Invest Like the Best with Patrick O'Shaughnessy
· Patrick O'Shaughnessy, Palmer Luckey
[00:00:00] S0: You said higher for your rig or this is… [00:00:00] S1: And I think that, that's probably for example -- I'm thinking what's the counter example, what's the game I would least want people to be playing. And it would be like other games that I do like but probably are not useful is probably Quake III or Unreal Tournament or maybe Call of Duty, ones where it's primarily based on repetition of twitch muscle reflexes. And I'm all about that. I think it's really fun. But someone having really good hand-eye coordination probably not going to be the thing you need to succeed at Anduril, probably. Maybe I'm missing something. [00:00:00] S0: Final two. What have you learned about marketing?
· Palmer Luckey
[00:00:00] S1: If you want to take a really cynical view of my career, it's that I have primarily accomplished marketing better than anything else. With Oculus, I know we did a lot of great technical stuff, but I think that the biggest thing we did differently was just figured out how to make VR seem appealing to people. That was the real win. And that was enabled by technology, being able to make a thing that was good enough. It was small and lightweight and low cost enough that it seemed reasonable. But we did so much work making it cool and making it desirable and making it something that people wanted to be involved with, both on the customer side and also the game developer side. Marketing is not just about your end user. It's about all the people you need to work with. When we market at Anduril, we're doing something very different. At Oculus, we were marketing primarily to game developers we needed to partner with and then to end users. And the thing about end users is you have to convince them one by one individually by the millions that they need this thing in their lives. In the defense space, you really are marketing to a very small group of people. You're marketing to people that you need to recruit, and that's the number one audience, people that you need to convince that Anduril is the place to be. Your other audience is politicians who pay for your stuff, and the third is your actual customers that're going to be buying your things. But with the DoD, I don't have to convince the end user. I'm not convincing the guy who is setting up the system in the field that this should be bought necessarily. I'm convincing the people who are more at the strategic level. Put it another way, every person in the U.S. who gets to decide what rocket motors get purchased, you could fit them in a conference room. There's not millions of people that I need to reach. It's actually a very small group of people. And so it's a very different kind of marketing that we have to do. I think that's probably the toughest lesson for me to learn coming into this from Oculus, where it was like one-to-all marketing as opposed to what we've done at Anduril, which is one to few and being hyper-targeted. It's really fun to make TV ads and do big trade show presences and to do a press tour where you meet with 50 press outlets over the course of four days.
· Palmer Luckey, Patrick O'Shaughnessy
That's what I would do when I was, for example, at Oculus, going to CES. I would go to CES, and we would give people 30-minute meetings each press outlet, and it would be 15 minutes of talking with me and then 15 minutes of getting a demo of the latest Oculus hardware and software. I would rotate through four press outlets an hour, doing this 15-minute spiel and we would start at about 9:00 a.m., and we would go till 5:00 p.m. So you'd have about 32 outlets a day, and you do that for four days, and you'd say, "Wow, I met with almost 200 people in the last few days. That was really fun." And of course, now you're reaching the millions of people that read those 150 or 200 outlets. But as fun as that was, it's not something that makes sense for Anduril. [00:00:00] S0: What do you think of AGI?
· Palmer Luckey
[00:00:00] S1: I love it. I'll be honest, I've been less focused on AGI because I think it's a very hard problem and I don't think that Anduril is in a position to be the company that figures it out. There's so many people trying so many different approaches, unlike VR headsets, where I think there's a clear tractable road map to solving the problem. It's very clear what you need to do to make it work really well. You can road map today a technical plan that gets you to a headset that is indistinguishable from reality on the visual side, at least, in some environments. It's very hard to reflect -- some things are harder, sun glaring off of snow, very, very difficult. I'd say there's a clear technical road map to a headset that provides you with an experience of sitting in a fluorescent-lit office room that is indistinguishable from the real deal. We know what we need to do. That is just not the case for AGI. There is not a clear path, and I suspect that whatever ends up working is going to be a bit different than what people are betting it is today. I'm going to let everybody else work on that, and I want to be in a position where the second someone figures it out, I can jump on it and integrate that capability into my system. So that means building systems that are optimized around autonomy in the first place, building systems that can be well controlled by a computer, building systems that are made more useful by more computational intelligence, things that can take advantage of advances in autonomy. And Anduril is in a really good place. If someone came up with a perfect AGI that was as smart as the smartest person in the world than anything, Anduril would be able to make great use of that, and I am looking forward to it. I hope John Carmack figures it out. If he does, then it means we get to work together directly again. That's the Palmer Luckey fantasy. It is that John Carmack is the one who figures out AGI with his Carmack-led crack team of geniuses. But whoever it ends up being, we're in a pretty good place to take advantage of it. The only other thing I'll say is, I am not scared of AGI the way that a lot of people are. I think that AGI is a technology where the benefits do accrue more to the defender than on the offensive side.
· Palmer Luckey
Obviously, people can pick examples here and there, "Aw, but here's this horrible thing that it could be used for." I still think the benefits actually accrue more to the defender. People will say, assuming that we get AGI, first of all, I think it's likely, especially in the near term, that it's going to require computational models that are extremely heavy that are going to be easier to run in basically just better-off nations that have strong regulatory environments and a general desire to do good than out at someone's garage. It could be someone figures out how to do it in their garage and they use it to bioengineer the next super weapon. But let's assume that AGI just exists in the more likely sense, and I think for a while it's likely to remain the domain of the people that you would want to have it. I think the better way to look at it is, "Yes, AGI could be used to engineer really terrifying bioweapons." But that's already the case today. There's already really good bioweapons you can make without AI. We struggle to defend against things like that. I think AI would be very helpful in being able to defend against. In other words, if AGI is so good that it can come up with anthrax two and it's just an incredibly powerful thing, I would hope that AGI can also come up with a thing that immunizes people against it or that builds nanobots that are in our blood stream at all times that are able to, on the fly, compute exactly what they're going to do to stop any biological threat, natural or man-made. It's a case where the defensive uses like that are actually going to outweigh the offensive uses in the long run. The thing to worry about is you have -- let's say, North Korea or Iran figures out how to do it with a relatively small number of people. They have the brilliant insight that nobody else does, and then they use it to create super weapons. But at the end of the day, I think that actually comes down less to an AI problem and more of a people problem. I'm way more scared of really evil people with existing weapons than the possibility that they get slightly better weapons. If they were to do something, the only thing stopping some of these nations is the threat of massive retaliation.
· Palmer Luckey, Patrick O'Shaughnessy
I'm not sure that a nuclear-equipped Iran is less spooky than an AI-bioweapon-equipped Iran, mostly because the way you stop them from deploying either of those capabilities isn't actually tied to the technology itself. It's tied to everything around it. [00:00:00] S0: I sincerely wish I had two more hours with you. This has been so much fun, but I'm forced to ask my traditional closing question, what is the kindest thing that anyone's ever done for you?