November 6, 2025 • 48 mins
The Sydney Dialogue (TSD) is just weeks away! If you haven't yet registered, head on over and do so now: https://tsd.aspi.org.au/enquire
In the second episode of TSD Summit Sessions, David Wroe speaks to Anduril co-founder and CEO Brian Schimpf. This is an episode for our defence wonks - Brian talks about the need to build capabilities quickly, affordably and in ways that mean it can be built using a country's existing industrial capacity for years to come. He also talks about deterrence, using Ukraine and Taiwan as real time examples, and the need for companies to be on the side of liberal democracies, and to be thinking about the long-term strategic needs of liberal democracies.
And, of course, the conversation also covers autonomy and artificial intelligence, the need to keep human decision making in the loop and what human-machine teaming looks like. It's a great conversation that covers a lot of ground, and provides an excellent lead-in to the discussions that will be had at TSD on 4-5 December in Sydney.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:01):
Stop the world. Welcome to Stop the World.
I'm Olivia Nelson and. I'm David Rowe.
The defence company Andrew was founded less than 10 years ago
but is now valued at more than 30 billion U.S. dollars and has
contracts around the world, including with the Australian
Defence Force. For the ADF, they're making the
Ghost Shark underwater drone, which is about 6 metres long and
(00:22):
has just started rolling off theproduction line and will provide
all kinds of undersea capabilities.
Yeah, Andrew is one of the more interesting companies around in
national security at the moment.Live was cofounded by the
colourful Palmer Lucky. He's literally colourful.
He's known for his ponchant for tropical shirts, and he is also
famous for creating the Oculus Rift virtual reality headset.
Which he sold in his early 20s. Right, correct.
(00:44):
I think he was possibly the youngest billionaire in the
world at the time, self made. He went on to all sorts of fame
and fortune. And then along with a group from
the defence tech firm Palantir, including the now CEO Brian
Schimpf, Palmer Co founded Andrew.
So it's with Brian, who's our guest today on Stop the World.
He's a great customer. So Brian's in Australia for the
(01:05):
Indo Pacific Maritime Expo at the International Convention
Centre in Sydney, which is the same venue where Aspy's own
Sydney Dialogue will be taking place in just a month's time.
Grab your tickets while you can,folks.
Yeah, Andrew's philosophy is basically to make stuff that
works, make it quickly, make it affordably, and design it in
ways that mean it can be built using a country's existing
industrial capacity for years and years.
(01:27):
That's important because it recognises that we may be
entering a danger period sooner rather than later, and things
like long delays, excruciating supply chains and dependencies,
ballooning costs and so forth just aren't practical in this
strategic environment. So Brian talks about all of
that. He also talks about the need to
deter advert, to deter adversaries by projecting a
(01:49):
sense that you have the nationalcapacity to be a difficult
target not just for a week but for years to come, so the cost
of progression rise. He talks about Ukraine and
Taiwan as well as real time examples.
Yeah, and he also talks about being a company that, a bit like
Palantir, in which Andrew has its roots, is unashamedly on the
side of liberal democracy. So they're thinking about the
(02:09):
long term strategic needs of liberal democracies and that's
what's going to shape the world for the better, not just selling
stuff in the here and now. Agree, and Brian also talks
about the growing role of autonomy and artificial
intelligence. Your favourite topic, Dave?
I see that it was left to the end of these voice overs.
Best to last. About the need to keep human
decision making in the loop, human machine teaming, and about
(02:30):
having software systems that actually process all the
information that is coming in from millions of sensors around
the world. Yeah, it's a great one for the
defence nerds. It was a really good chat with
Brian. I, I very much enjoyed it.
Just for Full disclosure, I recently made a trip to Andrew's
Facilities in California, spent a couple of days touring there
and my travel was partly coveredby the company.
(02:50):
So that said, let's hear the thoughts of Brian Schimpf.
Brian Schimpf, welcome to Canberra.
Thanks for coming on Stop THE World.
Hey. Thanks for having me.
I want to start with an overviewof Andrew, relatively new
company, I think 2017. You cofounded it with Palmer and
some of the others taking a new approach to defence industry.
(03:10):
It's fair to say I'm going to lead you a little bit down a
particular direction here. But I'm interested in speed and
affordability, about building a massive force in a strategically
useful time frame, and this ideaof harnessing the industrial
capacity of a nation so that it can sustain, you know,
potentially itself for years during a crisis or a full fight.
I mean, just talk a little bit about what you're doing and why
(03:32):
it's new. Yeah.
So like you said, founded in 2017 and there was a lot of
principles when we got started that we, we really believe, but
you know, we've really leaned into this kind of class of next
generation, more autonomous, lower cost systems.
And so there's a lot of different parts of that.
One is, you know, the business model of how you can build them
(03:52):
can be very, very different. You know, you don't necessarily
need to serialise it in the way these things have always been
done where the government carries all the risk.
We've been able to lean in on our own dollar for building
factories, you know, Co developing capabilities or just
building them off balance sheet.And that's enabled us to scale
to just a huge number of different capabilities very
quickly. So we've got something like 20
different products, everything from, you know, smaller scale
(04:15):
systems through to things like go Shark that are, you know,
school bus size, autonomous submarine, we're working on
autonomous fighter jets, huge span of capabilities.
But this other idea was really around how do we think about
mass production in a different way.
My analogy for a lot of the traditional defence systems was
we went down this path for a long time, which was technical.
(04:35):
Exquisiteness was the most important capability.
If we had a missile that was 10%better, you know, could could
reach a little bit higher, a little bit further, find
slightly stealthier targets, youknow, that was the right answer.
And we asked industry to build the Swiss watch, you know,
luxury good missiles, aircraft, all these things.
(04:56):
And they did it. They were responsive to that,
but nobody asked the question, can you scale a Swiss watch
factory? And turns out you cannot.
And and so in a lot of ways, this was, you know, kind of
inevitable where we ran into this problem where you see it
today. There are massive multi year
backlogs on these foreign military sales deals out of the
US and a lot of allies and partners have seen worldwide are
(05:19):
feeling like the urgency to rearm, the urgency to build
capacity is very real, but the availability of of actual
systems they can acquire quicklyis just not there, right.
Like you got to wait four or five years to get a lot of these
capabilities. It's a huge problem.
And talking with folks in the commercial world, you know,
learning lessons from automotive, consumer
electronics, we've pulled in a really interesting blend of
(05:40):
people from both traditional defence as well as a lot of the
commercial space. The reality is how mass
producible something is really set when you design the systems.
And if you design it with these commercial processes and you
find smart ways to get, you know, nearly identical
capability, but at a much more producible process, you can
really scale quickly. You can lean into the automotive
(06:03):
and industrial suppliers that doexist.
You can take advantage of composites and metallic
providers, all these different things that that really do.
The capacity is already there. You just got to be able to
design your systems to take advantage of it.
And when you look at like World War 2, the US didn't, you know,
really was able to scale up very, very quickly.
But it was because when you design these systems, they took
(06:26):
advantage of what we could already make.
Like that was the key, right? It wasn't, well, here's this is
super exotic material that is unmanufacturable.
And I that's the linchpin, right?
And, and the reality with manufacturing is you can have
1000 parts on a thing and if onedoesn't show up, you're still
not producing. So it's always these relatively
few defence exotic things that you really got to solve for, but
really taking advantage of designing those out, getting to
(06:48):
commercial supply chain and really being able to scale
quickly, that's that's the formula we've really found has
worked well for us. My to add a personal touch, my
grandfather on my dad's side in Lancashire in northwest England
worked in a fireworks factory atthe start of World War 2.
So you can imagine what he started doing in a very quick,
(07:09):
quick pivot. Now, I mean, what what you're
talking about makes a lot of sense during a fight and a
crisis. It also has deterrence
advantages, of course, because if an adversary can see that
this is what you're able to do, then they have to know that if
they get into a fight with you, then it's going to be over a
longer term. I'm interested in some of the, I
mean, the, the real life examples that we're seeing
(07:30):
unfold at the moment and Ukrainebeing the most obvious one.
I mean, they've reorientated a bunch of their capacity, their,
you know, national capacity to make drones and that's been
decisive in keeping them competitive for 3 1/2 years.
In fact, I'd I'd go so far as tosay there's a real question
about whether Putin would have actually invaded if in 2022 full
(07:53):
scale invasion if he had known that it was going to go the way
it's gone. The same time we've got Russia
sending cheap drones over the border into NATO countries,
forcing them to deploy very expensive F30 fives to match
very, very cheap drones, which is not sustainable over the
longer term. So it's an interesting
demonstration of the way in which sustainable management of
security challenges is actually going to be essential, including
(08:16):
to deterrence. So what are you sort of learning
and taking away from these examples that are going on
around the world at the moment? Yeah, so I think if history is
any indicator conflicts, most ofthese conflicts when you look at
state on state conflicts are protracted wars, right.
Like the reality is industrial capacity really matters.
Again, you know World War Two example, you know the US tanks
(08:37):
were worse than the German tanksjust could out produce them by a
massive quantity. And so the the industrial
capacity of this does really matter.
And when you look at a lot of the, you know, sort of war plans
and, and the way kind of the different services think about
these conflicts is the goal is to have it over in two weeks,
but very few conflicts are over in two weeks.
(08:57):
So the question of, you know, what sort of happens after that
phase is really critical. And this deterrence is always
this capability. And well, and I think the, you
know, Western allies have just this immense capacity to deliver
very decisive, very strong kind of initial salvos.
(09:17):
It's like just sort of the best in the world.
There's no question we can kind of do that very, very quickly.
But I think there is a legitimate question of how long
can it be sustained If you're China, if you're Russia, that's
a legitimate question, right? So, you know, Ukraine's now
producing something like 4 million drones a year.
I think Russia's at like 5 DoD official estimated that from the
(09:38):
industrial base, they believe they can get to 100,000 drones a
year. Seems like a problem, right?
And so the this this issue is, is very severe where, you know,
most of the war gaming shows a lot of the key munitions are out
in something like 8 to 9 days like it is, it is a real
problem. And you know, why do most people
start wars? Well, there's a lot of reasons,
but I think the biggest one is they thought they were going to
(09:59):
win. And if you can make it a little
more clear that we have the the will to fight back, that we have
the capacity to do so and not just the sheer, you know, volume
of force we could bring to the fight, but the sustained
political and industrial capacity to keep that fight up
well in. Fact.
I'd refine it to win at a politically acceptable cost.
(10:22):
That's right. Yeah, exactly.
And so, so there's the politicaldimension of sustainability on
this as well. And, you know, I think there's
been a lot of, you know, historically the, the sort of
free market view of, of, of economics, you know, in, in free
trade has, has been the dominantforce.
