The Ups and Downs of a Drone Startup [Entire Talk]

First, I'm going to start with a quick show of hands. How many people saw the Falcon Heavy launch yesterday? Okay, so this is not that story, but this is on the continuum of that story. So we now measure entrepreneurship by Musks. So I would say this is about a two milli Musk story, but you know, it's not a zero millimusk story. And I hope you can see glimmers of the courage and ambition and techniques that Musk used in SpaceX in what we did with drones, not just because they're both in the air. So I'm going to start by, I'm going to basically, over the course of this, I'm going to tell you sort of my story and I'll make it as narrative as possible, but I'm also going to be telling two stories. One is about the use of open innovation as a disruptive tool and then ultimately a business founder. And the other is going to be a sobering but I hope also inspiring story about what it means to compete with great Chinese companies and without ruining the surprise. We lost, but maybe we also won. So. So I'm going to start with the year is 2007 and I'll show you some pictures of my kids in a moment. But basically think about the aerospace industry as looking like any other industry, which is to say that it was a very mature industry. It had lots of big companies, Boeing, Lockheed, McDonnell Douglas, et cetera. It's pretty well populated around a slide that you've seen many, many times, which is that there's price and there's features and the more features, the higher price, et cetera. And, and you should assume that this is sort of fully populated. And then you might ask yourself, why would anyone want to enter the aerospace industry, in our case the air, in Musk's case the space. And the answer is that there's no units on this chart. And if you put units on this chart, you suddenly realize that it starts at expensive and goes to very expensive. Now, the second book, as Matt mentioned, I wrote in the sort of the Long Tail Trilogy, was free. It was a book about the use of zero marginal cost economics to that is the Internet digital distribution as a form of marketing and changing a business model to a freemium business model. But the point is that as everyone has learned in economics, there is the notion of price elasticity and that if you lower the price you should discover discovery demand markets customers that you might not have also previously found. Now, the reason why the aerospace industry sort of started at expensive and went to Very expensive is because they were assuming there was essentially one customer, at least in space. And also true for drones, which is the business I was in. And that is the government. The government is a. Sometimes it's military, sometimes it's NASA, et cetera, but it's the government. And there's a certain amount of regulation that comes with this procurement processes and all this kind of stuff. And basically it's very. If you ask Elon Musk, which I did, he said, Elon, Mr. PayPal, Mr. Tesla, Mr. SolarCity, what makes you think that you can go to Mars? He didn't say material science. He didn't say Moore's Law. He didn't say software. He said cost plus accounting. Cost plus accounting means that the economic model of aerospace is broken. The economic model of aerospace is that. Is that the government says, what's it going to cost to make this? And you say X, and they say, we'll give you X + cost +. So there's no incentive to lower X because the government will give you a profit on whatever you say X is if you win the contract. And Elon Musk said, this is a completely flawed economic model. If you. It doesn't incentivize lower cost. And if you lower the cost, you can discover markets that maybe are beyond the government. Maybe there's a whole bunch of people who would love to put things in space, but they can't afford to do it. And if you lower the cost by say 10x, maybe you'll find more than 10x as much demand. So I don't do space, but I did do something else. I did drones. Now, I didn't actually do drones, as you'll see. I did Lego with my children. But we'll get to that in a moment. But the point is that I am not an aerospace guy. I wasn't even a hardware guy. But it was pretty clear that something in the world had changed to lower the barrier to entry, to kind of basic computers, sensors, autonomy and things like that. And I sort of did this experiment. I said, what if you. What if you took the essence of a drone, which is an autopilot, and you basically said, that's a computer with software. And what if you said, well, computers with software these days cost about 10 cents. What if we just sort of put a computer with software on a little board and charge $19 and said you could put it in any kind of toy plane, it would be a drone? What would that be like? And the answer is that it was ridiculous. It barely worked. It was hard to use it was dismissed as a toy. And people rightly laughed and said, that's not a Predator, that's not a Global Hawk, that's not a drone. That's just an Arduino in a foam plane. And I said, okay, well that's cool. What if you made it a little bit better and charged $30? Anyway, eventually we made autopilots for about $90. And we, by the time they stopped laughing, we put 2 million drones in the air. We put more drones in the air every week than all of the aerospace companies in the world combined. And now at a certain point they said, well, I guess maybe these are drones. We hadn't really thought of them that way. And so we just said, okay, we'll just put them in this here, this is one of them. We said, well, ok, if you don't like the autopilots, we'll just put it in a vehicle that flies. And they're like, oh, I guess that is kind of a drone. And I'm like, and we're going to find useful things to do with it. And this is what my company does today, is we just use drones to scan the world. And now we're doing what satellites do and reality capture and laser scanners, et cetera. And now we're like part of the construction industry. And it didn't occur to anybody back in those days that drones would be useful for that. They thought it was a weapon. And the notion that it could be a way to monitor construction sites or insurance or wildlife was just not possible. But this is what happened. And by the time we started doing this and got the price from $10 to $20 up to about $10,000, the aerospace industry woke up and said, oh, I guess we need to lower the price. And they did their damnedest and they managed to lower it to $100,000. So this model is exactly what's happening in aerospace as well, that the incumbents, who spent their entire life as companies setting up procurement processes and ensuring that the regulatory barriers were such that nobody else could come in, suddenly are confronting somebody who came from outside with something that initially looked like a toy, but had accomplished the job through different means. So that's the economic over picture. Now, how did it actually happen? And the answer is it was in fact Lego. That's my nine year old daughter, Erin. At the time I was the editor of Wired. I have five kids. My wife and I are scientists by training and our kids are not interested in science. So we think that's a problem and we Keep trying to change that. And so one day at Wired, we got