How the Air Force Is Cracking Silicon Valley

Why Cheap Drones Redefine War

Speaker 2 0:00

You know, I think for a long time the US way of war was, you know, we want to build for a really big culminating event, and how do we get there? And that requires exquisite logistics and kind of this very well-controlled forward line of battle, and all those things go away when someone can sneak a drone and a truck into the middle of Russia and hold at risk, you know, some of their strategic assets.

Speaker 0:29

What the Ukraine war did was sort of demonstrate that out of necessity, you know, someone, you know, a country that that really needs to like defend themselves can sort of innovate their way out of like having this like requirement of being tied to exquisite war fighting systems.

Speaker 2 0:49

Even if you agree, maybe we shouldn't build such like long-term exquisite things. You may not agree with what should replace it. And so I think the entirety of the defense community is really saying, well, if not that, then what we have a real gift for you today.

Rajiv Parikh 1:11

Two amazing leaders in the Air Force are speaking with us about their experience in starting an Air Force studio that's run out of Stanford called the Department of Air Force Stanford AI Studio. And what they discuss is in their career, their collaboration that they've been working on for the last three years between academics, the military, and the entrepreneurial community. And what they highlight are many of the challenges that we face, especially in today's almost hyperkinetic environment, where in the era of Ukraine and Iran, where we are dealing with disruptive technology every single day. And they discuss how the exquisite nature of technology systems that build up our military today are being challenged by commoditized technology that's being deployed at scale, low-cost systems put together rapidly versus higher cost, much higher capability systems, and how one disrupts the other. And they're talking about their efforts to bridge all that. So it's a really powerful session about what they think. They're smart, they're geeky, they're thoughtful. This is something you have to listen to.

Meet The Air Force AI Studio

Rajiv Parikh 2:33

Hello and welcome to the Spark of Ages podcast. Today we're doing a special discussion on Defense Tech and the startup ecosystem that supports it and our national defense. Our guests today are the literal architects of autonomous defense, people who are focused on implementing the systems and hardware that are being created by the Defense Tech ecosystem. Two amazing guests, Jason Hansberger. Jason's been with us before, and we'll be glad to have him back again. Jason is a leader in technology and strategy, currently serving as the director of the Department of Air Force at Stanford AI studio and director of technology capability development for the United States Air Force. He also has served as a commander of the First Airlift Squadron. Jason holds a Master of Arts in Regional Studies with a Southeast Asia concentration from the Naval Postgraduate School and got his bachelor's degree in economics from the Air Force Academy. He is focused on applying state-of-the-art solutions in AI and autonomy to solve Air Force problems. We also have John Alora. John is an Air Force pilot and robotics expert currently serving as the deputy director for the Department of Air Force at Stanford AI Studio, their partners, and assistant dean of research for the United States Air Force Test Pilot School. He's also served as a D-52 aircraft commander and weapons and tactics officer. John holds a PhD in aeronautics and astronautics from Stanford University, a master's degree in aerospace engineering from MIT, and got his bachelor's degree in electrical engineering from the United States Air Force Academy. He is focused on developing physics-based machine learning techniques and leveraging AI to maximize human cognition in complex aerospace and defense environments. Jason and John, welcome to the Spark of Ages.

Speaker 4:20

Thanks, Rajiv.

Rajiv Parikh 4:21

Great to be here.

Speaker 4:22

Yeah, thanks for having us, Rajiv.

Rajiv Parikh 4:23

Uh Jason, you've been with us before. We talked before about maybe a year ago, a little over a year ago, and we really dug in on autonomous weaponry, and a lot has changed, obviously. Between them and now, you know, we have two active uh military situations or confrontations that are occurring, but this is actually very timely. We've also seen a massive explosion in uh defense tech investments. They've been coming for some time, and now you're really seeing them grow. I'm seeing numbers that are saying that we're in the you know from single-digit uh billions of dollars being invested from 2020 1.6 billion to uh something like just in the first five months of the year, 14.6 billion dollars. And if it's if you're talking about dual use, it's uh you know 49 billion dollars, depending on which thing you look at. So quite a bit of money going into defense, and so we'll get into all these.

The Cheap Kill Paradigm

Rajiv Parikh 5:14

So the first question is in Ukraine and now Iran and Lebanon, the nature of warfare has changed in that inexpensive, repurposed consumer drones can effectively destroy a hundred million dollar military assets. It's a shift from those who expected future wars to rely strictly on exquisite high-end technology. How is this cheap kill paradigm changing the type of startup founders the government is choosing to learn from and get behind?

Speaker 2 5:41

Maybe I'll I'll pick the word nature just a little bit. You know, just for military people, you know, we kind of think of nature as like why do people, why do organizations go to war? And I think the nature is is somewhat unchanged. It's, you know, you can use some definition from closets or whatever. It's an extension of politics by other means. And so I think people go to war or they go into competition in order to advance their own self-interest, often at the expense of some other competitor. And then while at the same time avoiding undesired escalation. That's kind of the nature of that's how I just thought that's how I define competition and really it's war it goes to war. But the character of war, I think, is changed a lot. And we see that you kind of may maybe most openly and lessons learned in in Ukraine are you know are the most proliferated. I think we're still learning a lot on Iran, but I think there's like you you had the right general sense, which is there's a democratization of the ability to hold at risk a competitor or an enemy is there's a democratization of the ability to hold at risk your competitors' means of of fighting. That makes a really big difference in how we think that you need to then define the problem, build a requirement, design a scheme of maneuver for that. All these things are being upended because of this change in character. And, you know, I think for a long time the US way of war was, you know, we want to build for a really big culminating event, and how do we get there? And that requires exquisite logistics and kind of this very well-controlled forward line of battle. And all those things go away when someone can sneak a drone and a truck into the middle of Russia and hold at risk, you know, some of their strategic assets, or overwhelm your you know, Patriot missile defense or whatever that a missile defense is with just the mass of some autonomous or extremely cheap platform. Because there's two, there's really important, I think, to keep in mind that autonomy does not equal drones, drones does not equal autonomy.

Rajiv Parikh 7:37

You're right. Those are two very different things, right? You can have a human-piloted drone, and then there's obviously, as you're mentioning, a big difference between autonomous, meaning they may and I think you even talked about it last year. There's ones that may fly together as a swarm, but still the human is controlling it, versus one in which you're programming and then it just does the whole thing itself and makes the final decision.

Speaker 2 7:59

Yeah, and so you know, all that to say, right, is I think the character of warfare has changed to the point that we cannot apply our legacy methods of thinking, method of organization, and methods of procurement and expect to come out winning on the other side. I have a saying that a lot of whether it's a military organization or civilian organization, anytime you have something that has a lot of investment in money and time and reputation, it is almost always the case that that thing will only be obsoleted by a competitor on a battle space. Like it's hard to self-obsolete. And I think my my sense right now is that there's a lot of resistance within the US and everybody everywhere else amongst great powers to self-obsolete the things that kind of come to define the way that they fight war. But if we don't do it, then it'll be done for us on a battlefield.