But I think people are waking upto the realisation that
(10:44):
countries like China have intentionally manipulated that
free trade system to create strategic strangleholds on the
supply chain, right. And so this, you know, the
obvious example everyone points to is exercising the restriction
on rare earth magnets and the massive impact that has had the
US utilising, you know, high endGPU export controls.
(11:05):
Like all of these are instruments of, you know,
national security that are becoming really important.
But the West has not had a practise in thinking through
this industrial policy question,right?
Which is, you know, I need to not just say free trade is the
ultimate goal, but a degree of sovereignty, control and
assurance that I have not just for defence purposes, but
(11:25):
commercial purposes as well. The ability to deleverage any of
these adversaries I'm dealing with is, is incredibly
important. And so I think we're defence
when we look at this, there's a lot of things that we can do
directly, right? Like we can tap into these
commercial supply chains, we canlook for alternative supply, we
can have a much more scalable approach that creates a lot of
(11:45):
benefit. You then have a set of things
that defence alone cannot solve.The volume does not exist, the
demand does not exist in a sufficient quality quantity to
solve rare earth magnets, to solve lithium production, to
solve germanium, right. Like all these things are, are
are quite critical. And so I think there's got to be
kind of a whole of, of government approach in all of,
(12:06):
you know, kind of the allied countries really around this
question of how do we mutually assure that the supplies we need
will exist, particularly upstream as you look into the
supply chains. And that in a time of crisis, we
are not going to be collapsed economically, that we we have a
more resilient, you know, kind of holistic strategy to this.
Yeah, Yeah. I mean, it's interesting to be
(12:27):
even thinking about these thingsfrom the point of view of a
defence company. You know, you, you, one might
think that as a commercial entity your job is to just
provide, provide, you know, capabilities or you know,
provide particular platforms, provide ammunition, whatever it
might be. I mean, just the fact that
you're thinking about these sorts of things as, you know,
very upstream raw materials in that case, like critical
(12:50):
minerals, I mean, I just reflecton that a little bit more for
me. Yeah.
I mean, I think, you know, our, our sort of view is 1, there's
just straight like kind of commercial interest in this,
which is we've got to be able toproduce these systems.
And so if that means we've got to go stockpile germanium or go
strike deals with people doing mining and, you know, refining
operations, hey, we'll do it if that's the, the right strategy
(13:11):
to take. But I, I think there's a broader
responsibility that we're typically never asked for, which
is a question of how would you be resilient in a time of a
crisis? And that is not something that
is ever written into a contract.Nobody knows how to really
measure that. Even the data to get at that
question is very opaque, right? And the supply chains
historically have been incredibly deep.
(13:32):
So our strategy on this is, you know, if the historic strategy
is I buy, you know, kind of my avionics module and my control
module from these vendors. And then they have been sub
vendors and sub vendors and somevendors you might go back 567
layers deep before you start getting to this question of, you
know, a short supply and raw materials.
It lengthens your lead time and makes scaling much harder and
(13:55):
you have a really critical problem.
What we've kind of tried to do is pull in more of that design
so that we can work with a broader base of suppliers that
are a little bit closer to the problem.
And we can start to get a littlemore transparency around this
question of what is actually scalable for both our own
commercial purpose. How fast can we ramp, you know,
how quickly can we respond to these things.
We don't want two to three year lead times on this, We want
(14:17):
three to six month lead times. So that becomes really critical.
But then this, this question of wartime, I think is an
obligation that every defence provider should be really
looking at, right? Like if, if we do not understand
the realities of what will it take to sustain this industrial
scale during conflict time, well, whose responsibility is it
(14:38):
then, right. And, and if we have the best
first hand knowledge, then we should be obligated to do this.
Now nobody asks us for this. I would love to see it in
contracts. I think it would be something
that would be very healthy. I don't know what I would write
to put that contractual clause in, but it it's a, it's a really
important concept. But I mean, this idea of taking
responsibility is a really interesting one.
(14:59):
I mean, I, I think something like supply chains, for
instance, they are inherently geopolitical and, and the the
recent rare earths deal between Australia and the US is a
classic example. We are working together to deny
China its ongoing stranglehold over something that is clearly
being willing to weaponize. So what you're what you're
talking about there ties in quite well to me with this idea
(15:21):
of almost sort of, I don't want to say patriotic industry, but
industry that does actually takea position in geopolitics.
Now, you know, obviously U.S. companies during the Cold War,
you know, worked with the US government very closely.
But what I mean, one thing that I, at the risk of sounding
ingratiating here that I do admire about Andrew, is that
(15:42):
you're willing to say, OK, we are here for liberal
democracies. We are here for, you know, to
enable them to provide security.We we are actually taking a
position on these things. Does that just work from a
commercial point of view, or do you find yourself having to
balance one thing against another in certain contexts?
Or or does it just a line of of its own accord?
(16:05):
Well, I, I think there's, there's a lot of nuance to these
things. But yes, I think that is so
we've, we've taken a very clear position that, look, we're
aligned to Western allies. We have a clear set of
principles and we are not going to do things outside of the
policies that the governments are looking for.
And I think that's our our sort of division on this is you don't
(16:28):
want, you know, kind of tech Bros setting policy on these
things. It seems like a very bad
strategy for the world. You want tech companies to
actually be accountable to saying, here's the best
information I can provide, Here's the trade offs and, and
you need the government accountable to actually making
policy decisions on these things.
(16:48):
And so we tried to be good partners in that, you know, and,
and really look at creative waysthat we can work within the
legal and regulatory frameworks to still create a set of policy
options. So one of these has been really
key to us is looking at kind of this question of sovereignty
export, you know, and how do we think about what's aligned to,
(17:08):
you know, the traditional approach rates buy through the
US government, it's export controlled.
That's it, right? And, you know, sometimes there's
some localization and offsets and production, but I think
there's a lot more creative strategies to this where, you
know, maybe we can take some of the technology that's built in
the US. Maybe not all of the most
sensitive pieces need to be exported because other countries
also have great investments in unique radar capabilities or,
(17:31):
you know, unique, you know, warhead capabilities or things
like that, that that may be justmuch more specialised sensitive
to where the US believes they need control.
But we can find win win strategies, right where we can
accelerate allies and partners on what they need to produce,
what they need to build up and give, you know, kind of US,
Australia, any of the countries we work with closely out of the
(17:53):
security assurances for their own military that they need to
have. And so there's, there's these
kind of very straightforward, you know, but require some
creativity strategies you can have around ways to get at this
first principle problem of how do we enable allies and partners
while still protecting the sovereign interests of every
country we operate in. And I just don't think you can
be a defence company and not have a view on these
(18:16):
geopolitical questions. It is the, you know, I think
everyone was allowed to forget for about 20 years that this
mattered when it was just so clearly a hegemony on, on, on
kind of international order. That's just not true anymore.
Every country, kind of every, every company needs to take a
bit of a side. And we've seen this in the tech
community where for a long time in the finance community as
(18:38):
well, they were so thoroughly compromised by the promise of
returns in China that nobody in the national security community
ever thought was going to materialise.
It's like, what does it mean to get a high return in China?
Are you ever going to get your money out?
What does it mean for your investment to go to the
literally 0? Because, you know, the autocrat
decided today that he doesn't want that industry or that money
is not coming out. And it was a big wake up call to
(19:00):
a lot of these guys who had justhuge interest in China and then
realising that that was not someone you could trust, that
was not a good bet to make. And that they kind of took for
granted the the rules of order that the, you know, the kind of
liberal democracies shared. So we've seen a huge change in
just even the tech community over the last few years where,
you know, when we started, I remember there was massive
(19:23):
protests and concerns that Google was working with the US
government on AI. Now last year, I think nearly
every tech companies like openlycommitted to supporting AI with
the, with the government like that is a huge change.
And, you know, it really, the world has really shifted in the
last few years where this is notoptional anymore.
(19:46):
You actually have to really think through these kind of key
strategic questions. The government's getting more
sophisticated, the the companiesare feeling more responsible and
there's just a much greater alignment around, you know, not
taking for granted that the the countries that have provided so
much peace and prosperity. Interesting.
I mean, the, you're, yes, you'rea defence company, but the, you
(20:07):
know, the distinction between a defence company and a tech
company is much less clear than it, it might have always been.
I'm just interested in how that that transition where the tech
company, where the tech industryhas started to I suppose come a
little bit closer to where companies like Andrew and
Palantir have been for some years now, what that experience
(20:28):
has been like from Andrew point of view.
I mean, you presumably feel somewhat vindicated by that.
Yeah, When we started, it was, you know, when we were going out
in 2017, remember, it was Trump won.
There was a massive backlash against doing anything with the
federal government. And our sort of view is been
(20:50):
that this is just, this is a bipartisan issue, this matters.
And the system is much worse than anyone sort of realised.
And the world's a much more dangerous place than anyone
realised. And I've been a lifeline
Democrat. A lot of the Co founders said as
well. We've had a very balanced kind
of partisan view. But everyone agrees on the
mission of what we're trying to do.
It's like best technology to ourallies and partners in the US
(21:13):
and our war fighters. Like that's what we need, right?
Like no question. And you know, also it was this
period where the sort of employee revolts over working on
anything controversial is massive.
So to start with a company that was working on, you know,
defence and we had investors saying commit to never making
weapons. Of course we're not going to
(21:34):
commit to that. Like we're a defence company.
Like this is an important aspectof it.
But having that clear vision of like, this is what we are going
to do. Do not come here if you do not
support this mission. But here is what we are doing
and we're clear with it. We have communicated it clearly,
publicly, transparently with ouremployees, with the media, with
(21:55):
government, and that that has worked really well.
So I think this sort of was a huge lack of sort of
authenticity and transparency inthese things.
And then it really what, you know, I think Vladimir Putin and
G have done an amazing job of convincing the rest of the world
that this actually matters, right?
They've really made it very clear.
Have you ever found it difficultto get people as a result of
(22:15):
that, you know, stated position?Especially when the demand for
highly qualified tech people is so high.
You know, it honestly, not really because I think the,
there was such a desire for likeclear, authentic views of, of
what your mission was at a time when there was just this sort of
very crazy, you know, view of, of, of your relationship to the
(22:39):
companies and the mission on it.So even when it was more
controversial, you know, kind ofpre Ukraine invasion, even then
there's the clarity of what we were doing, I think resonated
with a lot of people. This this idea that people in
tech don't want to work on defence, I think is a very
minority, but loud minority position.
