Palmer Luckey: Autonomous Weapons Are Ancient and Why Anduril Won't Build Humanoids | EP #295 — Transcript
Full transcript
- 0:05Wow, it's so much brighter out here.
- 0:07>> It is. It's, you know, the future is
- 0:09bright.
- 0:10>> There's lens flare in the future.
- 0:12>> Yeah.
- 0:13So, welcome everybody to the Oscars of
- 0:15optimism. That, of course, we're opening
- 0:18up the summit with the king of
- 0:20AUTONOMOUS WEAPON SYSTEMS.
- 0:22>> [laughter]
- 0:26>> ALL RIGHT. SO,
- 0:29I'M GOING TO start with the fact that
- 0:31we're awarding three X prizes this week.
- 0:34Palmer, congratulations for winning our
- 0:37wildfire X prize.
- 0:39>> Thank you so much.
- 0:41>> [applause]
- 0:42>> This one's been a long time in in the
- 0:44works.
- 0:45>> Let's get Let's get the the the slide up
- 0:47here. So, Palmer won $2.2 million for
- 0:50Anduril.
- 0:51>> As if you needed it.
- 0:52>> Um a million of
- 0:54a million of that from Lockheed,
- 0:57which was kind of interesting.
- 1:00>> We do a lot of work with Lockheed. We
- 1:01also compete with them, but we
- 1:04in the in the defense space, it's you
- 1:06know, we we we compete with their
- 1:07fighter jet division and we partner with
- 1:09them on some of their payloads. And so,
- 1:11we do a lot of stuff together. But, you
- 1:12me and you have both been trying to get
- 1:13this autonomous firefighting stuff going
- 1:15for a long time.
- 1:16>> mean, you got this prize announced back
- 1:18in 2019. It's been a long time.
- 1:20Anduril's one of our first two products
- 1:22was an autonomous firefighting tank that
- 1:25we started building in 2017, finished in
- 1:272019.
- 1:28>> It's insane that we let fires become
- 1:30conflagrations before we put them out.
- 1:32>> And it well, and it's and it's kind of
- 1:33insane that people are saying, "Oh, the
- 1:36you know, the technology's finally made
- 1:37it possible." And I don't know, I'm not
- 1:39going to correct them too often in
- 1:41public, but it's I mean, this has been
- 1:42possible for over a decade. It's
- 1:44>> Yeah. All right.
- 1:44>> It's just that
- 1:45it's become politically politically
- 1:47impossible to ignore the problem now.
- 1:49>> We'll watch a 1-minute video clip of
- 1:51what your team achieved. And if you walk
- 1:53us through it here.
- 1:56>> Sure. So, you
- 1:57this was
- 1:59an example of what we were doing at the
- 2:00test site where we were
- 2:02detecting fires, classifying fires. What
- 2:05you don't really see in here is that you
- 2:06also need [music] to reject false fires.
- 2:08That was part of the test as well. So,
- 2:10you have to know, you know, who's the
- 2:11guy smoking a cigarette? Who's the guy
- 2:13What's What's What's a campfire? What's
- 2:14a road flare? And then what's something
- 2:16that shows the characteristics of real
- 2:17fire? That's what you need to flag to
- 2:20people. That's what you need to flag to
- 2:22emergency services. Uh and then you need
- 2:24to match make the type of fire to the
- 2:27right asset that's in the area. And so,
- 2:29you want to basically use the most
- 2:31cost-effective thing that can respond
- 2:34and get there in time. And if
- 2:36something's growing faster, you might
- 2:37need something different. If it's in
- 2:39woods versus on a brush hill, you might
- 2:41need something different. And so, you
- 2:42know, it's it's about detection,
- 2:44classification, rejection, matchmaking,
- 2:47and then going and and hitting the
- 2:48target. And the last thing I'll say is
- 2:49that with most of what we do,
- 2:52uh our work with the military,
- 2:55you'd go through many of these same
- 2:56steps,
- 2:57but if you find a target, let's say it
- 3:00takes you, you know, an hour to an hour
- 3:03to to get to somebody versus 10 minutes,
- 3:05it usually doesn't make as much of a
- 3:08difference. When you're talking about
- 3:10fire, though, the difference between 10
- 3:12minutes and an hour response, totally
- 3:15different tools you're going to need to
- 3:16respond with. You know,
- 3:18a bomb is is is a is a bomb whether or
- 3:20not it takes a day or an hour or a
- 3:22minute to get there. Uh but when you
- 3:24when when a fire is is is is
- 3:25metastasizing, you need to pick the
- 3:27right thing that can put it out before
- 3:29it gets too large.
- 3:30>> Amazing.
- 3:31>> You know, uh
- 3:32we're going to be announcing something
- 3:33in a month's time
- 3:35uh that you're funding a new X Prize,
- 3:39and uh it's going to be secret until
- 3:41then, but we're going to tease it now.
- 3:44Uh
- 3:45another $10 million prize. Would you
- 3:47share share with us what we're going to
- 3:49do?
- 3:50>> Sure, you know, uh
- 3:52the
- 3:53you know, we we should we should keep
- 3:54the details secret cuz there's there's
- 3:56some there's some good twists, but
- 3:58suffice to say that I'm not I'm not
- 4:01interested in joining the race to push
- 4:04AI forward with my money. I think that
- 4:07the
- 4:08the artificial intelligence side of
- 4:10models is is is pretty well covered by
- 4:12plenty of people. I think we've not been
- 4:14doing enough understanding biological
- 4:16intelligence particularly beyond people
- 4:18and so this is going to be a prize
- 4:20around understanding how animals think
- 4:23and making so that we can communicate
- 4:25with them 10 times better than the
- 4:27current state of the art for the current
- 4:29smartest species and not just trained
- 4:30animals, but wild animals in their own
- 4:33in their own language.
