Creating the Uncrashable Car

Published Mar 2, 2023, 5:05 AM

When Austin Russell was 17 years old, he founded Luminar Technologies to work on a remote sensing technology called Lidar. 

Today, Austin is one of the world's youngest self-made billionaires, and Luminar may be on the verge of solving Austin's problem: How do you make Lidar cheap enough and good enough to use in millions of cars?

Pushkin. For a while there, the discourse was full of self driving cars. They were right around the corner. Then they were not right around the corner. Now we have robotaxis in a few places, but not in most places, And for normal cars, autopilot does not really mean autopilot. In all this binary will they or won't they drive themselves talk? We've lost sight of something. Lots of people are figuring out big, interesting ways to bring new technologies to cars, even before we get to self driving cars, which looks like at this point it could take a while. These new technologies could save thousands and thousands of lives. I'm Jacob Goldstein and this is What's Your Problem, the show where I talk to people who are trying to make technological progress. My guest today is Austin Russell. He's the founder of Luminar, a company that is developing a technology called lidar. Austin's problem is this, how do you make lidar cheap enough and good enough to use in millions of cars. LDAR stands for light detection and ranging. It's kind of a technological sibling of sonar and radar. A lightar system makes a detailed map of its surroundings by sending out pulses of light and analyzing the reflections that come back. Luminar's LDAR system will come standard on a new Volvo SUV that's going into production later this year. The company also has deals with Mercedes Benz and with Saic, a Chinese car company. Austin says LDAR can do a lot for cars even before we get to full autonomy. It's not about replacing the driver, it's about enhancing the driver. I think this is a really really important part around what we're doing is they we're not trying to do some moonshot of replacing the driver altogether and having are welcoming our robotaxi overlords in a city near you like that's actually a very very challenging problem out on its own. Our goal is about enhancing the driver, making cars dramatically safer, going back to the original vision around what all this stuff was supposed to be in the first place, except actually making it happen. The reality is is that right now, you know, as many as one and a half million people have their lives lost on the road every year as a result of vehicle accidents. And this is something that's totally preventable for that matter. Like you know, it sounds like a basic problem, but just don't let your car smash into the thing right in front of you. That's how you prevent most accidents. Turns out that's actually not as straightforward as a problem as you think, and to be able to have the confidence to you know, fully take over from the driver and those kinds of situations is not an easy thing. So that's where the lighter is coming into play for this kind of high performance lightar that gives you confidence to do that and can enable the car to take over the breaking system and steering wheel to get you out of those kinds of situations. Austin has been working on this problem for a long time. He founded Luminar back in twenty twelve. At the time, Austin was just seventeen years old. Also, at the time, the use of LDAR in cars was, like Austin, barely out of its infancy. You take a look back in the day, you have these hundred thousand dollars fitting systems, you know, out of the roof of these you know, test cars that are out there, and you know, generally low performance, you know, not something that's nearly as high performance is what it can be. So the moment you're starting to think about this is that moment when like what we kind of think of as like the LDAR on the roof of whatever, the early like self driving Google vans or something like this great big thing and it costs one hundred thousand dollars. I mean it feels sort of analogous to the like vacuum tube computers of the like nineteen forties fifties. I mean, is that the right way to be thinking about it? Like, Oh, it's this giant thing. It doesn't work that well, but it's clear that somebody should be able to figure out how to make it smaller and cheaper and better. Absolutely, yeah, I mean the key thing is is that you could see some times of things up to one hundred meters out with those systems. But the hard part was is seeing you know, low reflectivity objects like like dark objects like a like a plaque car, a tire on the road, at person in dark clothing and anything begin to is very very difficult. So you set out to make this thing smaller and cheaper, like about two orders of magnitude cheaper, right, You got to drive down the price, like it's got to customs of what it costs, and it's got to be better. Right, that's that's that's what you're setting out to do. It's got to be tenets better at one of the cost. Yes, very good classics tech problem a day. It's just hard because it's hardware. Um. So, so I mean, tell me a little bit about about doing that, right, Like, surely there were moments when things weren't working, when you thought it was harder than it seemed. Tell me tell me about one of those moments. Yeah. Absolutely, Um I would say, by the way, it's actually absurdly difficult. I wouldn't necessarily recommend it for anyone there too, just because in this case it's a convergence of multiple things all at the same time. As you pointed out, hardware is hard already on its own, very very challenging. There's two other aspects of it are unique. There's the hardware. There's the fact that we're building crazy lasers and you know, and receivers and new custom chips that go into the systems. This is not something that you're using official parts like, you're creating new kinds of you know, three five special materials. You know that are not you know, your everyday silicon, you know that you can just be able to go and use and fab with. So there's a lot of very