And I think the vast majority are in favour of a strong
(23:00):
National Defence. Like it's, it's kind of the, the
underpinning of safety and security we've had for a
century. So the I never found it that
challenging, right? And, you know, it's I actually
think it was like almost better when we were more controversial,
like, you know, the when, when too many MBA show up because
(23:22):
you're very cool, like you should really watch out.
Like it's, it's you might have jumped the shark at that point.
And so like when you're cool, you're not sure if people are
there for the right reason, Like, are you there because this
is a great stop in your career and it's a popular company, or
are you there because you actually believe in it?
And we want people that are actually, you know, kind of
thoroughly believe in the mission of it because it can
really corrupt you very quickly otherwise.
(23:43):
And so that's that's something we still philtre for still, you
know, very transparent about what we do for that reason.
And, and I'm I'm not sure being a popular company is actually
that much of an advantage in reality, right?
All right, Speaking of sharks jumping or otherwise ghost
shark, let's get on to that. The underwater UIV that you're
building in collaboration with Australia.
(24:06):
I mean, interested in a couple of things.
One is that it to me sort of it epitomises this, this issue
we've got about fast capability acquisition at the moment.
So you know, Australia is going to have fantastic submarines in
the 20 forties, right. But you know, there is a window
(24:27):
of danger that we're in now. Some people, you know, some
China experts say that, you know, 27 to 30 really is
potentially that sort of peak period when Xi Jinping might
say, OK, this is we're not goingto get any stronger than we are
now relative to other powers. So certainly, you know, quick
capability acquisition is something like that.
A country like Australia really,really needs to think about
given that we've got these more exquisite capabilities coming
(24:50):
online somewhat in the in the future.
I'm interested in your views on on that and how Ghost Shark sort
of fits into that. But also just in broad terms, as
much as you're able to in a, in a, you know, obviously an
unclassified podcast, give us your thoughts on why Ghost Shark
is useful for Australia, given its geography, given its
(25:10):
strategic challenges and so forth.
Yeah. So the the Ghost Shark programme
was for us as a company we really look for forward leaning
customers where there's this combination of creative military
thinking. So you know, how are you going
(25:30):
to solve these critical operational challenges in
different ways with a willingness to move fast and try
different models on acquisition capability development and
really experiment with more modern approaches.
And with Ghost Shark, we really found that with with the Rand
and ASCA and it was it was fantastic where you know, I
think we went very, very fast period from first conversations
(25:54):
through them saying this is a key capability, let's move out.
We entered a Co development programme, which is one of the
first we'd ever done. We're just sort of risk sharing
approach, very open on thinking through kind of governance
structures and how do we, you know, kind of break out of the
standard mould of destroying these schedules and costs with,
(26:14):
you know, obsolete requirements and, you know, over overburdened
regulation. But really just trying to trim
to what's necessary for the capability we're trying to
deliver. And that sort of creativity and
willingness to, to break the mould is, is, is rare.
I was really impressed with how this has gone throughout the
programme where, you know, I've seen this go so badly in US and
(26:37):
plenty of other places where, you know, kind of that risk
aversion, you know, this is the way we've always done it, you
know, coming in with, you know, man, submarine requirements on
to uncrewed capability. That's, that's a, that can just
destroy a programme, right? Like can destroy a capability
and so it required a degree of flexibility around kind of all
(26:59):
aspects and willingness to re examine all aspects of how do
you think about building, testing, deploying, operating
these capabilities. And so these are never just like
an acquisition question, right? Like I I think the the this
often gets, you know, kind of viewed as does the government
have the authority to acquire differently?
That's one of the five dimensions that need to be true,
(27:22):
right? It's, it's so many parts of
building and deploying these things quickly that you have to
solve simultaneously. And, and this is really key to,
you know, how are you going to actually get the mass and
capability in a rapid way, right.
So you're going to buy some things off the shelf, but it's
got the backlog problem that we've already talked about for
things that are in production aswell as getting new capabilities
(27:44):
through that traditional pipeline.
You know, it's gotten longer andlonger and longer.
There's there's really no way out of it.
And this is where I think this class of autonomous and lower
cost systems really can shine where the cost to develop is not
prohibitive. The cost to purchase is much
more affordable and then the ability to project, you know,
(28:05):
kind of power at scale is, is really, really compelling,
right. So we take kind of go shark and
what capability can this provideto Australia?
You know, obviously Australia issurrounded by ocean, right?
Like the, the maritime side has to be solved, the naval side has
to be solved. And the sheer volume of of
coastline and you know, kind of key choke points and straights
(28:28):
that you have to control and protect is is massive, right?
And at reasonable ranges, crude submarines are incredible,
right? Like I am a huge fan of Virginia
class. I think it is the most
impressive capable weapons programme ever done.
Like it is just a phenomenal capability.
I have nothing but good things to say about it.
But it is also the most complex weapon system ever developed,
(28:51):
which is, you know, part of the reason it's so spectacular.
And and you're just inherently limited by the scale that you
can actually deploy these app. And so as we look at where these
uncrewed systems can really shine, it really is about scale,
right? It is about the number of places
that you can be positioned, the amount of effects and
surveillance that you can acquire in the amount of
(29:13):
dilemmas you can create for youradversaries, where, you know, if
you can put dozens of these systems into the ocean, that
creates a lot of uncertainty foryour adversary around, you know,
particularly subsea where it's very hard to find things.
It is very hard to, you know, kind of mitigate these
capabilities. And it can create a massive
(29:33):
amount of uncertainty for an adversary trying to approach the
coasts, trying to conduct a manoeuvre where you can really
hold them at risk without reallya lot of, you know, capital
assets or sailors lives at risk.And that is a massive capability
enabler that I think the ran really understood and leaned
into as you know, as this reallythis bridge till you can get
(29:58):
Virginia class to where you wantit to be.
But even as an augmentation of just the sheer volume that
you're going to be need to deploy in any fight.
Let's broaden it out a little bit on on autonomy, autonomy
automation on crude systems. I mean, is there I, I, I, I see
(30:19):
and I pardon my ignorance, I should know this off the top of
my head better. But the way it would, you know,
there's something like the ghostshark could team with a Virginia
class, for instance. I mean that that just seems like
a natural use of it, a bit like the way air platforms we're
talking about doing in a similarway.
Just talk about human teaming a little bit, human AI teaming or
(30:41):
human autonomous teaming a little bit into the future.
I'm it seems to me quite possible that teaming is a
transition until automation and AI get to where, you know, the
role for human beings is simply,you know, our, our value
contribution starts to shrink tothe point of negligible.
(31:01):
Do you see it that way, or do you see the human role as being
sustainable forever? So I fundamentally this is a
question of sort of values and, and, and how you're going to
hold people accountable to what military systems do, right.
And so fundamentally to me, thisis the West will always want
(31:24):
human accountability around lethal decision making,
deployment of effects, right? Like all these things that are,
you know, have real political and societal consequence, we're
going to want human accountability around it.
We're never going to get to a world where we say that's
absolved. The AI told me to do it and it's
not my responsibility. And everything we've designed is
(31:45):
really about thinking that question of where is the right
point for humans to engage and how does that further enhance
human accountability, not degrade it, Right.
So the AI and the autonomy can be phenomenal doing a lot of the
more mechanical tasks. You know, I don't need a human
to plot the route that a missileflies or a submarine goes.
That is a perfectly suitable task for software to do, for AI
(32:10):
to do this. AI can recommend courses of
action and say, hey, I think youshould do AB and C.
And a human can approve that andunderstand the context and the
consequences of those decisions in a way that's very hard to
digitise. And so I think there will always
be this role for human accountability in it, but it's
always a matter of degree on these things where in a lot of
(32:31):
ways there have been lethal autonomous weapons for decades
now. The way a modern torpedo works,
you give it a box and you say anything that matches this
acoustic signature in this box that is a legitimate military
target. That sounds a lot to me like a
autonomous weapon. And so in a lot of ways, what
we're debating is really then these controls around what
(32:54):
context do we allow the machine to make these decisions?
What are the rules of engagementand how much freedom and
latitude do you provide it, right.
And then where does the human decision making and human
accountability really come in? You know, a lot of the Western
systems, whoever launched the weapon is accountable for the
results of what that weapon does.
I think it's a good system. We should probably keep that.
(33:15):
And so the so there will always be that part, right?
And so I that's how we really think about it.
Yeah, there's the stepping stonedimensions of how sophisticated
is the AI, how much control can it really do.
I think that is true. But I think the fundamental
thing is really around this question of values and how do
you want to provide accountability into the system.
(33:37):
But it's largely the values and ethics question rather than the
capability question. I think that that is the long
term question is values and ethics.
I think the near term question is kind of 1 of adoption, right?
So even if the technology was perfect to solve all these
problems, I think you are going to inherently see correctly the
desire to have a relatively short leash to start, right,
(33:58):
relatively prescriptive mission sets.
And then over time as trust is built and we understand what the
capabilities are, then you will release kind of more control to
these systems to perform more autonomous papers.
We've seen this with nearly everything we've deployed, even
counter drone systems. When it starts, typically the
operators want to look at, you know, moving the camera
(34:18):
themselves. Then they're OK now I'll let the
system move the camera. OK.
Now I'll recommend engagement patterns.
So you you kind of see a progression naturally as people
build more trust in the system of what that right balance is.
OK, just back on Ghost Shark fora moment.
It's described by Andrew as a sovereign undersea capability.
Just understand, explain to me what sovereign means in that
(34:40):
context. Does it mean that Australia can
do everything on its own at somestage in the force?
Well, for the foreseeable future, potentially including
one in which you know something happens in the Australia US
alliance, for instance, probablynot a hopefully not a political
collapse, but something else might just mean that Australia
needs to do it on its own for some defined.
(35:01):
Period. Yeah.
So for us, we the team that designed Go Shark is based in
Australia, it's Australian engineers.
We have tight collaborations with the US counterparts, but
on, but all of the core design expertise resides in Australia.
The production capacity is basedin Australia.
(35:23):
We can fully assemble these go sharks here.
And the supply chain, you know, there's a degree of you know
kind of AUCUS components to this, there's some UK suppliers
and things like that. But you know, the vast majority
of the metallic components, the composites, the battery systems,
you know, core electronics, assembly, all of those pieces,
we've built out an Australian supply chain around this.
(35:46):
So the ability to actually produce independently of really
nearly any other country is, is quite high.
And so the so for us, sovereign means kind of that independent
ability, a, a mindful understanding of where those
interdependencies exist and something kind of acknowledged
by the government of Yep, this looks correct in terms of the
(36:08):
interdependencies I want to have.
And so for us, it's very much about creating the ability for
Australia to act quite independently as needed.