Speaker 8:49

Yeah. And if I could just add to that, I think that we as well as our adversaries, particularly the ones that are sort of nation states, have always designed for conflict between ourselves, right? We have designed for conflict with a pure adversary, they have designed conflict, you know, their strategy for us. And because of the lack of just sort of, you know, these conflicts becoming a hot war or an operational reality, we collectively have operated under the assumption that we need to field the most exquisite systems, that these exquisite systems are the things that are going to allow us to have the upper hand against the adversary. And I think that what the Ukraine war did was sort of demonstrate that out of necessity, you know, someone, you know, a country that that really needs to like defend themselves can sort of innovate their way out of like having this like requirement of being tied to exquisite war fighting systems. The fact that we can take proliferated cheap drones and impose challenges on an adversary, hold at risk their strategic assets, and that the economics actually favor the country that is sort of using these cheap assets is like, you know, these are lessons learned that we as the United States, as other countries are sort of, you know, observing the the war in Ukraine, are learning and are realizing that, okay, you know, maybe our assumption that building exquisite systems is is not true, that maybe we we need to start rethinking about how we develop systems as a whole to to fight the next

Funding Startups Without A Clear Signal

Speaker 10:27

fight.

Rajiv Parikh 10:27

Are you finding that now the like with this shift there's greater openness to funding earlier stage companies or partnering even more? Like I know Jason and and both of you are basically gone into this role so that you can enable greater innovation and faster movement. This wasn't just planned as of last year. This has been done for some time. But are you finding a greater openness because of what's happening?

Speaker 2 10:52

I'd say yes, but these exquisite things require really long lead time thought and process to bring to bear. And, you know, that kind of goes back to this like self-obsoleting. You know, I think there is resistance anywhere, right? When you say, let's field this new thing. There's people who are invested in an existing program who are going to wonder, what does that do to my program? And, you know, this isn't just a military thing, but it is a military thing too. That we do run into, even if you agree, maybe we shouldn't build such like long-term, exquisite things. You may not agree with what should replace it. And so I think the entirety of the defense community is really saying, well, if not that, then what? And I don't think there's a perfect agreement on what that should be. And I was thinking about this last night, actually. I haven't even tested this with John, so maybe it's terrible thought. You know, sometimes when when we try to go get something done, we don't really like to start with requirements because a requirement is already an abstraction from a problem. It's going to incorporate someone's view or many, a number of people's views, and maybe you know, they don't align with yours to get to the requirement. Because the requirement should start from a problem statement. And so I think, you know, where where John and I really like to go is all the way down to kind of a very first principle. What's the problem you're facing? What are the physics of that problem, and how do we address the physics of what you want to do? And then we say, is there a technical bottleneck there? Is there a technical solution? And does that technical solution lie within kind of what the AI studio is well placed to try to solve? And so for us, we can't solve every problem, we don't have the means to that, we don't have the nobody knows that, right? And so I think, yes, for John and I, it's it's a really big deal for us to find the people who we think can address the physics of the problem. And oftentimes that is a startup or a you know a newer firm just because they haven't built around the old kind of you know process. And so I would say that yeah, we do go find some really like innovative, exciting, you know, kind of leaders. But at this point, I would say that it's it's more of a network than it is a well-established process, just because we I don't think there is one yet for how to garner and marshal a trillion dollars towards a new character in warfare.

Rajiv Parikh 13:04

That's right. So you know, you're really hitting upon the this notion that there's like this embedded installed base, right? You if you go to build a weapon, it's a weapons program, it's 20 to 30 years. It's suppliers all around the country and around the world, and then you're saying, well, this new thing that's coming, it may obsolete what's already out there. How do you do it? So, John, you were saying something?

Speaker 13:23

Yeah, I was gonna add that your original question was, or part of the question was, is the department more willing to fund startups or sort of innovative companies? And is there more money available to these companies? And I think, you know, in general, the answer is yes. But to sort of piggyback on what Jason was saying is that we as a department are still really thinking about like, okay, we want these guys because they move fast and they're building innovative technology, but we don't really know like very concretely as a strategy, as a sort of principle that we can organize our organization around what we should make bets on. And so what you see is a lot of sprinkling of funding here and there. We still, I would say, lack sort of a cohesive understanding of what investments to make. And so that, you know, in some sense that's that's challenging because you know, I think that people, you know, within the space want clarity of purpose. Like, okay, you know, what is the clear demand signal that I should be building towards? But at the same time, you know, like it's kind of hard from the the government side because the the character of warfare is changing, you know, our only data point is the stuff that's happening in in Ukraine. But at the same time, we're trying to make investments, you know, for for a conflict in in the Pacific. And so it's like, you know, what lessons learned translate from Ukraine to the Pacific, what doesn't, and how do we sort of like how do how do we marry these two two things? So we have a lot of work to do within the department to really try to understand uh and have clarity of like what we want to make investments in. And so, you know, while there's still funding, it's still a challenge to really put out that that demand signal out to the rest of the industry.

Speaker 2 15:12

We have lots of short-term funding, things like you know, the SIBR, SITR, all these kinds of granting, you know, non-dilutive sources of funding. That's hard because those things don't come with a sustainment contract. And so making that transition can be really hard because you have to convince someone to in a program office or to create a program of record to sustain that thing long term can be difficult.

Speaker 15:36

And I think VCs are like investors are becoming a bit more savvy about how the government works. And so no longer is it the case that you know you as a startup can, you know, go and get a bunch of SIBRs. A SIBBR is uh small business innovation research grant.

unknown 15:53

Okay.

Rajiv Parikh 15:53

I was thought of them as SBIR, so maybe I just got the wrong time. Everyone's got their own acronym.

Speaker 15:58

Yeah, exactly. So I think that uh folks are becoming a bit more savvy about how the government does business. And so no longer is it the case that you know you get a bunch of SBIR sibbers, you know, you show this revenue to an investor and they're like, wow, you you know, you're fully tapped into the government. I think people now understand that your long-term sort of potential as a company working in you know defense is tied to your ability to you know contribute to a program of record. And that's really, really hard.

Rajiv Parikh 16:29

We're actually gonna get into some

Government Equity Stakes In Defense

Rajiv Parikh 16:30

of that. There's a radical shift happening in procurement with the administration announcing a goal to buy one million drones in the next two to three years. Additionally, the White House is negotiating deals that would include taking equity stakes in startups to boost domestic production. Now, to put it in some perspective, the Ukraine says they're gonna build seven million per year of these drones. So what's your reaction to the US government acting as a direct equity investor in startups? Does their innovation risk creating artificial winners before the technology has been fully proven in the field?

Speaker 2 16:58

I don't know enough about it to have a really expertise opinion. I I would say, you know, my first reaction would be I'd want to know exactly how taking an equity position accelerates the development or, you know, benefits getting the technology from a problem statement to a prototype to a product that can be used in the field by a warfighter. Like the biggest question is how do we respond to the needs of a warfighter? I would say that by taking the equity position within a startup would be a means to that end. But I don't I don't see the direct correlation. Um but I I don't know. I haven't I haven't contemplated it very deeply, John. I don't know. Maybe John has an opinion on that.

Speaker 17:34

No, I I haven't, I'm I'm also not an expert in in this area. I actually didn't realize that the government was planning to take equity stake on startups. So this is news to me.