special things that go into this that we had to truly innovate from the physics up. So there's that part, and then there's also the automotive side of it, which by the way, is absurdly difficult out on its own. You know, just to be able to create something for automotive series production isn't by no means an easy task at all, high regulatory barriers, has to be super reliable, has to be super scalable. Yeah. Yeah. But the thing is that these are long design side goals here too. I mean you're talking about you know, five plus years of planning, you know, to go into it before where we're going to see the light of day in a car, and you know, very serious stuff, and you're and you're talking also probably the better part of a billion dollars worth of investment to be able to get there if you want to actually get into a production car. Good. So I am persuaded that it's really hard that you're setting out to do this very very hard thing. Yeah, I mean I can imagine there must have been multiple difficult moments, moments when some didn't work. Tell me, tell me one of those. Oh, that's probablys of like, you know, hundreds of things, but you know you don't in the earlier stages and something else, especially you know for some crazy sixteen or seventeen year old at the time you know, first started out, you know that not everyone was always on the same page. You know, they're too or even understood half the things I was saying, or ninety percent of the things I was saying. I mean, especially in the very early days. I mean, architecturally, tried every different type of combination of components, of wavelength of lights, of you know, semiconductor materials, of everything to be able to do this, and I think we actually counted, Yeah, it was over two thousand. I think it ended up being four thousand different ways of failing before you could actually have the right combination of components that can solve all of these different problems that you're trying to solve. For what does it even mean to say you're trying four thousand different things. That's a lot of different things. It's it's a lot of different things. I mean, in some of those cases we were prototyping it. In most of those cases, it was it was more on paper or in simulation, and you know, it turns out the really presumably to some extent, you're putting things on cars, you're trying things. You're putting Jane prototypes on type of cars and driving around the parking lot or whatever. Yeah, totally, totally, yeah, exactly. I think the first system that you have together, you know, took up a whole trunk, you know, of the car, just to be able to prove out the concept. By the way, when you first put together these prototypes, you know, without your own custom components in it, there too still would blow up. You'd have you know, receivers that would like suddenly fry themselves. When you say blow up, do you mean literally or metaphorically? Oh no, literally literally something blew up in the trunk of your car or something. Or tell me tell me about something blowing up. That's what I wanted. Yeah, well, I mean you used to have it where you know, the office shelf receivers, you know, before we building our own custom receivers, they would end up, you know, they couldn't handle the amount of the energy from the laser pulse and they would just fry themselves, you know, so it would be it would be stuff like that. Um, talk about problem and talk about trying to do a you know, a demo ten years ago. There two of what you're of, what you're putting together, and then next thing, you know, your your your receivers pride. So that's that's smoking smoke coming out of the trunk. Yeah, yeah, yeah, yeah, exactly why why did Chip let the smoke out? You know? Yeah, stay inside? It was like a mad scientist lab that we had there, you know, I mean it was it was the real deal. So so okay, how do you go from you know, things blowing up in the lab to a system that actually works where it's cheap enough to put on regular cars. The key was, you know, reducing the number of components you know, in the device. Normally you require these you know, hundred laser hundred receiver arrays you know to build you know, something that has any reasonable level performance. You know. Basically found a way to be able to do that using only a single laser in single receiver as opposed to requiring one hundred or hundreds. And that was part of the magic. But we had to end up creating the most you know, special and powerful, you know, and most sensitive receiver of its kind in the world, and these special custom chips and a very special laser and all these things to make that all possible. Is that the key innovation, for lack of a better word, is figuring out how to get more range and sensitivity with fewer lasers and receivers. Yeah, that was a key part of the critical innovation and secret sauce and other stuff to be able to make that happen. Absolutely well, how do you do it? I mean presumably everybody else would have used fewer lasers and receivers if they could have, Like is it software? Is it hardware? Is it? But yeah, I mean we literally had to make our own custom chips in our own systems there too, to be able to make this possible. Otherwise, as you said, everybody else would have done it. It's not out So you're not you're not just like monkey with the laser. You're you're designing a chip. You're designing the semiconductor that's going to go inside of it. Yeah, oh yeah, yeah, we're crazy like that, Like that's that's what we had to do, and there was no other way. Everything else would just cost thousands or tens of thousands of dollars, and we had to even work at a completely different wavelength of light and completely different special materials. Those to the existing stuff just wouldn't get you still to come on the show and still to come for Luminar. How do you build and sell wide our systems at scale and make a profit. That's the end of the ads. Now we're going back to the show. Is there a moment when things and I know it's not binary, but is there some moment when it sort of flips from basically not working