You know, there's several payloads that I don't get to be
brief to. I'm not an Australian, so that's
perfectly fine with me. I think that's a good thing.
Good, good. All right.
One thing we haven't talked about yet is Lattice, which is
(36:30):
really the core of your company.That's the the the the software
system that there's essentially a command and control system.
Just talk about how that forms the core and how you sort of
build around it. And interestingly, I mean,
we'll, we'll ghost shock. Will ghost shark employ lattice?
(36:50):
Is it, is it? I don't think lattice is just
for Andrew capabilities, correct?
You can use it for other things as well.
But just talk a little bit aboutthe role of lattice.
So stepping back slightly, the the way modern software has been
built in Silicon Valley is this idea, these very general purpose
platforms and the these are a lot harder to build with quality
than most people would appreciate.
(37:12):
You have to build immense tooling around your development
infrastructure and simulation. How do I test my software?
I, you know, I have to simulate sensors and autonomous
behaviours and all these networkconditions, all these different
pieces. You know, I have to build all
this tooling and instrumentationfor monitoring and deploying all
this controlling configuration. Just a, an enormous amount of
(37:32):
software you end up having to build.
And that is below sort of the tip of the iceberg, which was
what people, you know, kind of think of with these, which is I
get a drone to go from point A to point B into a mission.
So the complexity of these modern software stacks is just
immense. And so really every major tech
company has thought about this is how do I share that load
across a number of different capabilities and build these
(37:56):
kind of very flexible software platforms that allows me to
tailor to any, any application. And when we look at the the span
of problems we're trying to solve, it's really this question
of how do I build for a world where I have much larger numbers
of sensors on the battlefield, of drones on the battlefield,
subsea in the air, and I'm pairing those with manned
(38:16):
systems. So solving all those fundamental
problems. And that's how do I think about
networking these. How do I use AI to process that
information at the edge? What's the right way to move
that information around to create this holistic picture of
a battle space? How do I express an intent to a
robot? Like what does that mean?
And how do I represent that in adigital way that can be
(38:37):
transmitted over a network? You know, how does a human want
to interact with us, right? What do I want to see?
What decisions do I need to make?
How do I make sense of a increasingly complex
battlefield? So we have this view to where a
lot of these future battlefieldsgo.
Much more systems, much, much more disaggregated.
Everything is networked, but youhave constant jamming.
(38:59):
So how do I have resilience baked into that?
And I need to create this transparent picture as best as I
can for all the information I can get.
And that at the core is what we've tried to build Lattice to
solve. Now, obviously making a subsea
drone is very different than making a you know, a small
quadcopter is different than making a fighter jet, and it's
very different than talking to amanned F35 S the literal
(39:21):
components. Yeah, those vary from project to
project, right. And but that core infrastructure
around, you know, how do I network this represent
information and think about managing it that is shared
across. And so Ghost Shark, nearly
everything we do has these kind of core lattice pieces into it.
And the greatest part of this isbecause we've shared this code
(39:42):
base across so many systems, allthose learnings fold back in and
we can provide these updates to every system as it gets smarter,
better. And we take all those learnings
and distribute it across the whole fleet of all the
capabilities we're building. I mean, I find the idea of eyes,
ears and brains to be a useful analogy just to sort of more
instinctively grasp it. I mean if I had 10,000 eyes and
(40:06):
5000 ears. That might sound great, but if
I've got one brain, unless it's supercharged in some way with
AI, which may come down the track.
And I suppose, I mean, actually I'm, I'm let's talk about that
for a moment. What's what's the role of AI in
Lattice and what's the role of AI going forward?
Do you think in managing all of that sensor data, all of that
sort of decision making complexity, which is going to
(40:29):
which is already already exists for, for, for war fighters, but
is going to become more and more, I suppose potentially
overwhelming unless we we manageit well with AI.
Just tell us a little bit about the role of AI there.
Yeah, so I kind of think I always like to simplify these
problems down a little bit. So like, what are you really
trying to do for the vast majority of kind of military
(40:50):
applications for what you're trying to do with the sensors?
You're trying to figure out where your friends and allies
are. So where's the blue force?
And you're trying to figure out where everything else is in the
environment. So what's the map?
Where's the bad guys? What are they doing and how do I
find them? So when you build that, boil
that down, it's this question of, you know, how do I interpret
sensor data to know, hey, is this a ship?
(41:10):
Is this a tank? What am I actually looking at?
And then you can kind of then think of relatively simple ways
of moving this information around.
So a lot of what I think we've done well is think about these
relatively simple abstractions, how to do it.
These are the AII think the we've been doing what's called
the prior generation of of AI where this was the computer
(41:31):
vision using machine learning for detecting signals of
interest in, you know the electromagnetic spectrum, what I
think we would call now kind of more discriminative AI.
We've been doing that since the founding right so we deployed
border security towers and counter drone systems that have
been using computer vision and sensor fusion, you know for
eight years now and it's proven to work incredibly well right.
(41:52):
We probably the largest number of deployed production, you
know, AI systems out there. The and then now there's more of
this question of generative AI and these more reasoning systems
and what implications does that have?
And I'm extremely optimistic around this.
So, you know, I, I've been really fascinated by things like
(42:12):
alpha go alpha star. A lot of the things that, you
know, kind of came out of GoogleDeepMind around the ability to
learn these strategy games. And when we think about, you
know, fighter jet tactics or, you know, how would you have a
tech helicopter or anything, it is so there's so many variables
these pilots process. It's so hard to code into like,
you know, a, a particular algorithm and the ability for
(42:36):
these machines learned kind of generative strategies to outpace
human level performance or certainly any of these
traditional approaches is, is massive.
And we've seen this in some of the research we've been doing in
house with just kind of these reinforcement learning
approaches, just massively within like 4 weeks of work can
massively outpace any of the alternatives we've tried.
(42:56):
And so I think the ability to get to these superhuman levels
of performance on, you know, kind of these tactical
behaviours is very, very near term.
What I don't think works in practise is sort of this idea
that we're going to have this massive central brain kind of
puppeteering this large scale operation.
It's not resilient. I don't even think it works with
(43:19):
the laws of physics with how much information you'd have to
move over networks to get that data back.
I just, it's no amount of star link that's going to solve that,
just too much data to move. And you've created this
incredibly not resilient system all of a sudden.
And so when we think about this,it's really about how do I
create these layered intelligentsystems that I understand how
(43:40):
they operate in concert, but they're each making smart,
independent decisions. That's a resilient strategy.
And it also, by the way, is how the Western militaries are
organised. They figured out this mission
command structure is the right approach, give a degree of
autonomy and discretion down to your lower echelons.
You get phenomenal results, right?
And you kind of train for that. You, you operate that way.
(44:03):
You, you let them have initiative and that tends to
deliver much better results. So we kind of mirror that and
how we think about the technology perspective, which is
the right answer probably looks like something that is more
resilient, more distributed, more at the edge than this sort
of mythical central brain planning out the whole battle
(44:23):
space. I just, I don't see it.
It doesn't make sense to me. I don't know why you would do
that. It speaks to my lack of world,
real world experience either in engineering or war fighting that
I my head immediately starts going to a, you know, the the
human president or Prime Minister with one button, you
(44:43):
know, shall I win the war? You know, yes, slash no.
Which one do you want to click? That's where I see it eventually
sort of graduating towards. But you're doing a good job of
dissuading me that that that's going to be the case.
And rather it's to be down at more of a, you know, a lot of
decisions contributing in in a tactical, operational and even
(45:06):
strategic levels. That's correct.
That's right. And I think the, you know, you,
maybe your systems are implemented like so the way I'm
saying it is really about thinking of how the, you know,
future robot armies and these command and control different
echelons will work. I think it has to mimic the
structures that we have, that we've learned over a century of
what is an effective way to organise and operate a military.
(45:28):
I think those are sort of timeless principles that likely
are going to transcend the specifics of the technology.
Now maybe the degree to which that a human at that echelon,
the effect they can have will bemassively amplified, right.
So, you know, 1 pilot might be able to control 100 droughts.
Well, that is a massive increasein the amount of capability that
(45:49):
we can get at every echelon. And that is a a huge change that
technology can bring forward, but it doesn't remove the need
for that initiative, that creativity and that sort of
mission command structure at every echelon.
The other side of this, though, is what you can start to do with
simulation to really understand what is possible, what are the
likely outcomes of any given conflict, what strategies
(46:09):
actually work. There's sort of this game
theoretic notion or thinking about capability development
that we've really leaned into around we, the way we operate is
when we think of a new capability, for example, we
will. So here's the war fighting
problem I'm trying to solve. We'll run dozens of different
strategies of technology. And if we had this range or this
(46:30):
capability or it flew this fast,what if it was on the land?
What if it was in the air? And we can simulate those and
you can say, hey, here's how these potential conflicts and
these engagements would go. So the state of the art today is
reasonably complex. You can kind of get a, you know,
a, a fairly complex engagement and simulate that out.
But there's a huge amount of potential with generative AI to
(46:52):
very rapidly lay down super complex scenarios over a huge
number of variables. And you can get a much more
comprehensive idea of which are going to be the dominant
strategies and how likely am I to prevail with which strategies
and how to create those turn effects.
Now the, the hard part with thisis so many of these decisions
(47:13):
are really in the brain of one human at the end of the day
right of will China invade Taiwan?
Well, you can do all the military analysis you want, but
ultimately GE will have his reasons for doing it or not
doing it based on his perceptions of likelihood of
success, the political reality around him and what he needs to
do. You know, same with Putin right?
And the ability to use AI to forecast that humans brain.
(47:36):
That doesn't seem like it's going to happen anytime soon.
And so I think that the goal we think of with these things is
really around, you know, what are we trying to construct as a
company? We're trying to provide that
capability dimension in that deterrence equation.
And if you can make the outcome of these conflicts quite clear,
AI probably helps you can make these things much more obvious.
(47:58):
That is a good thing, right? Like that is what we were trying
to do. So that conflict does not
happen. And and that's ultimately what I
think our our goals and ambitions are on these things.
It's a great message to end withand and wrap up.
Brian, thanks so much for comingon Stop the World.
It's been a great chat. Thank you.
Appreciate it. Thank you.
Thanks for listening to this special episode of Stop the
(48:20):
World, the TSD Summit Sessions. We'll be back with another
episode next week.
Stop the world. Welcome to Stop the World.
I'm Olivia Nelson and. I'm David Rowe.
The defence company Andrew was founded less than 10 years ago
but is now valued at more than 30 billion U.S. dollars and has
contracts around the world, including with the Australian
Defence Force. For the ADF, they're making the
Ghost Shark underwater drone, which is about 6 metres long and
(00:22):
has just started rolling off theproduction line and will provide
all kinds of undersea capabilities.