these products in for review. LEGO sent the new LEGO NXT robotics kit, and somebody sent a radio control airplane. I thought, that's going to be awesome. We'll make robots on Saturday, we'll fly a plane on Sunday. Something will stick. And. And so this is Aaron reading the instructions. This is her brother Daniel hoping that it'll work. The good news is it worked. The bad news is that it was completely unimpressive to a kid. If you have seen Transformers, this doesn't do it. They're like, where's the freaking lasers, right? Why doesn't it walk? Real robotics is incredibly disappointing compared to computer graphics. So that sucked. And then we flew a plane and I flew it into a tree and that sucked. And so I said, well, you know, how could that have gone better? How could I have had a cooler robot and a better flying plane? And I was like, what if the robot flew the plane? Well, and so I Googled flying robot, and if you Google flying robot, the result is drone. I'm like, oh, yeah, I guess that's what a drone is. It's a flying robot. Wait, what's a drone? You Google drone and it's a aircraft with an autopilot. I'm like, oh, yeah, that would be a plane with a brain. I get it. Wait, what's an autopilot? If you Google autopilot, it turns out that it's sensors and compute and software, and it basically has to sort of figure out where down is, then it has to figure out where it is, and then it has to kind of follow a path. And I'm like, oh, that kind of sounds like what's in the LEGO Mindstorms kit. It's computer and sensors and software. So let's just do it. So here it is, the world's first LEGO autopilot. These are gyroscopes, accelerometers, magnetometers, arm core processors. Stuck a GPS on it via Bluetooth, stuck it in an airplane, and it flew badly. But that airplane is actually now in the LEGO museum in Berlin, Denmark, as the world's first LEGO unmanned aerial vehicle uav. It turns out that drones and autopilots are regulated as as missile cruise missile controllers, and they're export controlled, and that the act of putting this on the Internet was a violation of export control. We weaponized lego, which is one of my proudest moments. So at this point, my kids lost interest and went back to video games. And I'm like, whoa, what just Happened around the dinner table. My kids and I made a drone out of Lego and like balsa wood, you know, this. Something in this world has changed. What? So I created a website called DIY Drones. And there's a lot to be said for putting the letters DIY in front of traditional industries. You know, it really expands the mind. And so what happened is that just like a ton of people were thinking about the same thing at the same time as part of the maker movement. And it took off and this community came together and they started working on the software and working on the electronics and. And doing what open source communities do in a very kind of informal way. And at a certain point, it became clear we had a thing. As a matter of fact, we had the biggest robotics community in the world accidentally. Remember, this is a hobby. I'm doing part time in weekends, and we're starting to kind of reverse engineer the military industrial complex as a hobby, which was not our intention. And yet it kind of happened. It wanted to happen. And so at that point, everyone's like, this is awesome. Code, code, code. Print a circuit board, solder, solder, solder. And then a bunch of people come along and they've heard about drones and they're like, I totally want to do this. And we're like, here, here's the code. Just compile that and populate your pcb, fab your pcb and then populate your PCB over here. And they're like, what, do you just have a kit or something that you could sell? We can do that. So here it is, the first production line. Those are, yes, my children, these parts are LEGO parts. The cool part was that when I wanted the motors fabricated, this actually is for robotic blimp. But when I wanted motors fabricated, I'm like, I know nothing about motors. And I said, I'll bet they're made in China. And I went to Alibaba and I went to order motors. Because one of the important economic lessons you learn when you're selling things is that you can't buy things at retail price. If you intend to sell at retail price, you have to buy at wholesale to sell retail. So I didn't know where to buy motors wholesale, so I went to Alibaba and I found a company that manufactured motors. And I said, I'd like to buy some boners. And they're like, what kind would you like to buy? And I'm like, what do you got? And they're like, what do you want? And I'm like, what do you got? And they're like, here, here's a little sort of widget by which you can pick all the elements, the windings, the shaft sides, the magnets, et cetera. And then just when you're ready, just pick the quantity. And I'm like, I can design my own motor. And they're like, yeah, and it's all real time translated between Chinese and English. And I'm like, designed it. And then they're like, oh, okay, terrific. And I got to the end and here's a price per motor. And I wanted like, you know, 100. And it turned out the price per motor for 100 was like three bucks each. But for like 10,000. They were like 60 cents each. No, they were less than that, like 35 cents each. And I'm like, I don't need 60,000 motors, but how can I resist? So I just picked that and like 10 days later this box comes and all of my custom designed motors have just shipped and been shipped from China. I was like, I can to PayPal. I was like, I just got robots in China to work for me designing a custom motor. And they took PayPal. This is something in this world has changed. I pay the children in strawberries and juice. We made this. So the good news was that it's sold out immediately. The robotic blimp kit. They had an autopilot, but it was obviously an easier one designed to sell at maker faires. Then we learned another important lesson about economics. So I went to the kids and I said, good news, the blimps are hit. They sold out immediately. And I'm like, yeah, whatever. And I'm like, you know what that means? And I'm like, no, it means we got to make more. And they're like, no way, that's not going to happen. I'm like, okay, that's a problem. I'm not going to make more. My kids won't make more. What am I going to do? So I found this guy in our community, one of the first guys who was there. And I'm like, he seems super smart. His name was Jordy Munoz and he was flying a helicopter with a Wii controller. Just. That was awesome. I was like, jordan, you seem super smart. You seem to know everything. Do you want to help me out here? And he's like, yeah. He says, I got a little spare time. I'm like, okay, cool. And he says, yeah, I'll solder to get some boards for you. And I'm like, cool, what do you need? And he said, lease the parts. So how much? About 500 bucks. So I sent him a check for 500 bucks. And he sends me back this picture. And there he is soldering the boards. I'm like, done. OK, hobby now fully off in the hands. Majority. 