- 4:34>> Amazing. [applause and cheering]
- 4:36>> And I think that there's actually a lot
- 4:37that we're going to be able to learn. I
- 4:38mean, if you look at large language
- 4:40models, they're kind of this, you know,
- 4:42they're they're they're based on
- 4:44humanity's way of thinking and
- 4:46communicating distilled and I think that
- 4:48there's plenty to learn from all of the
- 4:51other species that are on this planet
- 4:53and I wouldn't be surprised if it has
- 4:54applications for AI particularly AI
- 4:58where you need to be
- 4:59more efficient, you know, more power
- 5:01efficient, more time efficient,
- 5:03you know,
- 5:04a different point of view, two heads are
- 5:06better than one, whichever whatever
- 5:09metaphor you want to use.
- 5:10>> Love it. Love it. All right, I'm going
- 5:11to kick off the first official question
- 5:13here.
- 5:14P-doom, you're in the AI industry, the
- 5:16defense industry, what's your P-doom?
- 5:20>> I'm just not worried
- 5:23about AI dooming
- 5:25the I guess I'm not worried about AI
- 5:28doing the killing. I'm very very worried
- 5:30about evil people
- 5:31>> Yes.
- 5:32>> using moderately competent AI to do evil
- 5:35things.
- 5:35>> Exactly.
- 5:36>> And I am s-
- 5:39kind of frustrated by a lot of people
- 5:41who are new to the dooming hobby.
- 5:45It's It's It's sort It's sort of like
- 5:48how we had to deal with
- 5:49the war in Ukraine happens and all of a
- 5:51sudden all these people who have had the
- 5:53dumbest opinions you've ever heard about
- 5:55the military for their whole life
- 5:58and people who believed we lived at the
- 5:59end of history that we had moved past
- 6:02conflict that we lived in a world where
- 6:04you could just write a nasty letter in
- 6:06the UN forgetting that that only works
- 6:08if it's backed up by credible threat of
- 6:10violence and then all of a sudden they
- 6:11change their minds. You're seeing this
- 6:12with AI. People who have not thought
- 6:15about this problem for more than
- 6:17a couple years or maybe even just a
- 6:19couple months
- 6:21are saying oh my god this is the biggest
- 6:23threat to humanity. These are people who
- 6:25have spent almost no time actually
- 6:27analyzing what the largest threats to
- 6:29humanity are. They haven't spoken to any
- 6:31of the experts in this area. They
- 6:33certainly weren't buying freeze-dried
- 6:34food and building prepper bunkers, you
- 6:37know, like like like like me and many of
- 6:39my my my buddies who who have who have
- 6:42indulged fantasies about what doom could
- 6:44look like and also thought realistically
- 6:46about it. And so yeah, the AI doom is
- 6:49way down the list for me. I'm I'm I'm
- 6:51I'm a big optimist. I think it's going
- 6:52to massively reduce the cost of resource
- 6:54extraction, processing, transformation.
- 6:57You're going to see every part of
- 6:58society become subject to the same
- 7:01economies of scale and efficiency that
- 7:03automated industries that automated
- 7:05without AI have seen. So think like
- 7:08agriculture or or textiles. You can buy
- 7:122 weeks of calories for less than 1 hour
- 7:17of federal minimum wage labor. That's
- 7:19crazy.
- 7:20>> Let's give it up for
- 7:21Palmer here. You're going to see that
- 7:23for everything.
- 7:24>> David [applause]
- 7:26>> So I I got a really
- 7:27um
- 7:28pressing question on my mind. I'm kind
- 7:29of a feeds and feeds and ships and AI
- 7:31foundation models guy.
- 7:33And so when you're putting systems in
- 7:35the field, weapons in the field, uh
- 7:37you can you can put the intelligence
- 7:39inside it and it this has always been a
- 7:41big problem for the Air Force, like if
- 7:42we're going to start flying drones
- 7:43instead of pilots, who's going to
- 7:45unleash the weapons and how's that
- 7:46decision going to get made? And the the
- 7:48downside of transmitting it back to
- 7:50headquarters is you have latency.
- 7:53Uh the upside is you have a lot more
- 7:54compute. Uh what you can put locally in
- 7:57a device, in a robot in the field, in a
- 7:58plane is limited, but it's really really
- 8:01fast. So what's your philosophy on like,
- 8:03you know, I think a lot of people would
- 8:04say, well, are we going to have
- 8:06self-contained decision-making weapons
- 8:08wandering around enemy territories or
- 8:10are we going to have like are we going
- 8:11to have you know, control at the
- 8:13headquarters?
- 8:13>> Yeah, more or less. The the real concern
- 8:15isn't so much the latency or or the
- 8:19speed of computation. The problem is
- 8:21that if you rely on
- 8:24you know, a central a central location
- 8:25to do your compute, you now need to have
- 8:28command and control that has a good
- 8:31high-bandwidth communication back and
- 8:33forth between you and potentially
- 8:35thousands or tens of thousands of active
- 8:37assets. And if your enemy jams your
- 8:39communications or intercepts your
- 8:41communications or hacks your
- 8:42communications, you're done.
- 8:44Uh or worse, they just skip targeting
- 8:47all the stuff in the field and they go
- 8:49for the place where it's all being
- 8:51controlled. You know, we saw we saw this
- 8:53I there was a there was a there there
- 8:55was there was this documentary that
- 8:57George Lucas was working on and and the
- 8:59Trade Federation, they had they had
- 9:01their central AI command and control
- 9:03blown up and that was how the Battle of
- 9:05Naboo actually resolved was all all the
- 9:08battle droids lost
- 9:09>> Oh yeah, they all fell over.