Yeah, Andrew is one of the more interesting companies around in
national security at the moment.Live was cofounded by the
colourful Palmer Lucky. He's literally colourful.
He's known for his ponchant for tropical shirts, and he is also
famous for creating the Oculus Rift virtual reality headset.
Which he sold in his early 20s. Right, correct.
(00:44):
I think he was possibly the youngest billionaire in the
world at the time, self made. He went on to all sorts of fame
and fortune. And then along with a group from
the defence tech firm Palantir, including the now CEO Brian
Schimpf, Palmer Co founded Andrew.
So it's with Brian, who's our guest today on Stop the World.
He's a great customer. So Brian's in Australia for the
(01:05):
Indo Pacific Maritime Expo at the International Convention
Centre in Sydney, which is the same venue where Aspy's own
Sydney Dialogue will be taking place in just a month's time.
Grab your tickets while you can,folks.
Yeah, Andrew's philosophy is basically to make stuff that
works, make it quickly, make it affordably, and design it in
ways that mean it can be built using a country's existing
industrial capacity for years and years.
(01:27):
That's important because it recognises that we may be
entering a danger period sooner rather than later, and things
like long delays, excruciating supply chains and dependencies,
ballooning costs and so forth just aren't practical in this
strategic environment. So Brian talks about all of
that. He also talks about the need to
deter advert, to deter adversaries by projecting a
(01:49):
sense that you have the nationalcapacity to be a difficult
target not just for a week but for years to come, so the cost
of progression rise. He talks about Ukraine and
Taiwan as well as real time examples.
Yeah, and he also talks about being a company that, a bit like
Palantir, in which Andrew has its roots, is unashamedly on the
side of liberal democracy. So they're thinking about the
(02:09):
long term strategic needs of liberal democracies and that's
what's going to shape the world for the better, not just selling
stuff in the here and now. Agree, and Brian also talks
about the growing role of autonomy and artificial
intelligence. Your favourite topic, Dave?
I see that it was left to the end of these voice overs.
Best to last. About the need to keep human
decision making in the loop, human machine teaming, and about
(02:30):
having software systems that actually process all the
information that is coming in from millions of sensors around
the world. Yeah, it's a great one for the
defence nerds. It was a really good chat with
Brian. I, I very much enjoyed it.
Just for Full disclosure, I recently made a trip to Andrew's
Facilities in California, spent a couple of days touring there
and my travel was partly coveredby the company.
(02:50):
So that said, let's hear the thoughts of Brian Schimpf.
Brian Schimpf, welcome to Canberra.
Thanks for coming on Stop THE World.
Hey. Thanks for having me.
I want to start with an overviewof Andrew, relatively new
company, I think 2017. You cofounded it with Palmer and
some of the others taking a new approach to defence industry.
(03:10):
It's fair to say I'm going to lead you a little bit down a
particular direction here. But I'm interested in speed and
affordability, about building a massive force in a strategically
useful time frame, and this ideaof harnessing the industrial
capacity of a nation so that it can sustain, you know,
potentially itself for years during a crisis or a full fight.
I mean, just talk a little bit about what you're doing and why
(03:32):
it's new. Yeah.
So like you said, founded in 2017 and there was a lot of
principles when we got started that we, we really believe, but
you know, we've really leaned into this kind of class of next
generation, more autonomous, lower cost systems.
And so there's a lot of different parts of that.
One is, you know, the business model of how you can build them
(03:52):
can be very, very different. You know, you don't necessarily
need to serialise it in the way these things have always been
done where the government carries all the risk.
We've been able to lean in on our own dollar for building
factories, you know, Co developing capabilities or just
building them off balance sheet.And that's enabled us to scale
to just a huge number of different capabilities very
quickly. So we've got something like 20
different products, everything from, you know, smaller scale
(04:15):
systems through to things like go Shark that are, you know,
school bus size, autonomous submarine, we're working on
autonomous fighter jets, huge span of capabilities.
But this other idea was really around how do we think about
mass production in a different way.
My analogy for a lot of the traditional defence systems was
we went down this path for a long time, which was technical.
(04:35):
Exquisiteness was the most important capability.
If we had a missile that was 10%better, you know, could could
reach a little bit higher, a little bit further, find
slightly stealthier targets, youknow, that was the right answer.
And we asked industry to build the Swiss watch, you know,
luxury good missiles, aircraft, all these things.
(04:56):
And they did it. They were responsive to that,
but nobody asked the question, can you scale a Swiss watch
factory? And turns out you cannot.
And and so in a lot of ways, this was, you know, kind of
inevitable where we ran into this problem where you see it
today. There are massive multi year
backlogs on these foreign military sales deals out of the
US and a lot of allies and partners have seen worldwide are
(05:19):
feeling like the urgency to rearm, the urgency to build
capacity is very real, but the availability of of actual
systems they can acquire quicklyis just not there, right.
Like you got to wait four or five years to get a lot of these
capabilities. It's a huge problem.
And talking with folks in the commercial world, you know,
learning lessons from automotive, consumer
electronics, we've pulled in a really interesting blend of
(05:40):
people from both traditional defence as well as a lot of the
commercial space. The reality is how mass
producible something is really set when you design the systems.
And if you design it with these commercial processes and you
find smart ways to get, you know, nearly identical
capability, but at a much more producible process, you can
really scale quickly. You can lean into the automotive
(06:03):
and industrial suppliers that doexist.
You can take advantage of composites and metallic
providers, all these different things that that really do.
The capacity is already there. You just got to be able to
design your systems to take advantage of it.
And when you look at like World War 2, the US didn't, you know,
really was able to scale up very, very quickly.
But it was because when you design these systems, they took
(06:26):
advantage of what we could already make.
Like that was the key, right? It wasn't, well, here's this is
super exotic material that is unmanufacturable.
And I that's the linchpin, right?
And, and the reality with manufacturing is you can have
1000 parts on a thing and if onedoesn't show up, you're still
not producing. So it's always these relatively
few defence exotic things that you really got to solve for, but
really taking advantage of designing those out, getting to
(06:48):
commercial supply chain and really being able to scale
quickly, that's that's the formula we've really found has
worked well for us. My to add a personal touch, my
grandfather on my dad's side in Lancashire in northwest England
worked in a fireworks factory atthe start of World War 2.
So you can imagine what he started doing in a very quick,
(07:09):
quick pivot. Now, I mean, what what you're
talking about makes a lot of sense during a fight and a
crisis. It also has deterrence
advantages, of course, because if an adversary can see that
this is what you're able to do, then they have to know that if
they get into a fight with you, then it's going to be over a
longer term. I'm interested in some of the, I
mean, the, the real life examples that we're seeing
(07:30):
unfold at the moment and Ukrainebeing the most obvious one.
I mean, they've reorientated a bunch of their capacity, their,
you know, national capacity to make drones and that's been
decisive in keeping them competitive for 3 1/2 years.
In fact, I'd I'd go so far as tosay there's a real question
about whether Putin would have actually invaded if in 2022 full
(07:53):
scale invasion if he had known that it was going to go the way
it's gone. The same time we've got Russia
sending cheap drones over the border into NATO countries,
forcing them to deploy very expensive F30 fives to match
very, very cheap drones, which is not sustainable over the
longer term. So it's an interesting
demonstration of the way in which sustainable management of
security challenges is actually going to be essential, including
(08:16):
to deterrence. So what are you sort of learning
and taking away from these examples that are going on
around the world at the moment? Yeah, so I think if history is
any indicator conflicts, most ofthese conflicts when you look at
state on state conflicts are protracted wars, right.
Like the reality is industrial capacity really matters.
Again, you know World War Two example, you know the US tanks
(08:37):
were worse than the German tanksjust could out produce them by a
massive quantity. And so the the industrial
capacity of this does really matter.
And when you look at a lot of the, you know, sort of war plans
and, and the way kind of the different services think about
these conflicts is the goal is to have it over in two weeks,
but very few conflicts are over in two weeks.
(08:57):
So the question of, you know, what sort of happens after that
phase is really critical. And this deterrence is always
this capability. And well, and I think the, you
know, Western allies have just this immense capacity to deliver
very decisive, very strong kind of initial salvos.
(09:17):
It's like just sort of the best in the world.
There's no question we can kind of do that very, very quickly.
But I think there is a legitimate question of how long
can it be sustained If you're China, if you're Russia, that's
a legitimate question, right? So, you know, Ukraine's now
producing something like 4 million drones a year.
I think Russia's at like 5 DoD official estimated that from the
(09:38):
industrial base, they believe they can get to 100,000 drones a
year. Seems like a problem, right?
And so the this this issue is, is very severe where, you know,
most of the war gaming shows a lot of the key munitions are out
in something like 8 to 9 days like it is, it is a real
problem. And you know, why do most people
start wars? Well, there's a lot of reasons,
but I think the biggest one is they thought they were going to
(09:59):
win. And if you can make it a little
more clear that we have the the will to fight back, that we have
the capacity to do so and not just the sheer, you know, volume
of force we could bring to the fight, but the sustained
political and industrial capacity to keep that fight up
well in. Fact.
I'd refine it to win at a politically acceptable cost.
(10:22):
That's right. Yeah, exactly.
And so, so there's the politicaldimension of sustainability on
this as well. And, you know, I think there's
been a lot of, you know, historically the, the sort of
free market view of, of, of economics, you know, in, in free
trade has, has been the dominantforce.
But I think people are waking upto the realisation that
(10:44):
countries like China have intentionally manipulated that
free trade system to create strategic strangleholds on the
supply chain, right. And so this, you know, the
obvious example everyone points to is exercising the restriction
on rare earth magnets and the massive impact that has had the
US utilising, you know, high endGPU export controls.
(11:05):
Like all of these are instruments of, you know,
national security that are becoming really important.
But the West has not had a practise in thinking through
this industrial policy question,right?
Which is, you know, I need to not just say free trade is the
ultimate goal, but a degree of sovereignty, control and
assurance that I have not just for defence purposes, but
(11:25):
commercial purposes as well. The ability to deleverage any of
these adversaries I'm dealing with is, is incredibly
important. And so I think we're defence
when we look at this, there's a lot of things that we can do
directly, right? Like we can tap into these
commercial supply chains, we canlook for alternative supply, we
can have a much more scalable approach that creates a lot of
(11:45):
benefit. You then have a set of things
that defence alone cannot solve.The volume does not exist, the
demand does not exist in a sufficient quality quantity to
solve rare earth magnets, to solve lithium production, to
solve germanium, right. Like all these things are, are
are quite critical. And so I think there's got to be
kind of a whole of, of government approach in all of,
(12:06):
you know, kind of the allied countries really around this
question of how do we mutually assure that the supplies we need
will exist, particularly upstream as you look into the
supply chains. And that in a time of crisis, we
are not going to be collapsed economically, that we we have a
more resilient, you know, kind of holistic strategy to this.