500 bucks. Maybe it was a lot of money, but I'm sure it'll work out okay. And he sends me back another picture and he says, so I hired some people. It's selling really well. We're now selling other autopilots and other sensors and things like that. And I'm like. He says, we've got a little industrial space. I'm like, that's incredible. They got shelves. He has a bookkeeper. It's really good. In the back, they have a reflow oven for printed circuit boards. There was a toaster oven from Target that was repurposed. And I'm like, who knew he was such a pro? He sends me another picture and he says, yeah, so we scaled up. These are pick and place machines and reflow ovens and stencil and printers and CNCs in the back. And now we have about 30 people. We're on track to do about $5 million this year. Remember, I've sent him a check for $500. I have not met him yet. And I'm like, okay, I think I need to go down and meet him. So I met him, and it turned out that when I first contacted him, he was a teenager in Tijuana and he had dropped out of college and he and his wife were just moving to the US So that his child would be born and become a citizen. So he had some spare time waiting for his green card. And so this was like a perfect fit for him. And then he said, oh, by the way, we've opened up a second factory in Tijuana, and here it is. And now they're wearing smocks with our logo, which I think I'd drawn in Microsoft paint or something. And the thing, you can't see it on this picture. But what I love about this picture is that at this point, he's really pro. And he's got electrostatic discharge protection. So there's these cables that connect you to the machines, so the static electricity has dissipated. And my friends who, like me, don't know anything about electronics manufacturing are like, you chain your workers to the machines. So cruel. And then at this point, we're now the biggest drone manufacturer in America. This is like the second plant in Tijuana, by the way. I think the Tijuana drone factory is a great name for a band. It's not taken yet. Totally go for that. So I'm like, I think I need to quit my job. I think we have accidentally created a different form of aerospace company. So we raised some money. I quit my job, moved the headquarters to Berkeley where I'd gone to school, where I live, and off we went. So when I say we disrupted the aerospace industry, this is what we did. This is where we launched right there. This is the price of helicopters, import price of helicopters in the United States. We basically destroyed the helicopter industry because it turned out, not because we make helicopters, because we made another way to levitate sensors in the sky. And it turns out a lot of helicopters didn't want to be helicopters. They didn't want to have pilots. They just wanted to put a camera in the sky, like traffic helicopters or firefighting helicopters or lidar scanning. They just want to put a sensor in the sky. And there's better ways to do it than putting a man in the sky, a human in the sky. So you don't see that very often. Remember, this is me. I'm right here at the dining room table with my children in Lego, right? This is me right there sending $500 to a teenager in Tijuana. This is the teenager in Tijuana building a factory on his own by buying used pick and place machines on ebay and downloading the manual from the Internet. And this is the US helicopter market. Now, we didn't do it alone, I have to say, at this point, a lot of other people were starting to use our autopilot, and we created an industry of people who are this kind of bottoms up, grassroots approach to aerospace. But that's pretty profound, unfortunately, because of the scale you can't see. It starts at $1 million, and it pretty much looks like it goes to zero. It didn't go to zero. It went to about $1,500, which is where we came out with this. But that same force that we used to disrupt the aerospace industry was used to disrupt us. A very good Co. Named DJI out of Shenzhen just did what we did, but did it better. They brought the cost down, they brought the innovation base up. And the price of consumer drone went from $1,300 to $500 in nine months, which I don't think the world has ever seen either. And that is. That's the neutron bomb, right? That is the end of the American drone industry, or at least consumer drone industry. That's it. Boom. It's gone. Everyone is gone. We're gone. GoPro's gone. You know Lilly and everybody, it completely vaporized the American drone industry on the hardware side. So I'll talk about that in a moment. So this is pretty dramatic, right? I mean, this is like, we talk about disruption. This is the future of aerospace is being played out at light speed on a global scale using open innovation, using all sorts of ways to get around regulatory barriers without breaking the law. Things like export control and FAA regs, all of them had exemptions for what we were doing. So that's pretty dramatic. Basically, when you pull back, this is what happened. There was an industry that was going like this and was going to continue going like that forever, until suddenly we came out of nowhere and we're like, hey, it's just developer stuff. It's just boards, it's just autopilot, it's just open source code. And they laughed. And then we're like, hey, you know what, you can actually put that in some plastic and batteries and motors and it's a pretty cool. It actually works. It's really easy to use. And maybe the audience, again, discovering new audiences. Maybe the audience is not, you know, the Air Force or the CIA. Maybe the audience is like kids at Christmas. Or maybe the audience is like construction workers, or maybe the audience is like, you know, scientists looking at land management. So, you know, if you made it really cheap, maybe you'd find something there if you'd only made it easy for them. And then we went a little further and they're like, you know what? Maybe, maybe this isn't even about drones. Maybe this is just about measuring the world. Maybe this is more like satellites and the Internet of things than it is like aerospace. Maybe, maybe we're just extending the Internet into the sky and that becomes a kind of really big deal. So this is what's happening. We're right about here right now, which is to say the consumer one is still bigger, but the enterprise side of drones are becoming a thing. And so when you look at how we did this, we basically disrupted an industry by doing everything they did. We did the opposite. Right? So, and this is pretty much. It's not Elon Musk exactly, but it's pretty close. So rather than high price, we're like, how about zero or as close to zero as we can get? Rather than the aerospace industry, which has an impeccable record of safety. It has never been safer to fly a fly on a jetliner. Today we're like, how about if we crash all the time in the beginning? Because we're learning. But don't worry, no one's on board, so nobody gets hurt. And so it's like. So the aerospace industry is like even the notion of nines, five nines, six nines of reliability. We spent three years going for 1:9 because before initially it would crash every time we flew it. And then by the time we were at about 20, 10 or so, it crashed about once every 10 times, which is pretty good. That's 1.9. Then it crashed about once every hundred times. And that was pretty much where we were ready to go to consumer side. Because remember, your phone probably crashes one out of 100 times used, and now we're shooting for about three nines, crashes