- 9:10>> That's right, they all fell over. And so
- 9:12I think that that's that that's the
- 9:13practical reality that's driving the
- 9:15need to have autonomy living not back
- 9:18home, but forward. There's other areas
- 9:20like you go underwater, you can't talk
- 9:23to things. And sometimes you could talk
- 9:25to things, but you don't want to. Having
- 9:27radio emissions actually exposes you
- 9:29necessarily. And so what what you
- 9:32for better or for worse, you're going to
- 9:33need this. I actually don't think it's
- 9:34as much of a shift as people think.
- 9:37I there's a lot of people who look at
- 9:38this and say, "Oh my god, Pandora's Box
- 9:40is being opened. You're going to have
- 9:41weapons that are making their own kill
- 9:43decisions." And they don't recognize the
- 9:46reality, which is that that's been the
- 9:47case for decades. You look at the You
- 9:49look at the You look at the Aegis system
- 9:51on destroyers. It has a fully autonomous
- 9:53mode. You can flip a button and it will
- 9:55shoot down anything that comes within
- 9:57whatever range you programmed it with
- 10:00without any human confirmation. You look
- 10:02at SeaWiz and SeaRAM for shooting down
- 10:05incoming missiles, incoming incoming
- 10:07artillery fire, incoming missiles, or
- 10:10incoming speedboats.
- 10:11Same thing. You can put them in fully
- 10:13autonomous modes. And I mean, it Going
- 10:15back to into Vietnam, we had missiles
- 10:17that you could launch from an aircraft.
- 10:18They flew over the horizon, tried to
- 10:20identify the right radio emitters that
- 10:22they believed corresponded with a
- 10:24surface-to-air missile launch site, and
- 10:26then struck them without any The person
- 10:28had to decide, "I'm going to send it in
- 10:29this area. They're going to look for
- 10:31this signal, and I know that there's
- 10:32this probability they're going to get it
- 10:34right." That decision is going to remain
- 10:36with people. But, autonomous weapons are
- 10:39widely widely There's a whole other talk
- 10:42we're not going to have today, but I
- 10:43will I I do have a rant about this where
- 10:46actually autonomous weapons have existed
- 10:47for thousands of years. If you look at
- 10:50Defined as kind of weapons that can
- 10:53execute the intent of what you
- 10:55programmed to do independent from the
- 10:57operator to the engineer. You look at,
- 10:59for example, pitfall traps that allow
- 11:01people to walk over them, but then
- 11:02capture the commander on his horse when
- 11:04he walks by and it's heavier. You look
- 11:06at certain poison traps that would that
- 11:09would kill people who had who had
- 11:12certain characteristics, not kill people
- 11:14who had other characteristics, or not
- 11:15kill animals.
- 11:17You look
- 11:18many types of landmines and sea mines
- 11:21and even even torpedoes. And so,
- 11:23it depends on on your exact definition.
- 11:26But, even going into myth where people
- 11:27are creating golems that can you go out
- 11:29and fight on their fight on their behalf
- 11:31or or defend you defend their city.
- 11:33I I guess my argument is we've had
- 11:35autonomous weapons for a very very long
- 11:37time. It's just a new set of technology
- 11:40that's that's making them better.
- 11:41>> So, technical question on that. The
- 11:42foundation models that go into those
- 11:44those remote systems
- 11:46>> Or the pit traps.
- 11:47>> Yeah, well so that that
- 11:49right right now the open source models
- 11:51are all Chinese. So, I'm pretty sure
- 11:53you're not going to be putting Chinese
- 11:54models into remote weapons. Are you
- 11:56going to build Anduril foundation models
- 11:58or how how are you going to get AI in
- 12:00those devices?
- 12:01>> I mean, we've been doing it since 2017.
- 12:03So, we're way ahead of everybody. We're
- 12:05we're a decade beyond that question.
- 12:08Um
- 12:09I think that we also don't the
- 12:11foundation models that people are
- 12:12building are not the types of models you
- 12:14want to be doing this. They're non that
- 12:16they're largely non-deterministic.
- 12:17They're non-auditable. It it's it's a
- 12:19very different category. The thing that
- 12:21we are building, unfortunately, has very
- 12:23few consumer applications.
- 12:25>> Mhm.
- 12:25>> And vice versa.
- 12:27Um there of course are LLM use cases for
- 12:30defense that make a lot of sense, but
- 12:31it's not guiding my missile or or
- 12:34classifying radar your radar radar radar
- 12:36feeds. I mean, and we we're also we
- 12:38can't build models that ingest text and
- 12:40then spit out text. I need to process
- 12:43raw data radar data feeds. I need to
- 12:45process sonar data from passive
- 12:47listening stations on submarines. That's
- 12:49it's it's pretty highly specialized. And
- 12:51I I will take some credit here.
- 12:53Uh all these people who are talking
- 12:55about how AI is going to change
- 12:56everything
- 12:57they are mostly uh
- 12:59recent converts.
- 13:01Anduril Industries, you might notice the
- 13:03acronym is AI. Um so, and that that was
- 13:06not that was not a mistake.
- 13:07>> Okay.
- 13:08>> Alex, jump in.
- 13:09>> Yeah,
- 13:10I think maybe dovetailing on our
- 13:12discussion of driving bench, we're now
- 13:14in an era when it's possible to take
- 13:16frontier models, drop them into a
- 13:17robotic embodiment, and get pretty
- 13:20stunning results. So, I'd love to ask
- 13:21you a question maybe about the scaling
- 13:23hypothesis as it applies to autonomous
- 13:26weapon systems or autonomous systems in
- 13:28general. If you have to extrapolate out,
- 13:30is there a Moore's law? Is there an
- 13:32experience curve for what as obviously
- 13:36the capabilities of underlying frontier
- 13:37models, whether they're from Frontier
- 13:39Labs proper or from AI,
- 13:42as as those progress,
- 13:44do we see any equivalent of a Moore's
- 13:47law for the autonomous systems
- 13:49themselves? And and maybe the the second
- 13:52half of the question would be, do you
- 13:54think ultimately we achieve some sort of
- 13:56convergency? Anduril it makes a variety
- 13:59of embodiments in a variety of different
- 14:01form factors. Do you see ultimately some
- 14:04general purpose category killer robot
- 14:07that is equivalently the generalist
- 14:10robotic embodiment just like we now have
- 14:12generalist models that are capable of
- 14:14doing substantially any task?