Speaker 2 17:43

I mean, I think it would give them runway. If you're a startup, you're looking for investors who are well aligned with your vision. I would say probably a big part of it's going to depend on who's directing the equity investment. And if the person in directing the equity investment is very well plugged into the need of the warfighter, able to provide runway and take risk on, you know, take adopt some of the technology risk and do it in that that that equity position is a better way to do it than uh granting a loan or or or just giving a grant, just a straight-up grant, then I would be, I would probably be a proponent for it. I think the problem could be that they would maybe crowd out like a really savvy investor who now can't, you know, afford to get into the round because obviously there's a supply and demand issue. If there's a large demand of investors, then if I'm the if I'm the the founder, then you know, all of a sudden my valuation is gonna be higher than it otherwise would. And so that may price out somebody, you know, who's trying to get into it. So there may be a crowding out effect, like any government intervention. And that could that could be a negative, but all in all, I mean, I think the thing that I'm most concerned about, I'm really concerned about this. And I we talk about this actually a lot at the the studio in our in our meetings, is I'm very concerned about the like our ability to defend against these widely proliferated things and our ability to field in an economical way, mass that allows us to you know impose the same on our adversary. And I think for me, I'm pretty excited by any means necessary within like the confines of the law to get this capability to our warfighter. And so if that involves a novel technique that's that's seemingly contrary to kind of like traditional past practice, I'm totally fine with that because I am really nervous about this. I I think we take we can take, for example, you know, the the people who we've lost, you know, in the war with Iran as evidence of the need to continue to advance in any way we can.

Speaker 19:38

Yeah. I think the fundamental question is what is the value proposition for the for both the warfighter, right, and the startup? So, like if you know, if the value proposition is high for the warfighter in the sense that, you know, this this equity stake now opens them, you know, opens the startup up to go and you know build the relationships with the program office and there's a uh a more direct pathway for them to be able to contribute this technology and field it, then that's like, sure, that's great because that's also good for for business for the startup, right? If the idea is that, you know, this the government is going to take equity stake into companies because you know now we can kind of profit from the upside and profit from this like building ecosystem, then I'm not sure about that, right? Because it's not clear to me how that translates into real outcomes that that make us a more lethal force. So I guess it depends.

Speaker 2 20:34

Yeah, I mean, and maybe one other thing that concerns me would be as a government program, you're there's a there's a big incentive to spend the money that's been granted to you, right? And so what I would be nervous about too would be okay, if the dollars are are authorized to be spent this year, then the person is gonna face some kind of pressure to spend it or they're gonna lose it. And are they gonna, are they gonna allocate it in an efficient and effective, effective way? Or is it gonna create market distortions that take us away from you know the ability to really get after something that's that's creative? And something that Ukraine does that's super creative, that maybe if we're gonna do this, maybe a good idea is they do they almost run these like startup units, like the units at the front line of Ukraine don't run like a traditional military unit. They run like a startup, and they're they're granted money based on their effectiveness. Maybe if we're going to have a great evolution in the way in which we fund defense production, then maybe we really need a matching way in which we organize, train, and equip and then go fight as a as a military too.

Edge Compute Beats The Cloud

Rajiv Parikh 21:41

So now Silicon Valley's AI boom is built on massive, energy hungry cloud data centers, but the AI studio is hyper focused on compute at the edge, where connectivity is a luxury. So, how do you convince leading commercial researchers and tech companies to care about severely constrained edge environments when all the commercial Commercial money is in the cloud.

Speaker 2 22:01

I'll let the roboticist answer this. This is a near and dear to the heart of robotics.

Speaker 22:06

I'm not sure that I need to really work very hard to convince folks to do research or do work in this field, primarily because I think that the requirement for edge computing is already a part of like the self-driving ecosystem. People understand that, you know, we can't just offload compute to the cloud for inference, that a lot of the things that, you know, we need to do computation on the edge so that we can, you know, we can endow these systems with the ability to make decisions. And the boom in robotics is actually making this requirement sharper. These systems are becoming much more size weight and power constrained vis-a-vis your self-driving cars. And the fact that there's a huge boom in robotics, people are building startups all over the place addressing manufacturing, industrial applications. They just don't have the luxury of being able to offload that compute to the cloud. And then, you know, obviously from the government standpoint, it's great that I can sort of like, you know, build on top of this because, you know, our applications are very much edge, right? We need to do computation on the field. We need to do computation in space. I would say that it's actually not very difficult for us as an organization to try to convince folks to build edge applications.

Rajiv Parikh 23:23

So you're not you're not seeing like a talent constraint or funding constraint. Because I because I could imagine that, you know, like if you're talking about from the Air Force perspective, things that fly, so like drones or versus things that move on the ground. You know, these are very small devices. Talent is going towards AI chips where you can waste power essentially. And you don't have to waste power. It may not even be silicon, it may be other types of compute you want to utilize to enable that efficiency to occur so that you can fulfill the mission.

Speaker 23:54

Yeah, I think that I would say the mainstream right now is uh certainly there's a lot of talent being like focused on training these models, building at NVIDIA, you know, massive like GPUs that can train massive models. But uh, you know, from my perspective, I see a lot of talent, you know, also sort of going towards the at doing inference, doing uh inference on the edge, and going into robotics and adjacent industries. And my sense is that actually to like be able to differentiate both in academia and industry in terms of compute, you need to be like sort of working on really hard problems and edge and like inference on the edge. This is my sense of like where a lot of these companies are sort of you know differentiating from Nvidia, let's say. Yeah, I I think there's a lot of really great opportunities for these guys to build in that space. And we as the AI studio are are really looking to capitalize on it.

Speaker 2 24:49

I think the tech, you know, the the there's really a talent question of are there the world's greatest engineers excited by edge compute? The answer to that's yes. I think the harder question is are they excited to build for defense? Are they excited to build as part of a defense firm? Oftentimes the answer to that is no. And so part of you know, John and I's job is to convince them that this has a really compelling application in defense. They should think about how it applies to defense and then integrate that into you know the defense community. How do you persuade them? I think that it's twofold. It's one that there's a very compelling your technology will enable this, and this is a good thing. If that's an agreement, then you have somebody get excited. And then it's we have the ability to connect you to the range that you could test this technology in, the test infrastructure to, you know, so that when you go to the conference, you're not presenting a paper, you're presenting theory, you're presenting a paper based on like the actual embodiment of the technology that gets people pretty excited. And so I think that's that's you know, one of the one of the things that the studio so over the course of three years, like anybody else trying to break into defense, we are totally an organization that's attempted to break into defense and it's taken us three years to get into to become our own program of record. And over the course of that, like it's been a lot of like any other startup, chewing glass, near-death experiences, and making a compelling argument to a collaborator that might not include a bunch of money. Well, actually, not might definitely did not include a bunch of money.

Rajiv Parikh 26:17

Right. It's a lot of helping them connect the dots, persuading them. It's the network model that Silicon Valley thrives on, and you're helping to put all these pieces together to interconnect to the network.

Speaker 2 26:27

Yeah, and I've we found that that's been pretty compelling. That, hey, you know, yes, we can't, we may not be able to fund this thing for you, but we can tie it to a you know, have really strong traceability to an actual problem that's occurring. We have access to kind of defense unique resources like aircraft at Edwards or a test program. And, you know, I think you know, those kinds of things have have helped compel people that yes, we're we're a reasonable like organization to collaborate with.

Rajiv Parikh 26:52

You're helping to smooth the path. It's not necessarily that there's like this moral question. It's more of how do I make my life easier or how do I get to take the technology I'm building and get it out to to a great customer and a great market?

Speaker 27:05

Yeah. So there's like two components to this. It's like, in order as the government to be able to provide a value proposition to these startups or these technologists, it's like, do you have money to make their thing a reality so that they can build it? Or, you know, are you as a government willing to put your skin in the game so that you can co-develop the technology to be useful? Because we didn't have the the former initially, you know, we really relied on putting our skin in the game, helping code develop the technology, exposing it to the resources that we did have available to us, and then showing that we can de-risk this technology in a very cost-effective way so that the folks that did have funding could see it and you know make a decision financially about whether to invest in the technology or not. So that I think that has always been our operating model. It's more fun that way. Like I actually don't like having a lot of money because now people want to like pitch me things and it feels very transactional. Our job feels more fulfilling when you know we're we're really helping to co-develop the technology and being useful contributors to actually scaling that technology uh within the

How The Studio Scales Impact

Speaker 28:19

government.