about one out of a thousand times. But it's no big deal, nobody gets hurt, they're cheap. And that's the thing is if you take nines out of the equation, then the pace of innovation really accelerates. And if you don't have humans on board, then the risk of innovation, the cost of crashing, is so low that you don't have to emulate the aerospace model. The aerospace industry is highly regulated. It's export control, it's faa, it's fcc, it's Department of Transportation, the works. And we're like, let's not do that. Let's not replicate the aerospace model. Let's find ways where we can find sandboxes, where we can innovate that have no regulations. And so we found under 4ft within visual line of sight, no flight over people, no flight at night. Suddenly it's deregulated. You don't have to get these permissions. We said initially, if you sell to consumers, you have to get FCC approval. If you sell to developers, you don't. You have export control unless it's public domain, open source, public domain or at least. So we argued, exempted from export control. And everyone's like, what? But these rules were written in the 70s. It never occurred to them that cruise missile controllers were going to be created by nine year olds on dining room tables with Lego. And so the notion of public domain cruise missile controllers, which is essentially what an autopilot is, it just wasn't a thing. It never occurred to them that this could happen. And yet it did. So we just found ways where the regulations were low, the pace of innovation as a result was high, and the numbers of customers was super high. Again, millions of drones in the air, while the military industrial complex was putting hundreds of drones in the air. So this is what happened. We started with these worlds. We sort of said, hey, there's these guys who are interested in these people who are interested in robots. And that was kind of me. And then there were a bunch of people who were interested in radio control airplanes. And we said, what if we put the robot in the airplane? And that's that thing you saw with that yellow plane with the LEGO in it. And that was kind of interesting. So we merged two communities and that was DIY drones. And then we said, okay, well that's pretty cool. And then a bunch of people said, hey, can I just buy it? And that's created 3D Robotics, now known as 3 Doctor, which was just designed to commercialize the work of the community. And then the RC people who really like flying didn't want autopilots and they're like, you know, I just want to. I like sticks. I mean, I want the joy of flying. We're like, great, that's not for us. But still, that could benefit from autopilots to make it easier to fly. So the toys are manually piloted, video taking videos of things manually piloted. And drone racing is manually piloted. And that was good fun, but that's not the direction we were going. Then we were saying we want autonomy, we want to make these truly pilotless vehicles. So we're going to focus on the software and the hardware side and that's three Dr. And that created the solo. And then we got our. And then the video guys who like flying, they actually tried to start commercializing what they did. And so some of them became pilots who did drone services as a service, drones as a service, and that's great. That's not what we do. We on the other hand, decided to get out of the hardware business because DJI was amazing and we just focused on the software. And then there's the drone software that flies in the drone and then there's the software software to analyze the data. And we did the data side, which is called Site scan and that's using data to scan construction sites. And we spun off the software into the Linux foundation as an open source community called Drone Code. And today that's intel and Qualcomm and ARM and Situ and Flir and Unique and a bunch of others, about 30 companies. And today this is sort of the Android of drones and it's all open source and it is a consortium and it is run under an open governance model. And I'm the chairman. But so this is the company and this is the community. And then there's a second community which I'm going to talk about in a second called Archer Pilot, which is sort of a standalone and one Sort of all focused on business and the other one's more focused on sort of developer friendliness. And this was a really sad thing, what happened between the facility splitting of the community into two forks. I'm going to take a second to talk about this because it's a really important lesson in open innovation. Don't do what I did. This was, when you're doing open source, it's really important to structure the open source project, both in terms of its governance structure and its license in a way that can be commercialized, if that's what you ultimately aim to do. So normally when you look at the continuum of open innovation, it starts with things like App Store, APIs, SDKs, et cetera. And then as you go into actual open source software, there's a bunch of licenses and there's a general sense that the more sort of copy left, the more radical the license, the better it is. And the GPLv3 is the most radical of the licenses. And in 2009, when we were setting up the license, we had these community members who are amazing developers, just the best I've seen. But they were open source zealots and I'm an open source zealot, or so I thought. And they were like, GPLv3 is the most viral, it will get the most developer engagement. People are required when they use this to give back. And I'm like, that sounds good. That definitely spurs developer engagement. I hadn't thought about what else that might mean, but it turns out that the GPLv3, and this is super controversial, I'm going to get shot for this. Metaphorically, the GPLv3 is, I think, a crisis in the open source community. It has essentially been banned by all big companies here. I would go as far as to say it's toxic for business. I think it is fantastic for developers. But because of its viral nature, because everything, every time you change the software, you must publish it, it discourages companies from using it because it can infect everything else with this. This is forced to expose your IP and as a result, nobody will touch it anymore. Now, this is not to be confused with the GPLv2, which is what Linux is. The GPLv3 was a quite extreme deviation of it. And I think that fissure in the open source software community is going to be felt for years. And I made the wrong bet. I went GPLv3 with ardupilot. As a result, once we decided to commercialize it, no one would touch it. So we fortunately had another group called PX4 based out of ETH in Zurich, another university. And they had chosen the BSD license. And so this one here is copy left, and this one's permissive, the BSD license. The BSD license is also completely open source, but you don't have to re. You don't have to distribute if you make modifications. You can choose to, but you don't have to. And the bsd, the permissive licenses are the ones that companies embrace. And so we had to divorce. And so one community went off gpl and they're doing great. Fantastic software, fantastic community, but really hard to commercialize. And then we created a new competing stack on the BSD license, and that's the