- 14:16>> No, I don't think so on the defense
- 14:18side. I think that for manufacturing and
- 14:21for hospitality and medicine, you might
- 14:23see this because you get economies of
- 14:25scale that uh that that that that that
- 14:27are really that are really material for
- 14:29those types of businesses and you know,
- 14:30a humanoid robot may not be the optimal
- 14:33form for a burger griller and it may not
- 14:35be the optimal form for a firefighter.
- 14:37It may not be the optimal form for a for
- 14:38a nurse [laughter] or an assembly
- 14:40technician.
- 14:41>> I see you're smiling.
- 14:42>> But even if it's not optimal, the fact
- 14:45that it is shared with humanity in its
- 14:47form makes it a really really good form
- 14:50factor. And now I can make one robot
- 14:52that does all of these things and I I
- 14:54would probably rather have a let's say
- 14:56robot that's 5% slower or or you know,
- 15:00less reliable than the optimal burger
- 15:02flipping robot if I can share
- 15:04commonality with that with that nurse.
- 15:07In defense, that's not really the case
- 15:11because you are in this constant race to
- 15:13not just barely beat your adversary, but
- 15:15to massively overpower them. To try and
- 15:18make it where you know, the goal is is
- 15:20to just barely win a conflict on the
- 15:22margins. It is to fight unfair, to have
- 15:26overwhelming threat of violence
- 15:28underpinning whatever diplomatic
- 15:30overtures or negotiations you have. And
- 15:32and and everyone was always pushing the
- 15:34limits. So, if if if I if I have a jet
- 15:37that has a reliable 5% every time edge
- 15:40over the enemy, that's worth paying a
- 15:42lot of money for, right? Like I I I and
- 15:45and I I I think the the
- 15:48the loss in performance you would see
- 15:49through lack of specialization probably
- 15:52makes it non-viable. Like to to to go to
- 15:54the to the extreme like a you know, I
- 15:56build robot submarines and those have a
- 15:58specialized form that needs to dive to
- 16:006,000 m depth and and sit on the bottom
- 16:03of the ocean for months at a time. And
- 16:05then I have other stuff that's on the
- 16:07surface and and you know, they're
- 16:09robotic systems that you know, roll up a
- 16:10hill and and they loiter and then I have
- 16:12other things that are you know, aerial
- 16:13vehicles that are doing very specialized
- 16:15tasks and I just can't imagine a
- 16:16convergence that would make sense.
- 16:18Unless you're telling me that we're
- 16:19going to get some cool like like
- 16:20humanoids that have like you know,
- 16:22on-board fusion reactors and they can
- 16:24dive to 6,000 ft depth.
- 16:26>> So, Palmer, why aren't why aren't you
- 16:27building those? I mean, if anybody, you
- 16:30should
- 16:31>> [laughter]
- 16:32>> Look, you could potentially build such a
- 16:33thing, but then very quickly you realize
- 16:35that specialization would allow you to
- 16:37build something even better. You say,
- 16:39"Okay, I've built something that can
- 16:40dive to the bottom of the ocean and fly
- 16:42to the moon." Man, it could get to the
- 16:44moon a lot faster if it didn't have to
- 16:46also dive to the bottom of the ocean.
- 16:47So, I think I think defense is going to
- 16:48be this one where kind of like a
- 16:51you know, you you look at for example
- 16:52the difference between cars where
- 16:54everything has shared platforms within a
- 16:56manufacturer you know, manufacturer and
- 16:58they also share a lot of the same
- 17:00standards
- 17:01and the interfaces.
- 17:03I think that you compare that with like
- 17:05a motorsport where it's all completely
- 17:07different and it's continuously pushing
- 17:09forward. I I think the defense side is
- 17:12probably always going to be this this
- 17:14weird branch of exotic highly
- 17:16specialized robotic systems.
- 17:18>> You know, Palmer, I expected if anybody
- 17:20was going to be building humanoid
- 17:21robots, it would be you.
- 17:22>> No, no, no.
- 17:22>> just you you said you said absolutely
- 17:24not.
- 17:24>> I've I've opted out because um I because
- 17:27I believe in them. If I didn't believe
- 17:29that humanoid robotics had such a
- 17:30powerful future, then I would probably
- 17:32build them. But I want Anduril to build
- 17:34things that wouldn't exist otherwise.
- 17:36That's what I generally want to do.
- 17:40Correct. All the other people are
- 17:41already doing a great job. Like I I I
- 17:44want to use their robots.
- 17:45>> And and and do things with them. Like I
- 17:47think humanoids are going to be an
- 17:48incredible general artificial
- 17:51intelligence interface between AI and
- 17:54remote systems and a lot of legacy
- 17:56platforms that were designed for a
- 17:57person to you know pull the lever, push
- 18:00the button, read the screen. Like we're
- 18:02going to be able to take all those and
- 18:03automate them without replacing them
- 18:05entirely. So, you have a lot of things
- 18:07like
- 18:08radar systems that cost millions, tens
- 18:10of millions, even hundreds of millions
- 18:11of dollars.
- 18:12Uh nobody wants to man them during a hot
- 18:14conflict cuz they're the first thing
- 18:16that's going to go. But if I can put
- 18:17humanoid robots in them, I've now been
- 18:20able to automate them without that. So,
- 18:21yeah, I'm not going to build humanoids.