Yeah, Yeah. I mean, it's interesting to be
(12:27):
even thinking about these thingsfrom the point of view of a
defence company. You know, you, you, one might
think that as a commercial entity your job is to just
provide, provide, you know, capabilities or you know,
provide particular platforms, provide ammunition, whatever it
might be. I mean, just the fact that
you're thinking about these sorts of things as, you know,
very upstream raw materials in that case, like critical
(12:50):
minerals, I mean, I just reflecton that a little bit more for
me. Yeah.
I mean, I think, you know, our, our sort of view is 1, there's
just straight like kind of commercial interest in this,
which is we've got to be able toproduce these systems.
And so if that means we've got to go stockpile germanium or go
strike deals with people doing mining and, you know, refining
operations, hey, we'll do it if that's the, the right strategy
(13:11):
to take. But I, I think there's a broader
responsibility that we're typically never asked for, which
is a question of how would you be resilient in a time of a
crisis? And that is not something that
is ever written into a contract.Nobody knows how to really
measure that. Even the data to get at that
question is very opaque, right? And the supply chains
historically have been incredibly deep.
(13:32):
So our strategy on this is, you know, if the historic strategy
is I buy, you know, kind of my avionics module and my control
module from these vendors. And then they have been sub
vendors and sub vendors and somevendors you might go back 567
layers deep before you start getting to this question of, you
know, a short supply and raw materials.
It lengthens your lead time and makes scaling much harder and
(13:55):
you have a really critical problem.
What we've kind of tried to do is pull in more of that design
so that we can work with a broader base of suppliers that
are a little bit closer to the problem.
And we can start to get a littlemore transparency around this
question of what is actually scalable for both our own
commercial purpose. How fast can we ramp, you know,
how quickly can we respond to these things.
We don't want two to three year lead times on this, We want
(14:17):
three to six month lead times. So that becomes really critical.
But then this, this question of wartime, I think is an
obligation that every defence provider should be really
looking at, right? Like if, if we do not understand
the realities of what will it take to sustain this industrial
scale during conflict time, well, whose responsibility is it
(14:38):
then, right. And, and if we have the best
first hand knowledge, then we should be obligated to do this.
Now nobody asks us for this. I would love to see it in
contracts. I think it would be something
that would be very healthy. I don't know what I would write
to put that contractual clause in, but it it's a, it's a really
important concept. But I mean, this idea of taking
responsibility is a really interesting one.
(14:59):
I mean, I, I think something like supply chains, for
instance, they are inherently geopolitical and, and the the
recent rare earths deal between Australia and the US is a
classic example. We are working together to deny
China its ongoing stranglehold over something that is clearly
being willing to weaponize. So what you're what you're
talking about there ties in quite well to me with this idea
(15:21):
of almost sort of, I don't want to say patriotic industry, but
industry that does actually takea position in geopolitics.
Now, you know, obviously U.S. companies during the Cold War,
you know, worked with the US government very closely.
But what I mean, one thing that I, at the risk of sounding
ingratiating here that I do admire about Andrew, is that
(15:42):
you're willing to say, OK, we are here for liberal
democracies. We are here for, you know, to
enable them to provide security.We we are actually taking a
position on these things. Does that just work from a
commercial point of view, or do you find yourself having to
balance one thing against another in certain contexts?
Or or does it just a line of of its own accord?
(16:05):
Well, I, I think there's, there's a lot of nuance to these
things. But yes, I think that is so
we've, we've taken a very clear position that, look, we're
aligned to Western allies. We have a clear set of
principles and we are not going to do things outside of the
policies that the governments are looking for.
And I think that's our our sort of division on this is you don't
(16:28):
want, you know, kind of tech Bros setting policy on these
things. It seems like a very bad
strategy for the world. You want tech companies to
actually be accountable to saying, here's the best
information I can provide, Here's the trade offs and, and
you need the government accountable to actually making
policy decisions on these things.
(16:48):
And so we tried to be good partners in that, you know, and,
and really look at creative waysthat we can work within the
legal and regulatory frameworks to still create a set of policy
options. So one of these has been really
key to us is looking at kind of this question of sovereignty
export, you know, and how do we think about what's aligned to,
(17:08):
you know, the traditional approach rates buy through the
US government, it's export controlled.
That's it, right? And, you know, sometimes there's
some localization and offsets and production, but I think
there's a lot more creative strategies to this where, you
know, maybe we can take some of the technology that's built in
the US. Maybe not all of the most
sensitive pieces need to be exported because other countries
also have great investments in unique radar capabilities or,
(17:31):
you know, unique, you know, warhead capabilities or things
like that, that that may be justmuch more specialised sensitive
to where the US believes they need control.
But we can find win win strategies, right where we can
accelerate allies and partners on what they need to produce,
what they need to build up and give, you know, kind of US,
Australia, any of the countries we work with closely out of the
(17:53):
security assurances for their own military that they need to
have. And so there's, there's these
kind of very straightforward, you know, but require some
creativity strategies you can have around ways to get at this
first principle problem of how do we enable allies and partners
while still protecting the sovereign interests of every
country we operate in. And I just don't think you can
be a defence company and not have a view on these
(18:16):
geopolitical questions. It is the, you know, I think
everyone was allowed to forget for about 20 years that this
mattered when it was just so clearly a hegemony on, on, on
kind of international order. That's just not true anymore.
Every country, kind of every, every company needs to take a
bit of a side. And we've seen this in the tech
community where for a long time in the finance community as
(18:38):
well, they were so thoroughly compromised by the promise of
returns in China that nobody in the national security community
ever thought was going to materialise.
It's like, what does it mean to get a high return in China?
Are you ever going to get your money out?
What does it mean for your investment to go to the
literally 0? Because, you know, the autocrat
decided today that he doesn't want that industry or that money
is not coming out. And it was a big wake up call to
(19:00):
a lot of these guys who had justhuge interest in China and then
realising that that was not someone you could trust, that
was not a good bet to make. And that they kind of took for
granted the the rules of order that the, you know, the kind of
liberal democracies shared. So we've seen a huge change in
just even the tech community over the last few years where,
you know, when we started, I remember there was massive
(19:23):
protests and concerns that Google was working with the US
government on AI. Now last year, I think nearly
every tech companies like openlycommitted to supporting AI with
the, with the government like that is a huge change.
And, you know, it really, the world has really shifted in the
last few years where this is notoptional anymore.
(19:46):
You actually have to really think through these kind of key
strategic questions. The government's getting more
sophisticated, the the companiesare feeling more responsible and
there's just a much greater alignment around, you know, not
taking for granted that the the countries that have provided so
much peace and prosperity. Interesting.
I mean, the, you're, yes, you'rea defence company, but the, you
(20:07):
know, the distinction between a defence company and a tech
company is much less clear than it, it might have always been.
I'm just interested in how that that transition where the tech
company, where the tech industryhas started to I suppose come a
little bit closer to where companies like Andrew and
Palantir have been for some years now, what that experience
(20:28):
has been like from Andrew point of view.
I mean, you presumably feel somewhat vindicated by that.
Yeah, When we started, it was, you know, when we were going out
in 2017, remember, it was Trump won.
There was a massive backlash against doing anything with the
federal government. And our sort of view is been
(20:50):
that this is just, this is a bipartisan issue, this matters.
And the system is much worse than anyone sort of realised.
And the world's a much more dangerous place than anyone
realised. And I've been a lifeline
Democrat. A lot of the Co founders said as
well. We've had a very balanced kind
of partisan view. But everyone agrees on the
mission of what we're trying to do.
It's like best technology to ourallies and partners in the US
(21:13):
and our war fighters. Like that's what we need, right?
Like no question. And you know, also it was this
period where the sort of employee revolts over working on
anything controversial is massive.
So to start with a company that was working on, you know,
defence and we had investors saying commit to never making
weapons. Of course we're not going to
(21:34):
commit to that. Like we're a defence company.
Like this is an important aspectof it.
But having that clear vision of like, this is what we are going
to do. Do not come here if you do not
support this mission. But here is what we are doing
and we're clear with it. We have communicated it clearly,
publicly, transparently with ouremployees, with the media, with
(21:55):
government, and that that has worked really well.
So I think this sort of was a huge lack of sort of
authenticity and transparency inthese things.
And then it really what, you know, I think Vladimir Putin and
G have done an amazing job of convincing the rest of the world
that this actually matters, right?
They've really made it very clear.
Have you ever found it difficultto get people as a result of
(22:15):
that, you know, stated position?Especially when the demand for
highly qualified tech people is so high.
You know, it honestly, not really because I think the,
there was such a desire for likeclear, authentic views of, of
what your mission was at a time when there was just this sort of
very crazy, you know, view of, of, of your relationship to the
(22:39):
companies and the mission on it.So even when it was more
controversial, you know, kind ofpre Ukraine invasion, even then
there's the clarity of what we were doing, I think resonated
with a lot of people. This this idea that people in
tech don't want to work on defence, I think is a very
minority, but loud minority position.
And I think the vast majority are in favour of a strong
(23:00):
National Defence. Like it's, it's kind of the, the
underpinning of safety and security we've had for a
century. So the I never found it that
challenging, right? And, you know, it's I actually
think it was like almost better when we were more controversial,
like, you know, the when, when too many MBA show up because
(23:22):
you're very cool, like you should really watch out.
Like it's, it's you might have jumped the shark at that point.
And so like when you're cool, you're not sure if people are
there for the right reason, Like, are you there because this
is a great stop in your career and it's a popular company, or
are you there because you actually believe in it?
And we want people that are actually, you know, kind of
thoroughly believe in the mission of it because it can
really corrupt you very quickly otherwise.
(23:43):
And so that's that's something we still philtre for still, you
know, very transparent about what we do for that reason.
And, and I'm I'm not sure being a popular company is actually
that much of an advantage in reality, right?
All right, Speaking of sharks jumping or otherwise ghost
shark, let's get on to that. The underwater UIV that you're
building in collaboration with Australia.
(24:06):
I mean, interested in a couple of things.
One is that it to me sort of it epitomises this, this issue
we've got about fast capability acquisition at the moment.
So you know, Australia is going to have fantastic submarines in
the 20 forties, right. But you know, there is a window
(24:27):
of danger that we're in now. Some people, you know, some
China experts say that, you know, 27 to 30 really is
potentially that sort of peak period when Xi Jinping might
say, OK, this is we're not goingto get any stronger than we are
now relative to other powers. So certainly, you know, quick
capability acquisition is something like that.