one that's now going to be used by most companies, including air taxis. And for example, the new Udacity air taxi course is based on this stack, the drone code stack stack, which is BSD licensed. So that was super painful. I had to be a complete jerk and essentially fire the community. I started because I made a mistake on the license and I don't feel good about it, but it had to be done. And I think that was something that we've learned a lot about over the last few years, about the differences between those licenses. Okay, so I have got. I'm going to go another 10 minutes here, and then we're going to take questions. So that was my story. That's what happened. This is part of a kind of a bigger. I learned some bigger lessons about innovation and openness. But basically, if you pull back, this is an experiment in the nature of the firm. And you've probably seen these guys before. This is Ronald Coase, whose theory of the firm created this notion of transaction costs and that companies exist to minimize transaction costs. And Bill Joy from Sun Microsoft Systems, who challenged that by saying, whoever you are, the smartest people in the world don't work for you. So Ronald Coase says, the only way to get things done is if people work for you. And Bill Joyce says, well, that may be true, but the smartest people don't work for you. Challenge. So with terrible graphics. And I know these are not the worst graphics that have been in this room, but they're probably close. This is what Coase said. Coase said, hey, transaction costs are very high until you create a company and roles and responsibilities under one roof, easy to communicate, everyone knows what they do. And transaction costs drop as the company gets more bigger and more professional. And as you know, that's not true. It's true to a point. But at a certain point, companies get big and bureaucratic and then the transaction costs rise again. So that's kind of a problem. Right? You know, we just seen ge, one of the best companies in the world. This is ge, who sort of through great management, managed to keep themselves here, but they were competing with companies that were down here. And so even GE found that the disadvantages of being big and bureaucratic outweighed the advantages. So what do you do about that? And the answer, as I think we are collectively learning, is that companies are a fantastic way to get things done up to a point. And that the only way to get things done at scale that continues to be efficient, like a little company, is to merge a company with a community. So three doctor Is here and drone code is there and that merger and that notion. And by the way, this is Stanford, right? This is Silicon Valley. We understand that communities are a big deal. We understand that having open APIs and developer tools is the right way to do it. We understand that platforms are more important than products. We get that. But this is still a pretty radical notion out there in the world. And how to do it well is not obvious. So I think this is. If I were to get Coase and Joy into a room, I'll bet. I think COS is actually dead at this point, isn't he? Okay, okay. So it wouldn't be that fun a conversation, but at some point imagining it, that's where they would probably agree, is that they're both right. They're just right at different scales. So to summarize paradox. That's right to a point. That's the way it really works and that's how you merge the two theories. So my next. So, you know, Nobel Prize, please. Okay, we talked about SpaceX. I started on SpaceX. Here's just a little data. SpaceX is. So we were competing with the Lockheeds and the Boeings, et cetera, the aerospace industry. SpaceX is competing with something called the United Launch alliance, which is a consortium of big aerospace companies. And I just wanted to show you the numbers. This, by the way, is maybe it's a few months out of date, but basically the cost. So Falcon Heavy has 1/4 the cost of the United Launch alliance lifter, 1/4 the cost and twice the payload. So basically it's 8x better pound per dollar ratio. So call it an order of magnitude. And interestingly, that's very common for SpaceX is almost always about an order of magnitude factor of 10 cheaper than its alternatives. And that's something he's worked on. And what's Interesting is that there's price elasticity just in the, in the. So this is the number of launches. So in 2017, SpaceX did something like 12. It's probably about 14 launches. And the United Launch alliance did eight launches. So this is an example of him using price to discover new markets. Now, his first, you know, this is going to be microsats and it's going to be telecoms and it's going to be roadsters to Mars and who knows? But basically lowering the cost of access to space by an order of magnitude is a really big deal. And you're doing this with not just eliminating cost plus accounting, but also eliminating the notion of disposable boosters and all the rest. So you're seeing it playing out there, you're going to see it playing out in cars. So I told you that 10 years ago I started a community called DIY Drones to, to kind of bottoms up, transform the drone industry. Well done. It again. Now it's DIY Robocars. And we're going to. We're now one of the biggest autonomous car communities in the world. Something on the order of 4,000 participants who are racing almost every week wheel to wheel. But we're not doing it with full size cars and people on board in the road because I don't have a car and I don't want to die and I don't have permission to drive on the road. So we do it at subscale from Go Karts on down here. I'm going to go back, actually. So actually this is what it looks like. We just have these warehouses. We have hackathons every week. This is Carl Bass, the CEO of Autodesk, who we turned into a crash test dummy. The steering wheel is disabled. All he has is a red button that's a kill switch. And this is. And you know, it looks like a toy. Sound familiar? That's a billion, a couple billion dollars of R and D. But they're kind of on the same continuum. We use the same technology, we use the Cameras, we use TensorFlow, we use OpenCV, we use the cloud. The cost of this is about $200. The cost of this is, well, you can't buy one raspberry PIs versus a bunch of Nvidia boards, maybe a radar, maybe a 2D lidar, et cetera. But it turns out that you can kind of emulate a lot of what's going on in the full size autonomous car. Space for left for 200 bucks in a warehouse with a bunch of nerds and this is our data. We're doing pretty well. We've got a neural network team. This is just the latest races. The neural network teams have been getting our. Those are the, those are the blue dots and the traditional computer vision. So think of this as like a proxy war with the neural network teams being like Google Waymo and the computer vision ones being like Tesla. And so the Google Tesla proxy war is playing out pretty well. I would say that I think, actually I have updated for the next one. I think the computer vision team has just pulled ahead. Maybe it means Tesla's going to pull ahead. But what's interesting is that we're going to beat humans by probably about August. So a bunch of nerds in a warehouse in Oakland and 40 other warehouses around the world