- 18:23Uh this is also why Anduril is not
- 18:25building like small quadcopter attack
- 18:27drones. Too many other companies doing a
- 18:28great job. There's a there's a lot of
- 18:30a lot of areas. And and we're not going
- 18:31to do space launch. Like we do space
- 18:33products. We have a lot we operate a lot
- 18:35of our own satellites and a lot of our
- 18:36own sensors in space. Uh but I don't
- 18:38want to be a launch provider. That's for
- 18:39all the other billionaires to
- 18:41>> Yeah, yeah, let let them handle that.
- 18:42That's fantastic. Salim,
- 18:44um you know, when you're operating
- 18:47Anduril Industries, you're uh
- 18:49decentralizing defense, right? The
- 18:52Pentagon and ministries of defense
- 18:55globally have been the most centralized
- 18:57entity in the history of humanity.
- 18:59Uh you're sitting on the edge dealing
- 19:01with that. Uh
- 19:03you must have a massive immune system
- 19:05problem where people inside the entity
- 19:07are going, "Nah, we've done that before.
- 19:09We can do it better." They either try
- 19:11and suck you in or they try and spit you
- 19:13out. How are you navigating that
- 19:15interface because you're doing it in a
- 19:16way that I've never seen anybody succeed
- 19:19do it.
- 19:22>> I think that this is a problem that's
- 19:23very, very hard at the beginning.
- 19:26>> Oh, I said the wrong thing.
- 19:28>> [laughter]
- 19:28>> Wow.
- 19:29>> No idea what the
- 19:30>> Well, until someone gets on the intercom
- 19:31and tells us to
- 19:33>> Someone won a million dollars.
- 19:34Congratulations.
- 19:36>> Um what were we talking about? What what
- 19:38what what was it?
- 19:44Seriously, I knew the intercom was going
- 19:46to come on.
- 19:47What were we talking about again?
- 19:48>> Centralization of Pentagon versus
- 19:50decentralized government.
- 19:52>> I'm going to I'm going to keep talking
- 19:53till somebody tells me that I can't. Um
- 19:55So, we've gotten through this a bit. I
- 19:57think early on there was that concern
- 19:59that Anduril was not going to be able to
- 20:01get through that. Our Our hardest
- 20:03problem raising money
- 20:04>> Skippy, would you please shut this down?
- 20:06>> [laughter]
- 20:08>> I've been through enough of these to to
- 20:09know how it's going to work.
- 20:11>> There we go.
- 20:11>> I imagine.
- 20:12>> One of our early problems was that our
- 20:14investors said, "Palmer, you potential
- 20:16investors." They said, "Look, we believe
- 20:18that you'll be able to build things that
- 20:19work. We believe you'll be able to build
- 20:21products that are great, but we don't
- 20:23believe the government is capable of
- 20:25actually purchasing them at scale." Cuz
- 20:26I mean, you talk you talk about
- 20:27centralization, you know about the
- 20:29issues. For people who don't, uh until
- 20:32just a few years ago, you had about 80%
- 20:35of MDAPs, major defense acquisition
- 20:38programs. So, the basically anything
- 20:41that that that matters, uh 80% of MDAPs
- 20:44went to five companies, and that's it.
- 20:48And I mean, you've seen incredible
- 20:50consolidation, centralization. Anduril
- 20:52has been successful because we use we're
- 20:54we're we're less of a contractor and
- 20:55more of a product company. We use our
- 20:57own money to build products, then we
- 20:58sell them, and we've gotten to the point
- 21:00now where people in government have done
- 21:02very, very well betting on Anduril, and
- 21:04then miraculously we finish on schedule,
- 21:07we finish on time, and then the next
- 21:09program manager says, "I know how I can
- 21:11make my job easy. I just I just give the
- 21:14contract to Anduril."
- 21:15>> Got it.
- 21:16>> So, it it's got to take guts to enter
- 21:18the defense industry, right? The most
- 21:20incestuous,
- 21:21you know, self-licking ice cream cone on
- 21:23the planet.
- 21:24>> Yeah.
- 21:25>> And and and what insanity,
- 21:28you know, gave you the balls to do that?
- 21:31>> The insanity was the reason that I got
- 21:33into it.
- 21:35Uh the first page of our pitch deck said
- 21:37Anduril will save taxpayers hundreds of
- 21:39billions of dollars a year by making
- 21:42tens of billions of dollars a year. The
- 21:43idea being we're not just making
- 21:45products that we're selling, but we are
- 21:46displacing the systems that cost
- 21:49needlessly 10 times what they should.
- 21:52And
- 21:53Yeah, thank you.
- 21:54>> [applause]
- 21:56>> And so, it does seem crazy, but, you
- 21:58know,
- 21:59I don't I I I can't put myself on the
- 22:01same pedestal as him, but one of my
- 22:03mentors and idols growing up and then
- 22:06also early in my career was John
- 22:07Carmack, the founder of id Software.
- 22:09>> John is great.
- 22:10>> John is incredible. And for people who
- 22:11don't know, he's basically the inventor
- 22:13of the modern modern 3D game engine,
- 22:16first-person shooters, also started a
- 22:18rocket company that competed in the
- 22:20>> He he was competing in the X Prize, and
- 22:21I I remember
- 22:22>> And he was one of the, if not the only
- 22:25one to be doing reusable
- 22:27>> Exactly, he was doing vertical takeoff,
- 22:28vertical landing.
- 22:29>> else was even trying to do. He did it.
- 22:31>> I remember there's this this name on the
- 22:33side of his vehicles, Armadillo
- 22:35Aerospace.
- 22:36How do you pronounce that thing? I
- 22:37thought you'd only know. That's right.
- 22:38He was using So, yeah, he was using uh
- 22:39he was using uh he was building his own
- 22:40sensor fusion algorithms. He was
- 22:41processing them on modern, you know, or
- 22:43on you know, commercial-style graphics
- 22:45card hardware using his game programming
- 22:50experience. Anyway, the point is it was
- 22:52called Armadillo Aerospace. That's
- 22:55right. And the thing and yeah, they were
- 22:57doing vertical takeoff and landing long
- 22:59before SpaceX was
- 23:01was was was I mean, doing anything.
