China's activities in Central and South America and the Caribbean. Right? And saying, “Look, from an intelligence perspective, we need to go and get a solid understanding of their actions as they're unfolding.”

They’re also calling out, particularly where China is helping and abetting Russia in its information operations campaigns. So this is front and center for the House Intelligence committee. You can see that from the statements that they've made. We have technology now to go out and track and monitor and understand. All of that. And we've got technologies also to come out and help protect us from these things. 

And so, we have to realize that we're in the middle of an information war. It is the first artificial intelligence war that we're engaged with, and we have some advantages, but if we don't move quickly, we cede that.

And, y'know, China has become incredibly sophisticated. We saw the ham-fisted attempts they had in Hong Kong where they repurposed a bunch of spam bots, and the messaging and the dynamics were, were, were,  very, very primitive. And now we've seen the move, y'know, all the way through to kind of quite sophisticated campaigns unfolding in, in South America.

Niki: Right. And if we've got, when you say kinetic, you're talking ships, planes, tanks, like, actual arms, [Sean: anything physical] guns- anything physical. 

Sean: [interrupts] Kinetic is anything physical; information is obviously anything that's in the information space, which includes cyber. [Niki: right] Y’know, but it also importantly includes information warfare, right?

Niki: Right!  Which is…I think is, as you say, that's the frontier we're in. And, and by the way, I'm for kinetics. I mean, who, whatever I have sort of, I'm a hawk on this stuff, but I'm for having the kinetic buildup. But if you're not willing to use that and worse, if you've demoralized the entire population or, y'know, people, everybody knows that TikTok is surveilling them. They don't care. They use it anyway. Young people use it anyway.

Look, if you, if you control the information that a population is exposed to, you control a large part of that population. And look, there's a reason that Facebook didn't get a foothold in China. Right? And it wasn't because it wasn't the biggest or the most technologically advanced social network at the time. The CCP looked at that and said, “We don't want an American company from Silicon Valley, run by Mark Zuckerberg, controlling the information landscape of our population. We work very hard to control [chuckling] the information population, the information landscape. We, we have a lot of people that do that.” 

I think both China and Russia if you're an, an autocratic regime, you are gonna control information. And you're gonna get very good at controlling information. [Niki: right!] 

Sean: The U.S. is in a bit of a, a bit of a disadvantage in this, right? We don't, and we shouldn't control information because that's not how a democratic system works. It works on the debate and engagement of ideas and, the, the nature of freedom of discussion through that and the freedom of belief. Right? Very, very important ideals. 

However, it becomes very amenable to an information attack because you have to let ideas go free; you can't control 'em. And, and I think we have as a democracy, and indeed all democracies, have a big, big asymmetric disadvantage versus the autocratic regimes that can control information.

So, we need monitoring situations to kind of understand when this foreign attack’s unfolding. We also need to be very, very clear about defensive measures. If we're seeing bot networks, we've gotta be able to identify them quickly and have the social networks take them down. And we also need to be kind of very aware not just to the U.S. population but of what's unfolding in our neighbors and our allies around the world.

And y'know, one of the big games that China is playing at the moment is, is very much an information operation to take down the perception that Taiwan could be an, a, a standalone country. 

Niki: Right.

Sean:  Right! [Niki: Right!]  And, and this is, this is something that's being engaged globally to get the global kind of buy-in to the,  to the idea that, y'know,  that China has had, y'know, y'know, that Taiwan is even a country full-stop.

Niki: And this leads to, again, having sat inside the senior conference rooms of these companies. When we left China at Google, the conversation was about not having a bastardized version of Google in China. What's changed from that year to now is the market. You need that market, right? Financially. Apple is gonna do things to cooperate with the CCP that they won't do with the U.S. government. [chuckles] 

And so I don't wanna, like, I actually think some companies do a really nice job of saying like we're gonna work for the U.S. government. AWS is one. I think Microsoft does a good job with it. So there's certain companies participating, but you also have this economic reality where they're looking at their bottom line, and they say, like, “We need access to this population to make money.”  And they've, it's such a huge market. Everything from movies to sports.

If we wanna defend and protect having the things that we have in a democracy, like, we've gotta just get really serious about defense.

Sean: Yeah. Look, y'know, you can't fault a company for, for having customers in China. And that, that's absolutely fine. But you can't, at the same time, come back and say, “Y'know, it's a big technology company. We won't work with the U.S. government.”  Right? [Niki: Right!]  Like that's not a tenable position as a U.S. company. Right?

So I, I think that dynamic is, is one that has to go now, of course, that that's not as simple as saying, that because you say, “Well, oh, if, if you sell to the U.S. government, can you really sell under China? Or is they gonna create tensions,”  and so on. I, I think we have to address some of that complexity, but I don't think it's, it's tenable as, as a U.S. technology company to say, “We will sell to China, but we won't sell defense to the U.S. government.” Right. That's,  that's not gonna work, at least it's, it's not gonna work long term.

I, I think the second bit here is, y'know, we, we have work to do to bring people along and, and, and say, “Look, there is an AI arms race unfolding today. And we either win this and keep what we like here, or we lose this. And all the debates about what y'know, technology should do with defense will be made for us by China.”

And that's, that's a, a, y’know, I think a very simple, very clear message, but it's one that hasn't landed. And, y’know, we have a short window to get that to land. I applaud, y'know, the House Intelligence Committee. I think what they've put forward in their Intelligence Authorization Act last week was incredibly forward-looking and, and acknowledged the urgency of the fight that we're in. I think it's now up to the technology companies to take that baton and run with it. 

Niki: I agree! And I think the last thing I'll end on is, so my, my specialty is messaging, and I think you're right. The storytelling isn't landing. And I think it feels abstract to people like,  “An AI arms race?” It's like when people talk about the national, the deficit, it's like, “I know it's bad,” but I really think that we need concrete examples of what the future's gonna look like if we lose this. 

I think the work your company's doing is really important. I'm delighted that you're working with the U.S. government and the other allied nations, Five Eyes nations, and keep it up! Keep ringing the alarm bell. I mean, the people who listen to this podcast work in D.C., a lot of people get it, but I also think we're looking for leaders who get it and can communicate it in a really crisp way so that the people of these democracies understand what's at stake.

Sean: Yeah. Look, the, the, the message that kind of go- and I think for those that do get it right, is to go and have that conversation with people. Right? And it'll feel alarmist. It'll feel, [chuckles] it'll feel like, y'know, like, “Hey, can't we all just kind of get on and keep swiping through our social media” and y'know, all the rest of it.  

We need to win this and if we don't win this, we will be living in a world where China is the global superpower, and that comes with it a whole set of ideological differences. And if you like having the ability to have freedom of beliefs and freedoms of discussions. And freedom to democratically choose governments. You don't want China as the global superpower. [Niki: Yeah!] And, and that's what we're fighting for. 

Niki: Absolutely! All right. Well, I dedicate this episode to my cousin, Madeline, who continues to use TikTok, and on every vacation, I'm like, “You need to get that off your phone. You're being manipulated!”  And she's like, “Ugh! Seriously lighten up.” But it is serious. And I'm so glad that you are working on this. 

Thank you for taking the time to come out and talk about it. 

Sean:  My pleasure. Thank you for having me.

Outro:

[music plays] 

Niki: Thank you for listening this week. The stakes of building up an artificial intelligence arsenal for democratic nations and defensive tools to protect against information warfare could not be higher. If we fail to act, [chuckles] then nothing else we talk about on this podcast matters. 

My call to action is this: if you work in public sector sales or at a tech company that’s reluctant to get involved in defending our freedoms, consider forwarding this episode. 

As always, thanks for listening.