using open source software and Raspberry PIs are going to be faster than humans at scale at 16 scale up to go karts by the end of this year. Why do we do this? Well, because it's fun. Because we can. What's the point of reinventing a wheel that you know, that Google and Tesla are already inventing? And the answer is, I'm going to go right back to the same thing we did with drones. We're going to do things they don't do. First of all, we crash all the time and that's good because that means we iterate really, really fast and nobody gets hurt. Second of all, we race wheel to wheel. That's not actually happening anywhere else in the world. No. Autonomous cars are racing wheel to wheel. We do it every weekend when you race wheel to wheel. And we know this in the car industry, this is how cars innovate, is that the car companies, the Ferraris and the Porsches, etcetera, they race wheel to wheel in Formula one, et cetera. And this is how automotive technology has always advanced. Except for an autonomous cars where we don't do that. How about doing that? Why would you do that? Well, it turns out if you want to be safe, and this is for those of you, I think a lot of you have engineering backgrounds. For those of you with an engineering background who understand control theory, there's basically two ways to be safe. One is to be very gradual, very slow, very cautious. And the other is to be very nimble, very fast, very highly responsive, so you can drive like a little old lady. And that's the way all autonomous cars, all autonomous cars are the first, the second order effect of autonomous cars is going to be traffic jams everywhere because they're so conservative, they don't know what to do? They stop and we're like, well, what if you just are super fast and you can nimbly dodge through traffic? I don't know, maybe it's a terrible idea, but that's what we're doing. I got other slides, but I think on that note I'm going to. Oh yeah, I wrote a book about all this stuff which you can read. But on that note of destroying a bunch of amateurs getting faster than humans by the end of the year, I think I will stop and take questions. Thanks very much. I realize, by the way, I didn't dig deep enough into the whole competing with China thing. So if some of the questions are about China, that would be welcome questions. How do you compete with China? Fair question. A plan. How do I compete with China? So the question was, how do I compete with China? I described earlier on how DJI just crushed all the American companies, basically all the companies around the world by being really good at what they do. So DJI is a company that grew up in the Pearl River Delta in Shenzhen. I described the companies of that era as the ones who were making iPhones for 10 years and taking notes. Basically the iPhone processes and everything about Apple and the iPhone was something they were really well schooled on. Dji. So I lived in China for four years. My children were born there. And I knew that they were not, they were not just copying, they weren't just low cost manufacturing. I knew that there was a really strong engineering culture there. And so we were prepared for this. We were prepared for tough competition from China. And we had the theory, which I'll get to in a second. But everyone said, and you still hear this, I can't believe you still hear this in Silicon Valley. Everyone said, oh, don't worry about China, they can't do X. And sometimes X was innovate or sometimes X was like global distribution, or sometimes X was marketing or sometimes X was software or design or something. As far as I can tell, there's no X. We haven't found an X. DJI does all that stuff. They started global. Fantastic software, fantastic marketing, fantastic design. They innovate. Everyone said, oh, don't worry, China just copies. I don't know when that was true, but that has not been true in my life since I've been paying attention there. DJI is innovating faster than everybody else. You know, I talked to like the US trade reps and they're like, so did DJI dump products under cost? Do you want to file? I was like, no, they kicked our ass. Just fair and square. And it's like, by the way. And they kept their prices higher than ours. What they did is they just innovated faster. So what they would do is they would release a product at 1,500 bucks, which is higher than our price, and we're like, ah, our product's cheaper. Then, like, six months later, they release another product at 1600, 1500 bucks, whatever. And then that product they'd released six months earlier goes down to 500 bucks. And so they managed to get the high end of the market and the low end of the market by simply iterating so fast. And I'm like, you know, this is not dumping, right? This is just innovation. This is. This is, you know, a company, a country, doing to Silicon Valley what Silicon Valley did to everybody else. So what do you do about that? And the answer is we got the heck out of hardware. There was just no light at the end of the tunnel. I mean, we lost. I don't know. I don't know how much we lost. We raised almost $200 million. And we, in the course of doing all this kind of stuff, we spent most of it. Some large amount of that was lost in going head to head. GoPro lost at least $100 million. Parrot lost $200 million. I mean, there's probably. Well, DJI, meanwhile, raised about a billion dollars at something north of $10 billion valuation. And everybody else just got crushed trying to kind of compete with the efficiencies of the supply chain and all that. So we got out of hardware, and that is, I think, the right thing to do. DJI created, largely, I think, inspired by our model. They created an SDK and an API, much like Apple, kind of more of the Apple App Store model than the Android model. And that's good. And today we use the DJI vehicles, and they're fantastic. And our software runs on them as it went from the product, the hardware, to the data. Then things started to change a little bit. So people are very happy to buy Chinese hardware that's fantastically good and cheap. They're less happy having their data in a Chinese server. And so in November of last year, the Department of Homeland Security banned DJI for. Essentially banned DJI for government use. And today DJI cannot be used by the US Government. And increasingly other NATO governments are not allowing the use of DJI equipment. I don't think that's fair, by the way. I don't think DJI is spying, but there's a lot of paranoia about this, and it's One thing to have commodity hardware, it's another thing to have data. And I think that there is going to be a war. And I think the great firewall of China, which is very effective at protecting the Chinese Internet from outside competition, actually works both ways. I think it'd be very difficult for Chinese cloud companies to succeed outside of China. So DJI is a hardware company, Global First, Alibaba, Tencent, Baidu. It's going to be difficult to see them compete outside of China despite having a fantastic technology, because there's some real issues about not just trust but also legal frameworks. You know, it's one thing to say, hey, we have this terms of service, we promise to do right. But does the Chinese government, are they comfortable with these national champions