- 23:03>> Yeah. No, before it existed, for sure.
- 23:06>> And but anyway, the the point is he he
- 23:10I I he ended up becoming the CTO of
- 23:12Oculus VR. I managed to poach him away
- 23:14from Id, but after he left, he went to
- 23:18work on AI actually even before I did.
- 23:22And so he started he started his own
- 23:23company to work on on on AGI. And
- 23:27and I said why why are you why are you
- 23:29doing this?
- 23:30>> You know, that the odds of you
- 23:30succeeding and to be clear, I still
- 23:32believe AI was going somewhere, but he
- 23:34said, "I'm going to start a company
- 23:35that's just going all in on the end
- 23:38state. Like I'm not trying to
- 23:39incrementally get there. It's just all
- 23:40in on AGI." And the point that he made
- 23:42was, "Palmer, even if I've got a 1%
- 23:45chance of succeeding, which maybe is an
- 23:47overestimation, but suppose I have a 1%
- 23:49chance of succeeding, and suppose that I
- 23:51could pull it off,
- 23:55the the risk-adjusted
- 23:57outcome kind of demands that I do it.
- 24:01And that was how I felt about defense.
- 24:02If there was a 1% chance that I could
- 24:05help solve this problem, then it was one
- 24:07that would be massively impactful to
- 24:09work on versus, let's say, starting
- 24:11another gaming company, getting into
- 24:13space launch,
- 24:15you know, building a car company, which
- 24:18it it all of the billionaires, they end
- 24:20up doing the same thing. You don't have
- 24:21enough people getting into defense or or
- 24:23putting on a bat suit. It's just it's
- 24:26just cars and rockets and and they just
- 24:28go in circles like that for for for
- 24:31>> So, you know, when I interviewed you on
- 24:32the on the Moonshots podcast last time,
- 24:34you mentioned that you're part of a
- 24:36billionaire's boys club chat.
- 24:38>> Yeah, yeah.
- 24:39>> [clears throat]
- 24:39>> Oh, yeah.
- 24:39>> Are they working on anything worthwhile
- 24:41at this point? I mean, I gave I gave
- 24:43them for saying like, "God, you got
- 24:44all this capital. Solve the world's
- 24:46biggest problems."
- 24:47>> I think the B boys are doing a lot
- 24:48better now than we were a few years ago.
- 24:50>> Okay.
- 24:51>> Um
- 24:51>> They're they're stepping up.
- 24:53>> And and for people you think we're
- 24:54talking about is I I was giving a lot of
- 24:56people a hard time, especially some
- 24:57retired B boys. And uh I was telling
- 25:01them what are you what are you doing?
- 25:01You have cuz money is one of the
- 25:04resources they have. But one of the
- 25:06resources people don't talk about,
- 25:07especially for the retired B boys, is
- 25:09that they have credibility, which would
- 25:11allow them to gather people around them,
- 25:13teams around them, and capital. Like
- 25:16they could and it's just sitting there
- 25:17underutilized if they're doing nothing.
- 25:20They have the credibility in that people
- 25:22trust them as a leader to be able to
- 25:23lead an organization like that.
- 25:26They could have the same ideas and the
- 25:28same words as somebody who's equally
- 25:30smart or smarter and they'll still do
- 25:33better for that. And so my point to them
- 25:34is you have a responsibility not only as
- 25:36someone who's who's got a lot of capital
- 25:38yourself, but who's got this
- 25:39underutilized resource of the ability to
- 25:41do more.
- 25:44If if you're not doing that, then are
- 25:46you really like what are you even doing?
- 25:48Anyone can sit on a boat. Few people can
- 25:51start another multi-billion dollar
- 25:53company that does something that
- 25:54matters.
- 25:54>> [applause]
- 25:56>> And the good news is most of them I
- 25:58think I I I think a lot of people have
- 26:00have gotten the itch to to solve some of
- 26:02these problems that are out there.
- 26:04Nihilism is is is on is on the down.
- 26:10>> [laughter]
- 26:10>> Yeah, that's an interesting point.
- 26:12Although
- 26:13Like I said, I think most of the B boys
- 26:15I think they seem like they keep pretty
- 26:17busy.
- 26:18I don't I don't I don't have a problem
- 26:20with B boys who don't do enough. Like to
- 26:22some people they mis- mistake what I
- 26:23what I say like oh Palmer wants other
- 26:25billionaires to work on defense or
- 26:27energy or or biology.
- 26:30You like
- 26:31my point is not that people need to
- 26:33distract themselves by doing more
- 26:34things. A lot of people can only do one
- 26:36thing.
- 26:37You know, and that was what I did for a
- 26:38long time.
- 26:40It's what I hate is the people who are
- 26:42not doing anything. That's that that is
- 26:44that is like a repulsive to me.
- 26:46>> So maybe message to the audience, you
- 26:48get to join the B boys slack group or
- 26:50whatever this is,
- 26:51Telegram, if you're doing something new
- 26:54and differentiated.
- 26:55>> The group actually got shut down and uh
- 26:57we've been we've been talking about now
- 26:59that we need to do a B boys 2.0.
- 27:01Um but I think I think people post
- 27:03>> some B girls as well.
- 27:04>> I think people posted the wrong memes a
- 27:05few too many times.
- 27:07>> Boom.
- 27:08>> I I look uh there there was there was a
- 27:10time where there were no there were no B
- 27:12girls and then there was this Business
- 27:14Insider article uh whining about it and
- 27:17uh the the reality is like it's just
- 27:18like me and some guys.
- 27:20>> Like you know, it's not like there's
- 27:21this isn't like the Illuminati
- 27:22conference.
- 27:23>> [laughter]
- 27:23>> You know, it's like they're they're like
- 27:25oh, there's a group and you know, but no
- 27:27women have been invited. I'm like I it's
- 27:28not that no women have been invited.