Rajiv Parikh 28:19

Is there a stat you can throw at us about the difference that's occurred between three years ago and today with the AI studio?

Speaker 2 28:26

Well, I guess yeah, three years ago we had two people part-time. That was me and John, and then a couple other part-timers, maybe four or three of us, all working this as our side hustle. Today we have two full timers. This fall we go to six full timers. We had zero dollars. The next budget, if it the new NDAA when it gets when it gets approved, will be a four million dollar budget plus some other contributions. And then we have a table at a coffee stand, and then hopefully we'll have an actual portion of the Duran building uh dedicated to the studio here, you know, by the fall. So yeah, I think you know, there's there's some real material permanence.

Rajiv Parikh 29:03

And then in terms of maybe companies affected or technologies accelerated.

Speaker 2 29:08

How many would I claim victory on? I don't know. I would say I would say that we've had like five or six real measurable.

Speaker 29:14

One DARPA program at $69 million. This was a program that we had conceived here within the Duran building. You know, it's become this DARPA program that's now sort of like transformed. I would say the industry, like the physics-based uh modeling and simulation industry. Actually, two of the companies that were like started by professors here at Stanford, Aero Astro, are now performers on that program. You know, we we have partnerships with under in other industry partners who are investing in sort of the the technical thrust that we've outlined within the AI studio, and they're making significant investments in startups that you know we have vetted or at least um are performing in those areas.

Speaker 2 29:56

Yeah, I would say, I mean, we've we've gotten contracts to like six startups. We have strong collaborations with two primes. It's become very real. And you know, the the original idea for this was very much John was getting his PhD, I was at the Pentagon, and I was, you know, I thought, okay, this is gonna be amazing. We're gonna like push research. This would be great. All these problems are gonna show up at our door, people can't wait to bring them to us, you know, and it and we we briefed this what we call the studio engine. And the the the idea for this I evolved much further than just a research center. You know, that it's it to make this work well, it requires, okay, go, you know, knock on the door of the problem holder, find the problem, and and show up with a collaborator from industry and academia to help you triage just how much technical de-risking needs to happen here. And are there existing needs to the desired end that we could just make a small move and answer this problem? And so, really, to make this work requires a pretty big ecosystem of people from across academia, government, and industry to make it efficient and effective. Otherwise, you just don't know. You don't know. There's so much that I don't know. I feel like, you know, I'm constantly insecure in my in my ability to like properly identify the problem or properly identify a pathway for this to get to you know operational use. And, you know, I feel much more comfortable with the collaborators that we've established to try to de-risk those those mistakes or potential mistakes.

Speaker 31:22

I would add that we are sort of building on you know the ecosystem here within within the valley. Like Stanford is sort of the convening authority or the convening place where we can get industry and government and academic folks to really sort of show us what they're building and and collaborate in a sort of meaning, do this code development of technology. But the AI studio effort is broader than just Stanford itself. It's really leveraging the ecosystem around it such that you know we can field technology faster.

Rajiv Parikh 31:54

That's great. It's really helpful to put that into context. You're like a startup in the land of

Fixing The Military Data Problem

Rajiv Parikh 31:58

startups. So we've talked a lot about deploying AI, but Jason, you've uh pointed out a fundamental flaw, right? Legacy military sensors and platforms that simply weren't built to record, produce, or organize data in a way that's actually useful for modern machine learning. So while you were previously on the show, you explained that while AI works well in structured digital spaces, it breaks down quickly in complex physical domains where an adversary is actively trying to create unpredictable edge cases. However, both Ukraine and Iran seem to have advanced in this capability. What do you think has changed in the last 18 months? And what is the fastest way for the US to fix this data problem across the existing fleet?

Speaker 2 32:38

Yeah, I mean, it definitely starts with data. And we had a big pitch on creating a data program that would allow us to bring testing data into a useful configuration for us to do AI and ML training. And that was really John identifying. I'm talking for John, and maybe I should just let him answer the question. But you know, it was really like, okay, we keep finding this problem. Whereas, you know, so again, start at the problem. What's the problem? The problem was a data problem. And we couldn't build the basics, you know, to build an AI and ML um solution. And so what I think, you know, why is it being fielded effectively in in Iran and Ukraine? Because they are embodying AI and you know autonomy in limited fashion at those places. And I think that the way that it's being effective is that it's it's being deployed and the kind of like you simplify the dimensionality of the of the mission that you give it. You're not telling it, hey, I want you to solve the world's problems while flying as a wingman to this thing. What you say is here's a very narrow set of, and maybe actually maybe I'm pointing here, is computer vision as made they leave some bounds over the last 18 minutes. But you know, what you could say is here, you know, this is uh we can embody this computer vision algorithm in a way that meets the size, weight, and power constraints of this device. And we're gonna give it a mission of if you can recognize this thing, then you can execute according to some you know previous set of instructions. But, you know, so it's just kind of like you have this deterministic blend and then some small kind of operational design domain where you allow it to make some kind of inference and you're willing to accept it because you can bound it with some kind of geographical constraint or bound it with some, you know, that that you're not asking it to do super complex things. And so I think that's where you see autonomy being used in a pretty effective way, you know, where it's still limited, is if you try to increase like the domains that it needs to understand or the dimensionality that it has to navigate on its own, that it's it's not capable of doing that yet.

Rajiv Parikh 34:31

When you're talking about that level of autonomy, you're not basically saying, Oh, go get something, and it just the swarm goes out and gets it. You're still there's still a human in the loop, there's still there's levels of coordination that are occurring. I wonder if you could trace some of that. Because a lot of times when people hear autonomous, they think, oh, they just sort of say, go out and take out your adversary here. And it's it's much more there's levels to that. And maybe if you could just sort of step us through a little

What Autonomy Looks Like In Combat

Rajiv Parikh 34:56

bit of that.

Speaker 2 34:56

Yeah, maybe the vignette. Okay, you're you know, Rajiv, you're on the front lines of Ukraine and you have three drones. You would then, you know, you and your team would navigate them to some preordained box that you know that that that you're you're gonna go do your operation in. If during that time they lose communication, you could have already put on them an algorithm that says, if you recognize this Russian tank or this Russian vehicle, and you can verify at some percentage or some confidence level that that is what it is, then they could grant the autonomous ability to take a lethal effect on that tank. But only within this geographic box. Only, only, only, you know, to where they're confident that it will execute at some predetermined level of of performance, probably superhuman. And then, you know, that that would be a way that you could employ an autonomous system. You could, I think there's some of this, but not a lot. Swarming like communication that can happen autonomously between between the drones if they get jammed, you know, and can't get back to a human, that they can then communicate with each other in some way to you know effectuate. I see this tank, can you confirm that that you're getting the same thing? And then that those drones can communicate and say yes and to give it give a you know a backup. Oh, I can't navigate out now, I'm lost. Can you help me find my way? Some of those like more simple things. And then, you know, maybe the you know the the drone that's not that doesn't take the the lethal action comes back home. But that you know, the way to do it is you you make a a very bounded like mission in that essentially degrade out of human in the loop to you know pre preordained agency. Makes sense.