having operations outside of the role of Chinese law? Unclear. So I think it's a really interesting war being played out at the data side. So I think when it comes to consumer electronics, China owns it, we're done. When it comes to the software running on consumer electronics, you know, we're doing fine. Like apps on Google phones which are made in China. When it comes to enterprise grade data, that I think is still an interesting battle to be played out. Yes, in the back, can you talk a little bit about your view of where journalism is heading today? The question is, where is journalism heading today? So I used to be an editor, not a journalist, but I used to be an editor in the media. I don't even know what journalism means anymore. So there's these words like media and journalism and it's like a lot of other things. There's capital J journalism, which is a kind of a pretty well established set of church and state and fourth estate and responsibilities. And then there's lowercase journalism, which is just the Internet. So we have the best of times and worst of times. I think it's an amazing time for writing, for personal expression, for creativity, for voices to get out there. And also many of those voices are accurately reporting their world. Now their world may be a kid's soccer game, it may be something happened on their block, it may be something happened in their company or industry. And so I think lowercase small J journalism, it's fantastic. Capital J journalism, as in sort of newspapers with business models. Good luck. I mean, I got no idea where there's light at the end of that tunnel. I was in the magazine world, which is relatively protected, but even that it's based on kind of a broken business model, 100 year old business model which says that basically you've got a three A three party system. You've got publishers, you've got readers, and then you've got advertisers. And the advertisers, you subsidize the publishing to reach the readers and then the readers sort of implicitly promise to buy from the advertisers. So this triangular market, well, it turns out there's other ways for advertisers to reach readers that doesn't involve going through that third leg of the triangle that is Facebook, Google, et cetera. And they're much more effective and they're measurable, et cetera. So as banner ads died, as display ads died, as non measurable ads died, so did the business model, traditional media. And there's a lot of experiments right now with paywalls. And if you're the New York Times, the Wall Street Journal, you'll be fine. I think Wired is now experimenting with a paywall and I think they'll be okay as well. But the San Francisco Chronicle, I don't know. I mean, I don't see how that's supported long term. And what happens as it goes away? Maybe initially we have a huge vacuum. I mean, there's lots of amateur journalists who are very happy to cover the PTA meeting and the soccer game and the block. Are they willing to cover the city council meeting? I'm not sure. We may have a gap. Yes. The way the DIY communities got started and the idea of reverse engineering and understanding how these established products work, were there any legal battles that you had to fight and were there any learnings from that? Yeah. So the question is, as we, as we kind of reverse engineered, did we have to fight legal battles? Surprisingly, not. Everyone told me that I was going to go to jail. They couldn't decide whether it was like, you know, federal jail or state jail or Guantanamo Bay. But it's like whatever you're doing, it cannot be legal. I mean, among other things, you're, you're basically open sourcing weapons is the way they saw it. You're taking military technology and you're giving it to, I don't know, terrorists, to say nothing of IP and all this kind of stuff. And so we kept expecting the FBI, the CIA or the NSA to show up. And they did. They showed up every week and they showed up to say, what you're doing is really interesting. Can you tell us who's using your stuff? And I'm like, yeah, absolutely. It's an open community here, there you can read it for yourself. They're doing it in public. And by the way, if anyone's doing anything bad, we'll be the first to tell you we're totally committed to this. And I'm like, so you don't have any problem with what we're doing? And I'm like, I don't think it's against any laws. And so literally every time I kept saying, is this the time I get sent to jail? And every time they said, I think there's, I don't think you're breaking any laws. Keep it up. Just be a good citizen. So, so we were, we thought that we'd get sued on patent stuff. Turns out that one of the nice things about disrupting an industry 30 or 40 years after it was invented is that all the patents had expired. There was nothing. There was nothing. We never got sued. Nobody got sued. It was the weirdest thing. Nobody went to jail. Even export control, the inspectors would come by and they're like, you know, I don't know what you're doing, but it can't be right. And we had to fight and we occasionally had to like, explain this. And eventually the White House came out with a white paper that said that literally the White House came with a white paper about DIY drones as a model of innovation. And they said, this is exactly what we should be doing. It'll be lowering the cost, increasing the efficiency. And by the way, this is consumer off the shelf technology. If you can buy it in Walmart, it cannot be an export control violation. And so we ended up spending a lot of time educating them on this. But here I am, still not in jail. You already talked a bit about how regulation from the 70s is disrupting DIY innovators, that seems to apply to many industries is enabler. So regulation from the 70s, disrupting DIY inhibitors or disrupting, blocking them from doing things, in our case, it was just the opposite. Regulations from the 70s didn't anticipate the possibilities of what we were doing. So we found loopholes. But do you think you could go farther if the regulation was not as specific? Yeah, a little bit, yeah. Today we work really closely with the faa, et cetera, but by and large, we were able to put millions of drones in the air with no changes to regulation whatsoever, just using the loopholes that were already there. So we think of this as like WI fi compared to cell towers. So back in the old days of wireless, if you wanted to make a radio, you had to get a license, broadcast license, paid billions of dollars for it. And then the FCC and their wisdom created a sandbox of 2.4 GHz and certain power limitations and spread spectrum. And they said, look, if you stay within these limits, you can do what you want, no license required. And basically we had that. There was something called the recreational use exemption. And because of the radio control hobbyists, there's this little carve out for something that looks like a radio control hobby use which is under 400ft in visual line of sight. And it never occurred to them that radio control hobbyists would end up essentially making drones. So we didn't have to change anything. The only exemption, the only problem was is that commercial use was banned. So recreational use was totally fine. But commercial use required a special license from the faa. And