A country like Australia really,really needs to think about
given that we've got these more exquisite capabilities coming
(24:50):
online somewhat in the in the future.
I'm interested in your views on on that and how Ghost Shark sort
of fits into that. But also just in broad terms, as
much as you're able to in a, in a, you know, obviously an
unclassified podcast, give us your thoughts on why Ghost Shark
is useful for Australia, given its geography, given its
(25:10):
strategic challenges and so forth.
Yeah. So the the Ghost Shark programme
was for us as a company we really look for forward leaning
customers where there's this combination of creative military
thinking. So you know, how are you going
(25:30):
to solve these critical operational challenges in
different ways with a willingness to move fast and try
different models on acquisition capability development and
really experiment with more modern approaches.
And with Ghost Shark, we really found that with with the Rand
and ASCA and it was it was fantastic where you know, I
think we went very, very fast period from first conversations
(25:54):
through them saying this is a key capability, let's move out.
We entered a Co development programme, which is one of the
first we'd ever done. We're just sort of risk sharing
approach, very open on thinking through kind of governance
structures and how do we, you know, kind of break out of the
standard mould of destroying these schedules and costs with,
(26:14):
you know, obsolete requirements and, you know, over overburdened
regulation. But really just trying to trim
to what's necessary for the capability we're trying to
deliver. And that sort of creativity and
willingness to, to break the mould is, is, is rare.
I was really impressed with how this has gone throughout the
programme where, you know, I've seen this go so badly in US and
(26:37):
plenty of other places where, you know, kind of that risk
aversion, you know, this is the way we've always done it, you
know, coming in with, you know, man, submarine requirements on
to uncrewed capability. That's, that's a, that can just
destroy a programme, right? Like can destroy a capability
and so it required a degree of flexibility around kind of all
(26:59):
aspects and willingness to re examine all aspects of how do
you think about building, testing, deploying, operating
these capabilities. And so these are never just like
an acquisition question, right? Like I I think the the this
often gets, you know, kind of viewed as does the government
have the authority to acquire differently?
That's one of the five dimensions that need to be true,
(27:22):
right? It's, it's so many parts of
building and deploying these things quickly that you have to
solve simultaneously. And, and this is really key to,
you know, how are you going to actually get the mass and
capability in a rapid way, right.
So you're going to buy some things off the shelf, but it's
got the backlog problem that we've already talked about for
things that are in production aswell as getting new capabilities
(27:44):
through that traditional pipeline.
You know, it's gotten longer andlonger and longer.
There's there's really no way out of it.
And this is where I think this class of autonomous and lower
cost systems really can shine where the cost to develop is not
prohibitive. The cost to purchase is much
more affordable and then the ability to project, you know,
(28:05):
kind of power at scale is, is really, really compelling,
right. So we take kind of go shark and
what capability can this provideto Australia?
You know, obviously Australia issurrounded by ocean, right?
Like the, the maritime side has to be solved, the naval side has
to be solved. And the sheer volume of of
coastline and you know, kind of key choke points and straights
(28:28):
that you have to control and protect is is massive, right?
And at reasonable ranges, crude submarines are incredible,
right? Like I am a huge fan of Virginia
class. I think it is the most
impressive capable weapons programme ever done.
Like it is just a phenomenal capability.
I have nothing but good things to say about it.
But it is also the most complex weapon system ever developed,
(28:51):
which is, you know, part of the reason it's so spectacular.
And and you're just inherently limited by the scale that you
can actually deploy these app. And so as we look at where these
uncrewed systems can really shine, it really is about scale,
right? It is about the number of places
that you can be positioned, the amount of effects and
surveillance that you can acquire in the amount of
(29:13):
dilemmas you can create for youradversaries, where, you know, if
you can put dozens of these systems into the ocean, that
creates a lot of uncertainty foryour adversary around, you know,
particularly subsea where it's very hard to find things.
It is very hard to, you know, kind of mitigate these
capabilities. And it can create a massive
(29:33):
amount of uncertainty for an adversary trying to approach the
coasts, trying to conduct a manoeuvre where you can really
hold them at risk without reallya lot of, you know, capital
assets or sailors lives at risk.And that is a massive capability
enabler that I think the ran really understood and leaned
into as you know, as this reallythis bridge till you can get
(29:58):
Virginia class to where you wantit to be.
But even as an augmentation of just the sheer volume that
you're going to be need to deploy in any fight.
Let's broaden it out a little bit on on autonomy, autonomy
automation on crude systems. I mean, is there I, I, I, I see
(30:19):
and I pardon my ignorance, I should know this off the top of
my head better. But the way it would, you know,
there's something like the ghostshark could team with a Virginia
class, for instance. I mean that that just seems like
a natural use of it, a bit like the way air platforms we're
talking about doing in a similarway.
Just talk about human teaming a little bit, human AI teaming or
(30:41):
human autonomous teaming a little bit into the future.
I'm it seems to me quite possible that teaming is a
transition until automation and AI get to where, you know, the
role for human beings is simply,you know, our, our value
contribution starts to shrink tothe point of negligible.
(31:01):
Do you see it that way, or do you see the human role as being
sustainable forever? So I fundamentally this is a
question of sort of values and, and, and how you're going to
hold people accountable to what military systems do, right.
And so fundamentally to me, thisis the West will always want
(31:24):
human accountability around lethal decision making,
deployment of effects, right? Like all these things that are,
you know, have real political and societal consequence, we're
going to want human accountability around it.
We're never going to get to a world where we say that's
absolved. The AI told me to do it and it's
not my responsibility. And everything we've designed is
(31:45):
really about thinking that question of where is the right
point for humans to engage and how does that further enhance
human accountability, not degrade it, Right.
So the AI and the autonomy can be phenomenal doing a lot of the
more mechanical tasks. You know, I don't need a human
to plot the route that a missileflies or a submarine goes.
That is a perfectly suitable task for software to do, for AI
(32:10):
to do this. AI can recommend courses of
action and say, hey, I think youshould do AB and C.
And a human can approve that andunderstand the context and the
consequences of those decisions in a way that's very hard to
digitise. And so I think there will always
be this role for human accountability in it, but it's
always a matter of degree on these things where in a lot of
(32:31):
ways there have been lethal autonomous weapons for decades
now. The way a modern torpedo works,
you give it a box and you say anything that matches this
acoustic signature in this box that is a legitimate military
target. That sounds a lot to me like a
autonomous weapon. And so in a lot of ways, what
we're debating is really then these controls around what
(32:54):
context do we allow the machine to make these decisions?
What are the rules of engagementand how much freedom and
latitude do you provide it, right.
And then where does the human decision making and human
accountability really come in? You know, a lot of the Western
systems, whoever launched the weapon is accountable for the
results of what that weapon does.
I think it's a good system. We should probably keep that.
(33:15):
And so the so there will always be that part, right?
And so I that's how we really think about it.
Yeah, there's the stepping stonedimensions of how sophisticated
is the AI, how much control can it really do.
I think that is true. But I think the fundamental
thing is really around this question of values and how do
you want to provide accountability into the system.
(33:37):
But it's largely the values and ethics question rather than the
capability question. I think that that is the long
term question is values and ethics.
I think the near term question is kind of 1 of adoption, right?
So even if the technology was perfect to solve all these
problems, I think you are going to inherently see correctly the
desire to have a relatively short leash to start, right,
(33:58):
relatively prescriptive mission sets.
And then over time as trust is built and we understand what the
capabilities are, then you will release kind of more control to
these systems to perform more autonomous papers.
We've seen this with nearly everything we've deployed, even
counter drone systems. When it starts, typically the
operators want to look at, you know, moving the camera
(34:18):
themselves. Then they're OK now I'll let the
system move the camera. OK.
Now I'll recommend engagement patterns.
So you you kind of see a progression naturally as people
build more trust in the system of what that right balance is.
OK, just back on Ghost Shark fora moment.
It's described by Andrew as a sovereign undersea capability.
Just understand, explain to me what sovereign means in that
(34:40):
context. Does it mean that Australia can
do everything on its own at somestage in the force?
Well, for the foreseeable future, potentially including
one in which you know something happens in the Australia US
alliance, for instance, probablynot a hopefully not a political
collapse, but something else might just mean that Australia
needs to do it on its own for some defined.
(35:01):
Period. Yeah.
So for us, we the team that designed Go Shark is based in
Australia, it's Australian engineers.
We have tight collaborations with the US counterparts, but
on, but all of the core design expertise resides in Australia.
The production capacity is basedin Australia.
(35:23):
We can fully assemble these go sharks here.
And the supply chain, you know, there's a degree of you know
kind of AUCUS components to this, there's some UK suppliers
and things like that. But you know, the vast majority
of the metallic components, the composites, the battery systems,
you know, core electronics, assembly, all of those pieces,
we've built out an Australian supply chain around this.
(35:46):
So the ability to actually produce independently of really
nearly any other country is, is quite high.
And so the so for us, sovereign means kind of that independent
ability, a, a mindful understanding of where those
interdependencies exist and something kind of acknowledged
by the government of Yep, this looks correct in terms of the
(36:08):
interdependencies I want to have.
And so for us, it's very much about creating the ability for
Australia to act quite independently as needed.
You know, there's several payloads that I don't get to be
brief to. I'm not an Australian, so that's
perfectly fine with me. I think that's a good thing.
Good, good. All right.
One thing we haven't talked about yet is Lattice, which is
(36:30):
really the core of your company.That's the the the the software
system that there's essentially a command and control system.
Just talk about how that forms the core and how you sort of
build around it. And interestingly, I mean,
we'll, we'll ghost shock. Will ghost shark employ lattice?
(36:50):
Is it, is it? I don't think lattice is just
for Andrew capabilities, correct?
You can use it for other things as well.
But just talk a little bit aboutthe role of lattice.
So stepping back slightly, the the way modern software has been
built in Silicon Valley is this idea, these very general purpose
platforms and the these are a lot harder to build with quality
than most people would appreciate.
(37:12):
You have to build immense tooling around your development
infrastructure and simulation. How do I test my software?
I, you know, I have to simulate sensors and autonomous
behaviours and all these networkconditions, all these different
pieces. You know, I have to build all
this tooling and instrumentationfor monitoring and deploying all
this controlling configuration. Just a, an enormous amount of
(37:32):
software you end up having to build.
And that is below sort of the tip of the iceberg, which was
what people, you know, kind of think of with these, which is I
get a drone to go from point A to point B into a mission.