Speaker 36:36

Yeah, my my sense is that the systems that are deployed now are not super exquisite sort of autonomous AI systems. Like my experience talking with folks in Ukraine or talking with folks who are building these technologies, even within our program office, is that a lot of the technologies that we're putting into a quote unquote autonomy stack are technologies that we have built over the past like two decades, right? Like the planning algorithms were are algorithms that you know I was designing like 10 years ago as a master's student. These are a decade-old algorithms. And part of that is because you know, we've we've developed a lot of trust in in these algorithms. They've been useful in various applications. There's a lot of code out there that we can just uh apply. And so, you know, I think that we sort of we as the government or we as a society take autonomy and we think like the most exquisite, you know, we we attach that to the stuff that we're doing in self-driving. I would say the application of like AI and autonomy and defense right now is behind maybe four to five years, maybe even six years of what we would conceive in in self-driving now. Maybe like a counterexample to that or like a caveat would be like computer vision for a one-way attack. But other than that, you know, the the systems that we're deploying are relatively sort of, you know, the techniques are relatively well known and old, but we've just demonstrated that we can do it sort of at scale and with a certain amount of reliability.

Rajiv Parikh 38:14

So much lower cost. The algorithms are have been around, but now that you have the inexpensive computing or well-known ways for these systems to communicate, you can now deploy them faster and iterate faster.

Speaker 38:25

Right. And in and novel applications. I don't think that before the Ukraine war that we were thinking about how computer vision could be useful. I'm sure that people had been thinking about it, but we hadn't demonstrated it at scale for you know war fighting applications.

Speaker 2 38:39

Yeah, because it the, I mean, the the idea of a one-way attack isn't really that new. A cruise missile is essentially a one-way attack drone. It's just that it takes GPS coordinates as it's, you know, I'm gonna send you loose, and here's the GPS coordinates you're gonna go get whether or not you're talking to me or not. And so now it's the introduction of computer vision allows for a somewhat more dynamic. You know, okay, I didn't have the GPS coordinates of the thing I wanted to blow up, but now I have a location, you know, based on computer vision that gives me the same level of confidence that that thing is is what I think it is.

Manufacturing Bottlenecks And Attritable Systems

Rajiv Parikh 39:10

It could find its way. So the 2026 NATSEC Hundred Report, it's a JP Morgan sponsored report, it's an annual report that provides a comprehensive pulse check on the defense tech ecosystem, which was recently re-released at the end of May, and it makes it clear that the major bottleneck in defense tech today is physical production capacity. So, Jason and John, on the Air Force's perspective, how are military requirements adapting to favor startups or or even companies that are out in the field today that can actually scale manufacturing rather than just those delivering a single super cool prototype? How vulnerable are our frontline operations if we can't overcome this physical production bottleneck?

Speaker 2 39:48

I haven't read the report, but I agree production has and probably always will be an issue. The ability to open new means of production is highly correlated with the requirements and exquisiteness, if that's a word, of the demand for the platform. And so one of the benefits of producing platforms that are lower cost, more attritable, is that they're also way more manufacturable. That you don't need these extremely exquisite manufacturing techniques in order to produce them is in and of itself a very strong feature of the reason to depend on them. So you know let's let's it let's accept that the report is correct, then you have a few ways to solve the problem. One is to build more factories with the ability to meet the exact specifications required by the requirement or the design of the of the platform or the weapon. The other is to accept lower quality manufacturing processes for more attributable things. My assumption is that it's not an either-or, that the way that we that we really start addressing this problem is that it's a combination of both.

Rajiv Parikh 40:59

One of the previous guests on the program, Matt Biggie of Crosslink Ventures, one of the companies he's invested in is a company called Crossbow or XBOW. And they basically, instead of relying on the large factories, they actually have field units. They basically manufacture in the field, rocket boosters in the field. So maybe there's as part of your requirements, you're doing additive manufacturing in the field as a way of getting through the typical supply chain logistical challenge of building something more complex and getting it to the field.

Speaker 2 41:28

Okay, so I have a thought on this. Yes. I think when you think about like manufacturing, where should it take place in, you know, Ohio or in you know, some forward deployed or like next to the trench. There's some temptation to say if we can put manufacturing right next to the warfighter, it could respond, especially if it's you know a 3D printed kind of you know, like a multidimensional ability to do production, that it could respond to any warfighter need and its ability to manufacture the thing that the that the warfighter wants. But I don't get the sense that the key to victory is built into the geometry of the drone. Let's just say drone or a walking dog, whatever, whatever it is, some little bot. We'll call it it. Well, let's do the drone because that's what we we use. So, you know, if I could just change the shape of the drone in some way, in a new way, then I can manufacture it on the front line and then it'd be ready to go and I would win. The geometry to me seems probably not like the bottleneck, the limiting thing that allows me to win. What I I think the the path to winning would be that the the manufacturing need is to take a step back because to manufacture even at a low-quality manufacturer, I need some pretty constant power. I need environmental controls, like these are materials, so they're sensitive to these things. And so I think there's a balance like where to where to put the 3D printer or where to put the additive manufacturing capability, how close to the front line does it need to be to win? I think the thing that tackles the logistical problem best is how do I fit as much capability in a given volume and weight? There's a a large furniture retailer, uh Scandinavian known for packing things very tightly.

Rajiv Parikh 43:00

I can't.

Speaker 2 43:00

Yeah. I don't know. Okay, so yes, you know, to me, I'm like, why don't we, I think we need to think of like to the front, the you know, the last mile of the front line should be an IKEA problem. Like, how do I design systems that can be assembled in different ways to give me different capabilities without having to do the manufacturing in a place that I don't have environmental controls and and you know secure power? Because if it's a mechanical solution, now, you know, my frontline operator can say, well, what I really need is a little more range. Okay, I'm gonna snap on some extra long wings on this one, or I'm gonna select a larger tank for for this one or whatever it may be, you know, where we should design modularity, the ability to like apply different sensors or different capabilities to the drone to meet the requirement and have a suite to choose from so that our planner can then know that those things are available to them. But you know, to like ship a manufacturing connect really close to the front line, I don't think it solves the problem because I don't think that the geometry, yeah, this novel geometry is the thing that's going to be the problem. So, you know, I think there's some. Really exciting companies out there, and my my input to them is think about what solves that that last mile problem the best, whether it's the connects full of like different parts in a trench or a connex with a 3D printer in a trench.

Rajiv Parikh 44:12

Great point. This is a really fun question.

GPS-Denied Navigation With Computer Vision

Rajiv Parikh 44:15

So looking five to ten years out, what is one science fiction capability that you both believe will transition into standard operational reality for the US military?

Speaker 2 44:25

Maybe the ability to navigate in a comdenite environment using computer vision. Or a combination. Thanks to John. That's what I have confidence in John's, and therefore I have confidence in our ability to do alternative precision navigation and timing in a comdenite environment.

Speaker 44:40

You took my answer, Jason. I figured I would.

Speaker 2 44:42

Well, he said we both agree on.

Speaker 44:44

Okay, we both agree on. Great.

Speaker 2 44:46

Explain what you mean by that.