that was a real barrier towards that last step of adoption I described. And so the FAA to their credits, after they could see there was just an untenable number of applications coming in for waivers, they created an exemption for a certain kind of commercial use. It's called a part one of seven. If you get a used to be, you needed an actual pilot's license for commercial use. So think about, when you think about 1970s regulation, think about the paradox here. If I am a 7 year old child who gets a drone under the tree and opens up the box and walks outside, you can fly, no problems. If I'm a trained professional flying with expensive equipment, with all the gear, I am banned and I have to get a pilot's license to do that. And so why is it that we're enabling children but banning professionals? And the answer is you ban who you can ban. You can't send a letter to a child, cease and desist to a child, but you can send a cease and desist to a company. So we worked with them and we got an exemption. So we turned the pilot's license to basically a driver's license license and that's called a Part 107 exemption today that's enabled the commercial use of drones. So that was the. The FAA, by the way, has been fantastic in this and the FCC has been fantastic. Actually the regulators have been, have been surprisingly understanding and helpful on this. And by the way, you probably heard that when, when the Falcon Heavy launch went off, they thanked. Elon Musk was the first person they thanked and the FAA was the second. Second entity for letting this stuff happen. So they're doing great. Any other questions? One in the back and then you in the middle. Yes, in the back. So car used to have drivers and now they're autonomous ones. What do you think of the future of autonomous drones? When there are no pilots anymore. There are no pilots. So the way these drones work, there's no pilots, there's no sticks. They've been autonomous for five or six years. What about recreational and racing? Oh, they're piloting because they want a pilot. So drones haven't needed pilots for most of the decade. The way it works right here is that you just push, there's an iPad and just press a button and say go. And it does an autonomous mission. So there's no sticks, there's no piloting, it just does its own thing. It does it every hour, it's like a sprinkler. So for commercial use, they've been autonomous forever. What you're describing, things like drone racing, those aren't actually drones, those are called remotely piloted vehicles. But drones are autonomous. And cars just the opposite. Cars are actually going to maintain a driver for a long time because again, there's a human on board. And so there's a much higher risk, much higher requirement to have a fallback. A so called human in the loop fallback for safety. Right now, if anything were to go wrong here, there has to be a human present, there has to be a human in the vicinity who could take over control in case something went wrong. They never do with a car. It's quite important that the human be able to take over control instantaneously because the human's on board at risk. There was a question in the middle. Yeah, yeah, this is the last question.

So you talked about for drone and you talked about the new hobby with automobiles. You've also talked about the risk when you put the human on board and what's at stake with that. Do you think that there'll be kind of a segregation of solutions and market where there's one where you have the high risk tolerance and you fail fast, but then those devices or solutions will have to be kept away from where they can do the harm. And the ones where humans will be exposed to the risk or on board or whatnot, which is what Waymo and these other guys are solving, that these markets really won't have an intersection for say the Venn diagram won't overlap. Or do you see like within a decade or more that the failing fast and doing it at an order of magnitude cheaper will eventually subsume that. And now that the Waymo one is really only a temporary thing.

It's a really good question. So there's three domains, right? There's the domain of little things flying autonomously without people involved. There's not people on them, there's not People underneath them, they're just doing the thing. Then there's when you have little things or maybe even big things with no people on board, but there's people on the ground who could be hurt. And then there's things where you've got, like, people on board, right? So the first one of little things flying around people, flying without not around people, that's solved. We do that every day. The second one of little things flying over people, we're just getting permission to do that. The sort of. The second iteration of that is bigger things with no people on board, but flying over people. And that's like drone delivery, like Amazon drone delivery or Google drone delivery. And we're still trying to figure out the rules on that. Probably the easiest way to do is not fly over people, like fly over water and things like that, rural areas, et cetera. That's probably in the next five years. Then you get to the big things with people on board also flying over people. And that's the air taxi scenario. That one's going to be tough. You know, I'm one of the advisors on a bunch of. For several companies in this space. Uber's got a conference every year called Elevate about this. In particular, one thing we talk a lot about is the red button. So you're now in an Uber air taxi. So you took out your app, you said, I want to fly to San Francisco. And so it tells you that there's a rooftop around here somewhere. And you just get to the rooftop and the thing descends. You get in it, you take off. You're now on your way to San Francisco. Is there a red button? So if you get in the elevator, there's a red button you can press if you're not happy. A lot of machinery has kill switches, etc. Is there a red button? If things are not going well, is there a red button? And this is an interesting one to think about because let's say there's no red button. How do you feel about that? You're completely helpless, right? There's nothing you can do. Let's say there is a red button. How do you feel about that? What if you press the red button? What's supposed to happen? Does the red button do anything? In which case maybe a parachute comes out and you then gently land on top of a bunch of people. That would suck for them. Maybe the red button is just like a red button elevator that doesn't actually do anything, but it just rings a bell at headquarters, and headquarters says, hey, what's up? Maybe there's maybe the red button instantaneously switches to some manually piloted mode, but if a motor just died, it's not going to do any good. We haven't figured out what should happen. When you think about just the psychology and the negative externalities of the red button, it's super tricky, especially when there's no pilot on board. So that one, we're not going to solve it globally, we're going to solve it in specific places. So Dubai will say, we're totally cool with it. Go air taxi from the Dubai airport to central Dubai, whatever, we're cool. New York may not say that for decades. So I think the future is already here. It's just unevenly distributed.