So the complexity of these modern software stacks is just
immense. And so really every major tech
company has thought about this is how do I share that load
across a number of different capabilities and build these
(37:56):
kind of very flexible software platforms that allows me to
tailor to any, any application. And when we look at the the span
of problems we're trying to solve, it's really this question
of how do I build for a world where I have much larger numbers
of sensors on the battlefield, of drones on the battlefield,
subsea in the air, and I'm pairing those with manned
(38:16):
systems. So solving all those fundamental
problems. And that's how do I think about
networking these. How do I use AI to process that
information at the edge? What's the right way to move
that information around to create this holistic picture of
a battle space? How do I express an intent to a
robot? Like what does that mean?
And how do I represent that in adigital way that can be
(38:37):
transmitted over a network? You know, how does a human want
to interact with us, right? What do I want to see?
What decisions do I need to make?
How do I make sense of a increasingly complex
battlefield? So we have this view to where a
lot of these future battlefieldsgo.
Much more systems, much, much more disaggregated.
Everything is networked, but youhave constant jamming.
(38:59):
So how do I have resilience baked into that?
And I need to create this transparent picture as best as I
can for all the information I can get.
And that at the core is what we've tried to build Lattice to
solve. Now, obviously making a subsea
drone is very different than making a you know, a small
quadcopter is different than making a fighter jet, and it's
very different than talking to amanned F35 S the literal
(39:21):
components. Yeah, those vary from project to
project, right. And but that core infrastructure
around, you know, how do I network this represent
information and think about managing it that is shared
across. And so Ghost Shark, nearly
everything we do has these kind of core lattice pieces into it.
And the greatest part of this isbecause we've shared this code
(39:42):
base across so many systems, allthose learnings fold back in and
we can provide these updates to every system as it gets smarter,
better. And we take all those learnings
and distribute it across the whole fleet of all the
capabilities we're building. I mean, I find the idea of eyes,
ears and brains to be a useful analogy just to sort of more
instinctively grasp it. I mean if I had 10,000 eyes and
(40:06):
5000 ears. That might sound great, but if
I've got one brain, unless it's supercharged in some way with
AI, which may come down the track.
And I suppose, I mean, actually I'm, I'm let's talk about that
for a moment. What's what's the role of AI in
Lattice and what's the role of AI going forward?
Do you think in managing all of that sensor data, all of that
sort of decision making complexity, which is going to
(40:29):
which is already already exists for, for, for war fighters, but
is going to become more and more, I suppose potentially
overwhelming unless we we manageit well with AI.
Just tell us a little bit about the role of AI there.
Yeah, so I kind of think I always like to simplify these
problems down a little bit. So like, what are you really
trying to do for the vast majority of kind of military
(40:50):
applications for what you're trying to do with the sensors?
You're trying to figure out where your friends and allies
are. So where's the blue force?
And you're trying to figure out where everything else is in the
environment. So what's the map?
Where's the bad guys? What are they doing and how do I
find them? So when you build that, boil
that down, it's this question of, you know, how do I interpret
sensor data to know, hey, is this a ship?
(41:10):
Is this a tank? What am I actually looking at?
And then you can kind of then think of relatively simple ways
of moving this information around.
So a lot of what I think we've done well is think about these
relatively simple abstractions, how to do it.
These are the AII think the we've been doing what's called
the prior generation of of AI where this was the computer
(41:31):
vision using machine learning for detecting signals of
interest in, you know the electromagnetic spectrum, what I
think we would call now kind of more discriminative AI.
We've been doing that since the founding right so we deployed
border security towers and counter drone systems that have
been using computer vision and sensor fusion, you know for
eight years now and it's proven to work incredibly well right.
(41:52):
We probably the largest number of deployed production, you
know, AI systems out there. The and then now there's more of
this question of generative AI and these more reasoning systems
and what implications does that have?
And I'm extremely optimistic around this.
So, you know, I, I've been really fascinated by things like
(42:12):
alpha go alpha star. A lot of the things that, you
know, kind of came out of GoogleDeepMind around the ability to
learn these strategy games. And when we think about, you
know, fighter jet tactics or, you know, how would you have a
tech helicopter or anything, it is so there's so many variables
these pilots process. It's so hard to code into like,
you know, a, a particular algorithm and the ability for
(42:36):
these machines learned kind of generative strategies to outpace
human level performance or certainly any of these
traditional approaches is, is massive.
And we've seen this in some of the research we've been doing in
house with just kind of these reinforcement learning
approaches, just massively within like 4 weeks of work can
massively outpace any of the alternatives we've tried.
(42:56):
And so I think the ability to get to these superhuman levels
of performance on, you know, kind of these tactical
behaviours is very, very near term.
What I don't think works in practise is sort of this idea
that we're going to have this massive central brain kind of
puppeteering this large scale operation.
It's not resilient. I don't even think it works with
(43:19):
the laws of physics with how much information you'd have to
move over networks to get that data back.
I just, it's no amount of star link that's going to solve that,
just too much data to move. And you've created this
incredibly not resilient system all of a sudden.
And so when we think about this,it's really about how do I
create these layered intelligentsystems that I understand how
(43:40):
they operate in concert, but they're each making smart,
independent decisions. That's a resilient strategy.
And it also, by the way, is how the Western militaries are
organised. They figured out this mission
command structure is the right approach, give a degree of
autonomy and discretion down to your lower echelons.
You get phenomenal results, right?
And you kind of train for that. You, you operate that way.
(44:03):
You, you let them have initiative and that tends to
deliver much better results. So we kind of mirror that and
how we think about the technology perspective, which is
the right answer probably looks like something that is more
resilient, more distributed, more at the edge than this sort
of mythical central brain planning out the whole battle
(44:23):
space. I just, I don't see it.
It doesn't make sense to me. I don't know why you would do
that. It speaks to my lack of world,
real world experience either in engineering or war fighting that
I my head immediately starts going to a, you know, the the
human president or Prime Minister with one button, you
(44:43):
know, shall I win the war? You know, yes, slash no.
Which one do you want to click? That's where I see it eventually
sort of graduating towards. But you're doing a good job of
dissuading me that that that's going to be the case.
And rather it's to be down at more of a, you know, a lot of
decisions contributing in in a tactical, operational and even
(45:06):
strategic levels. That's correct.
That's right. And I think the, you know, you,
maybe your systems are implemented like so the way I'm
saying it is really about thinking of how the, you know,
future robot armies and these command and control different
echelons will work. I think it has to mimic the
structures that we have, that we've learned over a century of
what is an effective way to organise and operate a military.
(45:28):
I think those are sort of timeless principles that likely
are going to transcend the specifics of the technology.
Now maybe the degree to which that a human at that echelon,
the effect they can have will bemassively amplified, right.
So, you know, 1 pilot might be able to control 100 droughts.
Well, that is a massive increasein the amount of capability that
(45:49):
we can get at every echelon. And that is a a huge change that
technology can bring forward, but it doesn't remove the need
for that initiative, that creativity and that sort of
mission command structure at every echelon.
The other side of this, though, is what you can start to do with
simulation to really understand what is possible, what are the
likely outcomes of any given conflict, what strategies
(46:09):
actually work. There's sort of this game
theoretic notion or thinking about capability development
that we've really leaned into around we, the way we operate is
when we think of a new capability, for example, we
will. So here's the war fighting
problem I'm trying to solve. We'll run dozens of different
strategies of technology. And if we had this range or this
(46:30):
capability or it flew this fast,what if it was on the land?
What if it was in the air? And we can simulate those and
you can say, hey, here's how these potential conflicts and
these engagements would go. So the state of the art today is
reasonably complex. You can kind of get a, you know,
a, a fairly complex engagement and simulate that out.
But there's a huge amount of potential with generative AI to
(46:52):
very rapidly lay down super complex scenarios over a huge
number of variables. And you can get a much more
comprehensive idea of which are going to be the dominant
strategies and how likely am I to prevail with which strategies
and how to create those turn effects.
Now the, the hard part with thisis so many of these decisions
(47:13):
are really in the brain of one human at the end of the day
right of will China invade Taiwan?
Well, you can do all the military analysis you want, but
ultimately GE will have his reasons for doing it or not
doing it based on his perceptions of likelihood of
success, the political reality around him and what he needs to
do. You know, same with Putin right?
And the ability to use AI to forecast that humans brain.
(47:36):
That doesn't seem like it's going to happen anytime soon.
And so I think that the goal we think of with these things is
really around, you know, what are we trying to construct as a
company? We're trying to provide that
capability dimension in that deterrence equation.
And if you can make the outcome of these conflicts quite clear,
AI probably helps you can make these things much more obvious.
(47:58):
That is a good thing, right? Like that is what we were trying
to do. So that conflict does not
happen. And and that's ultimately what I
think our our goals and ambitions are on these things.
It's a great message to end withand and wrap up.
Brian, thanks so much for comingon Stop the World.
It's been a great chat. Thank you.
Appreciate it. Thank you.
Thanks for listening to this special episode of Stop the
(48:20):
World, the TSD Summit Sessions. We'll be back with another
episode next week.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Ruthie's Table 4
For more than 30 years The River Cafe in London, has been the home-from-home of artists, architects, designers, actors, collectors, writers, activists, and politicians. Michael Caine, Glenn Close, JJ Abrams, Steve McQueen, Victoria and David Beckham, and Lily Allen, are just some of the people who love to call The River Cafe home. On River Cafe Table 4, Rogers sits down with her customers—who have become friends—to talk about food memories. Table 4 explores how food impacts every aspect of our lives. “Foods is politics, food is cultural, food is how you express love, food is about your heritage, it defines who you and who you want to be,” says Rogers. Each week, Rogers invites her guest to reminisce about family suppers and first dates, what they cook, how they eat when performing, the restaurants they choose, and what food they seek when they need comfort. And to punctuate each episode of Table 4, guests such as Ralph Fiennes, Emily Blunt, and Alfonso Cuarón, read their favourite recipe from one of the best-selling River Cafe cookbooks. Table 4 itself, is situated near The River Cafe’s open kitchen, close to the bright pink wood-fired oven and next to the glossy yellow pass, where Ruthie oversees the restaurant. You are invited to take a seat at this intimate table and join the conversation. For more information, recipes, and ingredients, go to https://shoptherivercafe.co.uk/ Web: https://rivercafe.co.uk/ Instagram: www.instagram.com/therivercafelondon/ Facebook: https://en-gb.facebook.com/therivercafelondon/ For more podcasts from iHeartRadio, visit the iheartradio app, apple podcasts, or wherever you listen to your favorite shows. Learn more about your ad-choices at https://www.iheartpodcastnetwork.com
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com