Speaker 44:47

Yeah, I would say that one of the challenges that we're we face now and sort of demonstrated within the Ukrainian war is the fact that GNSS is not guaranteed. Our means of navigation, you know, GPS, GNSS, we we just can't rely on. And especially if we're going to embark in a uh in a conflict in the Pacific, it's just very well known that we cannot depend on these on these systems. So how do we how do we do so? Is that because of jamming? Because of jamming, exactly. And a variety of other things, in including potentially interdiction of our uh space asset, our GPS assets, so on and so forth. And so, you know, how do we how do we get around that? And I think that there are many solutions out there. They're essentially like a sort of smorgasboard of like technologies that we can you know throw at this problem. I think what is of interest to us is how do we endow like small UAS. So these drones that are operating close to the ground with the ability to overcome jamming, to do two things. One, to provide sort of high resolution high resolution intelligence surveillance and reconnaissance, and two, find fixed, track, target, engage uh, and assess adversarial positions, and then close the kill chain with our traditional conventional aircraft. And I would say that we want to be able to do this cheaply, right? The name of the game is how do I not build an exquisite system that like allows me to do this. I think that again, computer vision has come a long way, particularly visual-based navigation. So the the ability to sort of use a camera, look at the terrain features, and then basically map match to like Google Maps to get GPS derived derived coordinates and as well as correct your IMU, your really cheap IMU. What's an IMU, John? IMU's internal measuring unit. You know, this is the device that gives the aircraft roll pitch and platform. And so I would say that this is a really cost-effective way. This is technology that we've been building for the past, again, decade. You know, with the advancements in AI, machine learning, the systems are just becoming more and more robust. And we can really rely on these things that are working out in the real world. And we're seeing a lot of companies because of the proliferation of software building visual-based navigation, a lot of companies are taking sort of the open source stuff that has really the fundamental work that's come out of the academic labs into viable products that they can sell to the government. And so I think with additional sort of caveats and harnesses around these technologies that they can be sort of useful for the applications that we care about.

Rajiv Parikh 47:31

I can get from point to point without GPS.

Speaker 47:34

That's right.

Rajiv Parikh 47:35

That's the cool thing that we're gonna do and deal with adversaries and deal with the adversaries along the way. So there's changes in the environment and adaptation. That's that's fantastic.

Spark Tank Sci-Fi Autonomy Game

Rajiv Parikh 47:44

We're gonna go from this wonderfully serious discussion to our game. We're gonna have even more fun. So welcome to the Spark Tank. Today we're joined by two men who are literally engineering the future of flight, cognition, and machine intelligence at the highest levels of national defense. Jason Hansberger and Dr. John Alora. Jason is the director of the Department of Air Force Stanford AI studio. Sitting right beside him is John, the deputy director of the same studio and assistant dean of research at the US Air Force Test Pilot School. So, gentlemen, you spent your days building safe, ethical, and cutting-edge autonomous systems for the real world. But today we're throwing out the standard flight envelope. We are putting your tactical minds, your engineering pedigree, and your deep tech insights to the ultimate test with the tactical autonomy challenge. We are stress testing your knowledge of pop culture's most infamous autonomous systems with our version of two truths and a lie. So, Jason and John, are you ready to prove that your mental inference is as sharp in the history of science fiction warfare as it is in the future of the United States Air Force?

Speaker 48:57

I doubt it, but we'll try. Yeah.

Rajiv Parikh 48:59

That was a very resounding yes. All right, two truths and a lie. And I found these very difficult. So it's gonna be just more fun and logic than anything else. I'm gonna give you three, and then at the end of it, I'll count down three, two, one, and then you'll show me your finger if it's three, two, or one as to which one is the lie.

Speaker 49:20

Okay.

Rajiv Parikh 49:21

Here we go. This one is about Westworld. So hope you know of it. Number one, the park's code was so valuable that Delos executives attempted to smuggle the entire intellectual property portfolio out of the park by compressing the data into the subconscious mind of a single decommissioned host named Peter Abernathy. That's number one. Number two, the AI hosts are governed by a strict hard-coded fail-safe called the primary directive, which physically melts their internal processing units if they cross the invisible geographic grid coordinates of the park's boundary. Number three, Delos Incorporated operated Westworld as a massive loss leader for over three decades, absorbing staggering operational deficits while waiting for their secret guest data mining project to yield a commercialized human immortality project. All right, you ready?

Speaker 3 50:18

Yeah.

Rajiv Parikh 50:18

Right. Three, two, one. Oh, you both said two. Do you know the show well? I remember I remember watching it like six years ago or whenever it was on.

Speaker 50:30

I watched it a while ago, yeah.

Rajiv Parikh 50:31

All right, why would you think two?

Speaker 50:34

I don't remember them melting down. I remember like Dolores was able to get out of Westworld, right? And she didn't melt down.

Rajiv Parikh 50:41

So Alright, well, this is great because number two is the lie. You both win on this one.

Speaker 50:46

Yes.

Rajiv Parikh 50:47

Peter Abernathy, that one's true. So for Peter Abernathy, that is a major specific plot engine from season two. The massive loss leader, that's true, Delos Incorporated, charged guests $40,000 a day. It wasn't actually enough to cover the staggering overhead 3D printing costs for biological humans, cleaning up daily park massacres, and maintaining a massive subterranean tech facility. They tolerated the staggering operational deficits with the long-term corporate goal of mastering mind uploading and selling digital immortality. And of course, why number two is the lie. It sounds like standard science fiction security feature, but it's completely made up. The hosts are stopped by explosive charges in their spines, not a geographic code melt.

Speaker 51:28

Oops. Okay. Wow. We were right for the wrong reason.

Rajiv Parikh 51:32

That's right. You know, you're right on. We're gonna go do something that my younger brother grew up with, and that's transformers. Okay. So here you go. We get into the lore here. Number one, according to the definitive historical lore, the transformers did not evolve naturally. They were originally mass-produced as consumer military products by a sinister, five-faced alien corporate race known as the Quintessens. Number two, the faction known as the Decepticons originally evolved from lower-class military hardware and gladiator bots. Whereas the Autobots were built as commercial consumer goods and domestic labor units. Or number three, the war on Cybertron was triggered when the Decepticons attempted to weaponize a planetary supercomputer called the AllSpark Database, which served as a digital patent registry for every machine design in the galaxy. Which one is false? Three, two, one. Oh my god. Disagreement on this. All right, Jason, why do you think it's one?

Speaker 2 52:39

I don't I don't remember ever hearing any of those words.

Speaker 52:43

Pattern matcher.

Rajiv Parikh 52:45

All right, Joe, why'd you pick number two?

Speaker 52:47

I for whatever reason thought that the Decepticons came out of the Autobots. They were like the, you know, it's like the standard angel story. They used to be angels. I don't know. Maybe that's wrong. I don't know.

Rajiv Parikh 53:00

Or came from the engine, uh, the hand.

Speaker 53:03

That's right.

Rajiv Parikh 53:04

Actually, you both got it wrong. It's number three. That was a crazy one because I would have thought, like Jason, I've never heard of the Quintessines, and I thought I've seen the series and a series and the movie. And according to definitive Transformers lore, the species were manufactured as a global product line by the Quintessens. The Quintessons operated Cybertron as a massive factory until the robots achieved a sentient spark and led a violent worker uprising. Number two is also true. It's the Transformers manufacturing blueprint. The Civil War was deeply rooted in the planet's original manufacturing blueprint. When the system collapsed, these industrial classes organized into warring factions we know today. And why number three is false is while the AllSpark is a real and vital part of the franchise, it isn't a digital patent registry. It is a mythical physical artifact that generates actual mechanical life. The war was triggered by a fight to control the reproduction of their entire species, not a copyright dispute over machine designs.