- 27:30It's just like me and some some of my
- 27:32boys. But
- 27:33>> [laughter]
- 27:33>> uh but actually there there was a there
- 27:35was a there was a there was a girl in it
- 27:37by the end. So, uh we need Yeah, we need
- 27:39to we need to restart that. We need to
- 27:40restart that group.
- 27:41>> So, you're in Facebook commercials now.
- 27:44>> Oh, yeah. Yeah, yeah.
- 27:45>> Yeah, so so the new Meta VR is out, uh
- 27:49how does it compare against your vision
- 27:50of what you wanted to achieve? Is it
- 27:53Yeah.
- 27:54>> Look, it does a lot of the things that
- 27:56uh it's not exactly what I would make
- 27:57were I the guy who's waving the magic
- 27:59wand and and and and you got lots of
- 28:01magic wands spinning the CAD, but
- 28:04uh
- 28:04look, the new Meta VR glasses are a lot
- 28:07closer in a lot of ways. Uh so, I've
- 28:09long been a believer that you need to
- 28:11decouple the compute and the battery
- 28:14from the head. You should it should not
- 28:16be that and and I'm not even saying the
- 28:18head totally, it's more of the front.
- 28:19So, uh before Oculus Quest. So,
- 28:22Oculus Quest was the last thing that I
- 28:24worked on before they fired me. And uh
- 28:26the prototypes at the time were
- 28:28a prototype series code named Santa Cruz
- 28:31and uh we were sure actually showing
- 28:33that publicly before they fired me. And
- 28:36uh Santa Cruz had the battery and
- 28:37compute on the back of the head and then
- 28:40the light, you know, the part on the
- 28:41soft tissue on the front of your head
- 28:42was kept as light as possible. And then
- 28:44at some point after they got rid of me,
- 28:46they they jammed it all into one big
- 28:48giant lump in the front and they ruined
- 28:49the ergonomics. They forgot why you need
- 28:51a top strap. Uh,
- 28:53but you know, it it it it bothers
- 28:55>> Serve them right.
- 28:56>> It bothers me most Look, it bothers me
- 28:59mostly at this point because uh
- 29:01the Oculus Rift CV1, which was the first
- 29:03modern consumer VR headset, it got the
- 29:06ergonomics more or less perfect in 2016
- 29:10and you know, we we were showing it off
- 29:11in 2015 and literally nobody has done as
- 29:16good of as it and it's been over a
- 29:17decade. Every single person screws it
- 29:20up, even Apple. And I I I just I don't I
- 29:23can't comprehend it. How how could it be
- 29:25that the ergonomic design led by a
- 29:2819-year-old kid with no experience is
- 29:32better than anything that anyone has
- 29:34come up with since. Anyway, the new Meta
- 29:36VR glasses are the closest thing.
- 29:39And they're very lightweight. They've
- 29:40off-boarded the computer. They've
- 29:41off-boarded the batteries. They have
- 29:43really high-quality pass-through.
- 29:45Um they they It has a feature that I
- 29:47invented for CV1 called Transformo
- 29:49Fabric, which they do not call it.
- 29:51The basically using uh fabric uh a a a
- 29:55a soft flexible fab black black
- 29:58light-absorbing matte fabric to one
- 30:00minimize reflections, but two uh
- 30:02basically seal the cavity between and
- 30:04around the lenses. That means that you
- 30:06can move them uh and the fabric just
- 30:08stretches to meet where the lenses are
- 30:10for different interpupillary distances
- 30:12rather than having like sliding
- 30:13mechanical interfaces and plastic and
- 30:16they brought Transformo Fabric back for
- 30:18the first time since my my era. So, I'm
- 30:20I'm I'm a huge fan and uh you know, it's
- 30:22it's a third of the price of Apple's
- 30:24product and uh somewhat higher higher
- 30:27pixel density. I I think that's pretty
- 30:29extraordinary. Oh, and and and 1/8 the
- 30:32weight.
- 30:32>> Yeah. Amazing.
- 30:33>> Pretty amazing. Pretty good.
- 30:36>> Let's close on Let's close on this.
- 30:38We've got, you know, 1,500
- 30:39entrepreneurs, moonshot visionaries, and
- 30:43tens of thousands watching on the live
- 30:44stream.
- 30:46Uh
- 30:46what's your advice?
- 30:48What, you know, what did you learn in
- 30:50high school that made you succeed at the
- 30:54speed you did?
- 30:55We were talking about this on the plane
- 30:56right back.
- 30:57>> I I was really lucky to uh I guess maybe
- 31:00not lucky. I did it. Uh but
- 31:02>> Well, you are you are lucky.
- 31:04>> That's right.
- 31:05>> I am [laughter] lucky.
- 31:06>> It's in the name.
- 31:06>> I started an online community called
- 31:09ModRetro when I was 14 or 15. Uh it was
- 31:12an internet forum for people who were
- 31:15modifying vintage game consoles and
- 31:17computers, building handheld units,
- 31:19improving the performance in a variety
- 31:21of ways. We had a few thousand active
- 31:22members.
- 31:24And uh and a whole team of volunteer
- 31:26moderators and administrators with with
- 31:28me me at me at the top of it for some
- 31:30reason.
- 31:31And uh I learned a lot of valuable
- 31:33lessons about how you can build an
- 31:35organization, how you market an
- 31:37organization, how you get people to
- 31:39believe in what you're doing. I learned
- 31:41a lot about dealing with the types of
- 31:44problems you deal with when you're
- 31:45running a company, you know.
- 31:46>> With humans.
- 31:47>> Well, I mean, you you have you have two
- 31:48It's a tale as old as time. You've got
- 31:50two people.
- 31:51They're both brilliant. They're both
- 31:53very good at what they do.