Speaker 54:07

Yeah, John. You should have known that, John. I spent too much time uh focusing on the fighting scenes, the action scenes, and not too much time on the storyline. So that on me. I have I have to rewatch it.

Speaker 2 54:18

Yeah, I now know what the next Transformers movie is gonna be. What is it gonna be? It'll be the quintessence. We're actually the quintessence, and then the Decepticons and Auto Autobots are gonna have to like combine forces to defeat them. It's gonna be one of those like common enemy movies.

Rajiv Parikh 54:32

Or it could be like a prelude to it, could be like a prequel.

Speaker 2 54:35

Yeah, you could do like, oh yeah, a prequel, and you could get like four movies out of this idea, these quintessenes with one human who can survive it all, you know.

Rajiv Parikh 54:44

Well, that's gonna combine with this next round, which is Terminator. Yes. So here we go. All right, which one's a lie? So number one, Miles Dyson, the director of special projects at Cyberdyne, originally pitched Skynet as a commercial traffic management algorithm for major US cities before the Pentagon stepped in with a hostile venture acquisition. That's number one. Number two, Cyberdyne Systems was a relatively minor tech firm until they acquired the salvaged reverse-engineered CPU and ARM from the first destroyed T-800, using it to launch a massive RD leap in microprocessing. Number three, in the franchise's alternate timeline, the government entity that ultimately takes over the bankrupt cyberdyne assets and finishes the AI project is the military's cyber research systems or CRS division. Alright, so which one's a lie? Three, two, one three and one. Alright. So, John, why'd you pick three?

Speaker 55:47

I thought Skynet was based on real life and I didn't hear DARPA there. So no, I I I actually don't know. I that was like a complete guess. CRS doesn't ring a bell to me.

Rajiv Parikh 55:58

All right, fair enough. Jason, why number one?

Speaker 2 56:00

Okay, I remember when the Terminator and Sarah, what's her last name? Well, they go to Miles' house and convince him to bring that, you know, what the what they're working on. They found the CPU in the arm, that's what, and so he brought him back to the office, and then she told the Terminator to stop killing everybody, and so there's this scene where he's like injuring everybody. But they go to his house and tell him how bad this is gonna turn out, and so it wasn't that he was uh a traffic engineer, he was actually working to develop this.

Rajiv Parikh 56:29

Sarah Connor. Yes, Sarah Connor and John Connor, yeah. John Connor, right? The the savior that we saw in number two. You're right, Jason. It is number one. Number one is actually a lie. So, Jason, you get to win today.

Speaker 56:43

Nice.

Speaker 2 56:43

This is one of the many benefits of being born 10 years before John or 12 years before John, whatever that was.

Speaker 56:49

Jason is constantly bringing up like random facts of movies that I've never seen. So I'm not surprised that he beat me.

Speaker 2 56:56

It's actually aggravating. Like I do movie quotes and then he just looks at me with like a you know a blank stare, and I was like, You're gonna have to have like a movie marathon, right?

Rajiv Parikh 57:04

Here's like the the reference movie the marathon.

Speaker 2 57:07

I mean, in the Air Force, if the movie quotes is a big thing, it's part of the culture. You better, you know, you need to be good on your movie game.

Rajiv Parikh 57:13

I don't know how many times that happens to me where I where somebody will say something and it just triggers a movie.

Speaker 2 57:17

You'd fit right in, Rajiv.

Rajiv Parikh 57:18

And I hit a movement and I'm like, oh, and I give them a response and they're like, What?

Speaker 2 57:22

Surprisingly, most of the Stanford students don't get my movie quotes either, which is also irritating.

Rajiv Parikh 57:26

Too much TikTok. Have you not seen Back to the Future? I mean, come on. They were just showing that at the Alamo down the street. They were just doing a whole Back to the Future, Back to the Future session.

Speaker 2 57:35

Yes, believe it or not, Rajiv. I've run into people who have not seen Back to the Future. I can't believe it.

Rajiv Parikh 57:40

So, yes, apparently Cyberdyne didn't build their revolutionary AI from scratch. They bootstrapped the entire company by using a piece of salvaged future tech. Remember that? They reverse engineered the CPU mechanical arm left behind by the first T-800 in 1984. This is the ultimate, you know, super loop time loop. They bypassed decades of standard IRD to launch a massive breakthrough in microprocessor architecture. Number three was false because when their headquarters were destroyed in Terminator 2, the project didn't die. It simply went through a government acquisition. The US military cyber research systems or CRS division stepped in, purchasing the remaining intellectual property assets. And finally, the system integration that ultimately brought Skynet online. And so why number one's a lie. Jason, as you as you talked about, Miles Dyson wasn't building a municipal traffic app. His division was funded from day one to develop a revolutionary neural network processor for autonomous military hardware. The Pentagon didn't pull off a hostile acquisition. They were the primary enterprise customer driving Cyberdyne's product roadmap all along. Crazy future. Super fun to play

Lessons On Career And Mistakes

Rajiv Parikh 58:43

the game. I got a bunch of quick questions for you. You two are from getting to know you, upstanding leaders and members of the community. And I think folks would just love to hear your responses. So, Jason, if you could be guaranteed to be really good at one thing that you're currently terrible at, what would you choose?

Speaker 2 58:58

I've always wanted to be able to dunk a basketball. Who do you want to dunk like?

Rajiv Parikh 59:02

Uh Jordan.

Speaker 2 59:03

Jordan.

Rajiv Parikh 59:03

I like Mike.

Speaker 2 59:04

But for real, I would love to have like a PhD level engineering skills. That would be fantastic.

Rajiv Parikh 59:09

Well, I think you're kind of doing that.

Speaker 2 59:11

Not there yet. I'm not even a crappy master's student yet.

Rajiv Parikh 59:16

John, would you give him an honorary degree?

Speaker 59:18

I think so. We'll see. We'll see how he does in the next year.

Rajiv Parikh 59:21

You have a one-year timeline.

Speaker 59:22

Better get on it.

Rajiv Parikh 59:23

For your dissertation defense. So, John, what's a mistake you made that taught you more about yourself than any success ever did?

Speaker 59:31

It's hard to say like what was the wrong thing or the decision that I've made sort of throughout my life. It's I feel like all of the decisions that I've made have led me to where I am now. It's hard to sort of think of a decision where, you know, at the time, like it was a suboptimal decision. But then like, you know, looking back at it and seeing where I am now and being very content with with my life and feeling very fulfilled, thinking that, you know, if I hadn't made that decision, then what? Maybe I would have had another life that may not have been as fulfilling. So I have made many mistakes, many small ones, I guess many big ones, but I guess I don't dwell on it too much. So I don't have a good answer for you.

Rajiv Parikh 1:00:12

Even though you have a PhD in pre in a pretty complex scientific field, that's a good part about having a little bit of this goldfish mentality. Forget my mistakes and move, learn and move on quickly, right? And don't dwell on it too much.

Speaker 1:00:25

Yeah, I would say that I think it is important to learn from your mistakes absolutely. But you know, you make so many mistakes that you move on and then you like forget, you forget the mistake ever happened, right? Because you're focused on maybe the next mistake that you're gonna make and hopefully do better.

Rajiv Parikh 1:00:42

That's great. That's great. Jason, what's something you're grateful that your younger self did or didn't do that's paying off now?

Speaker 2 1:00:50

I mean, I think the the big fork in my life started with applying to the Air Force Academy. I'm extremely grateful that I had no idea. I had a chemistry teacher in 10th grade who said, Hey, maybe you should think about applying to the Air Force Academy. And I was, you know, my response was like, I think I'm gonna go to a college. You know, and he's like, It's a college. I'm like, you know, like a four-year degree thing. And he's like, Oh my god, go talk to the counselor. And so, yeah, I mean I was super lucky. I applied to one school because my plan was just to go to community college if I didn't get into the academy. So I'm super grateful that I went there.

Rajiv Parikh 1:01:19

What about that? Like, why did you choose specifically Air Force Academy from your 10th grade?

Speaker 2 1:01:25

You're gonna be disappointed in the answer. It's because it was free and I had to pay for college. It seemed cool. Flying would be cool, and it seemed like it had good academics. I didn't go on a visit. I never, I would I went on zero college visits. So it was a big intervening thing for me. It defined, you know, the last 24 years of where I've worked and what I'd done. And I met my wife while I was TDY in in Germany. She's also American, but you know, happened to be there. Like it set in motion my and you know, my entire life and meeting John. And I didn't take my time there seriously. I didn't I didn't have a plan. I was super naive. I didn't know what that opportunity was. So I think it took me like 15 years of my Air Force career to mature and to like not be so naive. I think it it it literally took that long to really get to a point where I was like kind of able to take the things that I was interested in and make them simultaneously useful. My kids ask me sometimes, you know, what would you go back and change? And you know, the problem is if I change anything, I'm nervous that I wouldn't be where I am today.

Rajiv Parikh 1:02:19

You wouldn't have your four amazing kids and this incredible.

Speaker 2 1:02:21

Well, actually, I have five now. Five.

Rajiv Parikh 1:02:23

Yeah, five, sorry.

Speaker 2 1:02:24

We had one more three months ago. So that's why I'm actually like in this casual gear. I'm on my paternity leave.

Rajiv Parikh 1:02:29

Oh, that's amazing. Congratulations. So now here's another question, John. What's something you thought you'd have figured out by now, but you're still completely confused by?

Speaker 1:02:38

Oh my gosh. Yeah, I would have thought that I would have figured out how to better transition technology, let's say, within the government. Like I thought maybe when we first started this organization that there were existing acquisition pipelines. You know, I knew my role, which was helping shape the, you know, what the what the technological requirements were, and that, you know, we should be able to sort of leverage the existing infrastructure within government to go from an idea to now, you know, a tech a concrete technology that's being useful. What I'm finding is that the government is, well, first of all, that I'm very naive and that the government, like our processes are very complicated. And I'm constantly sort of learning new things about this very complicated bureaucracy, oftentimes to to my detriment, but also maybe not to my detriment in the sense that I found or we found ways to be able to circumvent, you know, some of the things that are not fundamental bottlenecks and ideally get you know technology to you know to the real world. So I thought I would have figured that out like you know, within a year and I'm still I'm still learning.

Rajiv Parikh 1:03:52

That's actually a great answer because if you didn't go and do all those things, you wouldn't keep pushing. That naivete keeps you going. Here's a final question

Leadership Beliefs That Changed

Rajiv Parikh 1:04:02

for both of you. What's a belief about leadership you held early on that you've since completely abandoned?

Speaker 2 1:04:08

As a really young, you know, I don't know, cadet or lieutenant, even an athlete in in the high school. I mean the coach is, you know, your leader. I I guess uh if you would have asked me then of a belief that a leader, yeah, that leader must know pretty much everything in order to like be good at the job of leading an organization. I don't think they need to know everything. There's certainly some element of expertise they have to have to lead an organization. But I think the level of expertise needed to lead an organization is you have to know enough to know what you don't know. If you think you know something, you know, attach a lot of suspicion to that certainty. Good leaders are good at navigating uncertainty through an acknowledgement of just how much they can't know and constantly striving to reduce that.

Rajiv Parikh 1:04:53

Is there a time where that really hits you? Like, is there like a situation where you've been put in many leadership positions? So is there a time where that notion crystallized?

Speaker 2 1:05:04

Yeah, maybe it starts to crack a little bit when I was a commander and I just I had done a lot of different things in the Air Force and I'd never been in an executive officer. I've never really been in the administrative process of the Air Force, and I really had to rely on experts to help me navigate some of the administrative things as I spun spun up on those. But I would say this, I mean, this job, so there's two giant bets on people that I've made in my life. The first is my wife and making a bet that she's the right person to partner with forever. The other is John. I've never made and professionally made a bet on someone's ability to like really be the core value proposition that we're going to build an organization around. And I would say that if I've done anything well as a leader for the studio, it's recognizing where the value lies for the Air Force and the person who represents it. And I'd say that John, you know, is kind of our archetype for that, but it's not enough. Like that's one of the big things that we're working on is developing pipelines to produce. People who can have an expertise in operations and robotics. Because I compare what we're trying to do in the military now, adopting robotics, you know, autonomous robotics as part of our scheme and maneuver to the Army Air Corps trying to adopt aviation into Army scheme and maneuver. They did that with expertise, like engineering expertise, aeronautical engineering expertise. We're trying to do it without sufficient expertise in the engineering that we need. Yeah, I would say this job has taught me, you know, that I uh to be a leader of this organization. I don't have that PhD level expertise. I need I have a guy who does. And my job is to, you know, make sure that we uh empower that in in in as as many ways as we can and you know to try to connect it to maybe the strategy work that I've done before in a way to make it as useful as possible.

Rajiv Parikh 1:06:44

That's great. John, your thoughts?

Speaker 1:06:46

Yeah, I I I always had this sort of belief that leadership was this sort of position where the leader was the the guy that was or the gal that was directing the organization that maybe got all the credit and maybe the it was sort of a singular role, right? You were shaping how the organization was moving and you're the visionary for it. I actually think leadership is a collaborative role. The the leaders that are the best leaders are those that can collaborate the best with their subordinates or the folks around them. The less collaborative you are, basically the more you think that you have the expertise to do the things and you take on those roles, the less I think effective you are, you know, as a leader leading an organization. Going back to Jason, he's like one of those guys that that I really look up to because I think he's a the person that sort of embodies this idea of being very collaborative, being curious, very open-minded, and and trusting that the folks underneath him will do the you know, the the right thing. And so, you know, from my perspective, like leadership is all about collaboration and your ability to like drive the vision and then have the folks underneath you or the folks that you work with, you know, build out and realize that that vision in a collaborative way is what I strive to be.

Rajiv Parikh 1:08:04

John, I think that's a fantastic answer and really does end the show on a great note. So I I want to thank both of you for spending the time with us today for how you've taken your role in the military and used it to then work with the entrepreneurial ecosystem and the technology ecosystem so that we can better safeguard and protect our people. So I I want to thank you both for being on the show. I definitely from knowing Jason in the past, know that uh you guys are both super high character individuals and leaders. I'm I'm inspired by having you here today. So thank you so much. Thanks, Rajiv. It's been a pleasure. Thanks, John.

Speaker 1:08:43

Yeah, thanks, Rajiv.

Final Thanks And Where To Listen

Rajiv Parikh 1:08:49

All right, thanks for listening. If you enjoyed the pod, please take a moment to rate it and comment. You can find us on Apple, Spotify, YouTube, and everywhere podcasts can be found. The show is produced by Anand Shah, production assistants by Taryn Talley, and edited by Alora Ballant. I'm your host, Rajiv Perit from Position Squared. We are a leading AI driven growth marketing company based in Silicon Valley. Come visit us at position2.com. This has been an F Funny production, and we'll catch you next time. And remember, folks, be ever curious.