- 31:54And they hate each other. And they
- 31:56disagree over something. You need to
- 31:58somehow mediate the conflict or separate
- 32:02them in a way that allows you to get the
- 32:03value that that allows both of them to
- 32:06contribute value to the greater project,
- 32:08uh but not have them hate you, not have
- 32:11them lose faith in the broader thing,
- 32:12but also you not not kill each other.
- 32:15Sometimes that means you need to
- 32:16reconcile things between the two.
- 32:18Sometimes it means that you need to put
- 32:20one guy over here and one guy over here.
- 32:22And sometimes you need to make sure that
- 32:23they don't know that that's what you're
- 32:24doing as you do it. That's just
- 32:26[laughter] I'm not saying that's the key
- 32:27to running a successful business is the
- 32:29management of the two problem,
- 32:31but it's
- 32:33Yeah, and then and once you make it a
- 32:34three-body problem, now I'm out.
- 32:36>> [laughter]
- 32:37>> But uh but but it's it's that and a
- 32:39hundred things like it that I learned
- 32:42running Mod Retro that I was able to
- 32:44take into starting Oculus, to carry
- 32:46forward into what I'm doing in my other
- 32:48companies. And uh that that's why I'm a
- 32:50I'm a strong believer that the the the
- 32:52kids should should should get a job. And
- 32:54I I I also had a job starting when I was
- 32:5614, so I learned how the real world
- 32:57works, too. But I I So, I'm a huge fan
- 33:00of people doing things at a scale that's
- 33:02larger than just themselves. In other
- 33:04words,
- 33:05don't just do a project, do a project
- 33:07with someone else. Do a project with
- 33:09other people. Build in the open, learn
- 33:12the the the the good parts of that, the
- 33:14bad parts of that. It it it's been
- 33:16extremely useful to me. And
- 33:18>> Has building changed in the age of AI?
- 33:19What advice do you have? Again, you
- 33:21know, we've been saying on the pod for a
- 33:23while now, you know, forget about grad
- 33:25school, you know, even college is a is a
- 33:27maybe. You didn't go to college. You
- 33:29know, and high schoolers should what?
- 33:32Find a a purpose they care about and use
- 33:35AI to go and code it. What's your your
- 33:37closing advice here on that?
- 33:40>> You know, one of the reasons I have not
- 33:42been as upset as some people about uh
- 33:48Now, it's funny. There's something on
- 33:49the screen. It's like it's to not talk
- 33:51about
- 33:52>> [laughter]
- 33:53>> Just wanted to make sure you saw that.
- 33:54>> I I I did. We're not going to talk about
- 33:55Don't think about it. All right,
- 33:56company. All right, great. Great.
- 33:59>> [laughter]
- 34:00>> The uh
- 34:02There's the There's a lot of people who
- 34:03who who love their craft and they love
- 34:05what they do. And uh and and they are
- 34:08rightfully mourning uh the end of, for
- 34:11example, coding things by hand top to
- 34:13bottom, start to finish as anything but
- 34:15like a craft endeavor. Um and and I and
- 34:18I understand that. It's it's the same
- 34:19way that a lot of artists felt when the
- 34:21camera was invented. They They whined
- 34:23and complained about how it ruined art
- 34:25and it made it so that anybody could
- 34:26push a button. And and you know, the
- 34:28these things are always painful for the
- 34:29people who are masters of their craft.
- 34:32Um I already got screwed in that way a
- 34:34long time ago. When you're a founder,
- 34:35you don't get to actually work in any of
- 34:36the tech anyway. All the other people
- 34:38get to have all the fun. And so, you
- 34:40know, I I like electrical engineering. I
- 34:42like optical engineering. I like
- 34:43mechanical engineering. I don't get to
- 34:45do any of it cuz the job of a founder is
- 34:47to
- 34:48hire people to to displace them. And and
- 34:51and to do what they thought they could
- 34:53do or what they like they do like doing
- 34:56better. The job, I've said many times to
- 34:58people, if you want to work on
- 34:59technology
- 35:01work for somebody else. If you want to
- 35:03work on nonstop start a
- 35:05company.
- 35:07And so and so what I would say to people
- 35:08is AI doesn't actually change the
- 35:11part of the job very much.
- 35:13You're still on the hook legally. You're
- 35:15still on the hook for all the
- 35:16interpersonal stuff. And you could you
- 35:18you can delegate some of it, but at the
- 35:19end of the day it's your call, it's your
- 35:21decision. It's the same thing for
- 35:23fundraising. It's the same thing for
- 35:24general product architecture. And when
- 35:26people make the wrong decisions, it's
- 35:27your choice
- 35:29as to when you need to overrule them,
- 35:31override them, replace them. I don't see
- 35:33AI replacing that part of it anytime
- 35:36soon. Not because it couldn't, but
- 35:38because the gains everywhere else are so
- 35:40much larger that I think we've got a
- 35:42good 10 years of speed running
- 35:45all of the other parts of the business
- 35:47before they before they come to that.
- 35:48So, look, AI changes a lot, but I I I'm
- 35:51telling you there's a lot of people you
- 35:52are in for a rude awakening if you are
- 35:54like a hobbyist coder or a hobbyist
- 35:56engineer and you think you're going to
- 35:58start a company and you think you're
- 35:59going to be doing all that stuff? No,
- 36:01you're going to learn a hard lesson. It
- 36:04Being a founder was was always like AI
- 36:06took your job. It's just that other
- 36:08people took it.
- 36:09>> Yeah.
- 36:10Ladies and gentlemen, let's give it up
- 36:12for Palmer Luckey.
- 36:19>> [music]
About this transcript
This page contains the full transcript of Palmer Luckey: Autonomous Weapons Are Ancient and Why Anduril Won't Build Humanoids | EP #295 by Peter H. Diamandis, generated from the public captions YouTube serves with the video. The transcript has 7,244 words across 1,122 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.