to basically working? One of them? I didn't you still see the image online? It was it was down Palm Drive here and Stanford exactly exactly, Yeah, it's Stanford. And are you watching it in real time? Is the car like drive on a Palm drive and you're looking at the car controlling the system there too, I'm looking at it? Yeah, yeah, absolutely, So you're looking at it on your laptop. You're basically seeing what the system is seeing on your laptop in the car. Yeah. I had a monitor rigged up in the car there too, so I can actually go go watch and had a keyboard, had had a whole set up there too. Then, so you're sitting there, you got the thing rigged up, somebody's driving, you're looking at the laptop and what so they're driving. I'm there, I have a system set up, boot it up to go, turn it on and see what the output is. And then next thing, you know, you just have this just this beautiful image of everything, or and I say beautiful in an almost artistic way, because like I've just been you know, for ten years and even at that time for years, just so deeply embedded in the data. I almost like see data, you know, like the matrix, like like the matrix exactly. No, that's the thing. It literally looks like you're in the matrix. Because you can fly around this three D world. It's like totally abstract, you know, colorations and perspectives and everything that you could never do in the real world. So it actually is kind of wild. What do you see in in this instance? What do you see? So I'm seeing rows of these palm trees here, which I think it turns out, are just really really distinctive and really interesting looking in the live ar at least because you have these different palm fronts that then all become incredibly distinctive where you can just see even all like the texture and the detail of these leaves, where it can zoom and fly through them, you know, in three D because you could rerender it any angle. That's the part of the software that we had. You know that that was just absolutely mind blowing to me at the time. And I was like, oh my gosh, it's it's it really all came together to work because you don't know, right, you know, I mean, you you try this stuff over and over and over and over and over again until you can get like you can get it to be in a really good place. So not only are the pump chase beautiful, but your your thing is working, your system is working. Yeah, exactly, Like I always had these tears of joy, and I was like looking out of you know, because it's picking up stuff so far off, because it's you can see so small away. It's picking up stuff one hundreds of meters away, centimeter level precision detail. I mean, not to mention the whole thing's actually working in the first place. You know, it's it's um because the tech all came together and that there's stuff like that that that just makes all that different let's talk about kind of where you are right now in terms of what vehicles you're in and going into, and also where you've gotten the cost to and sort of what you still need to do that you haven't done yet. The overall goal of what we have from a big picture standpoint is to move. It is to build and enable the uncrashable car. So to say, you know something that where we can eliminate vehicle accidents and make any thing of the past, And is there I mean, do you have like a well one hundred year vision when you say the uncrashable car, do you have a year in your mind when it'll be possible to build the uncrashable car? Well, that's why I said the hundred year vision. But no, no, no, But but I think lifting already from a system that's there today much less for the future, could already start preventing massive amounts of accidents. Tell me about what what sort of the frontier for you? Clearly you've made this thing work, carmakers are starting to buy it. What haven't you figured out? Yet? What are you working on now that you still haven't solved? Execution scale? You know, continuing to evolve and mature of the business, you know, building software layers on top of the hardware. How is the how is the unit economics? The unit economics seem like they could be hard. You're not the only company selling lighter There are other big companies, like you know, there's a hard Wares is a famously hard business, partly for that reason, Like is that hard? That part's less hard? Actually? Um, I think that's something when you have a fundamentally differentiated and value technology there too, that just is distinctive. The only thing that actually matters is how you can successfully execute and how you can be able to build this up. Because right now, what's the percentage of cars out there that are that are currently lighter equipped or going to be lighter equipped? I mean in the immediate it's still like, you know, less than a percent, you know of the overall market. Like this I was going to get your market opportunity, Yeah, it's it's like, yeah, it rounds to zero, you know. So that that's the thing is that you know when you when it comes in with this whole next wave, you know there's that trillion dollars of opportunity, you know that you have ahead to do it. It's actually, I mean, is it is it hard to be profitable? Like, is it hard to get the profitable? Yeah? Yeah, yeah, yeah, I mean but here's the thing, here's the thing right now, it's a game of survival ever everything. I mean, you're you're you're familiar with the macroeconomic environment and everything for like growth technology companies there too. You can't raise money, you know, you're basically ninety percent of companies out there are totally screwed, you know, because there's no opportunity to be able to actually deliver on you know, a real product towards the real industry with real revenue and real capabilities or any real order book. Luminar is losing money, right, Luminar is not breaking even, No, we're not. We're not profitable overall as a company on a unit economics basis obviously, you know, for for the contracts of the stuff that we have signed up for that that is you know, uh, generating real money there too for us. But is it profitable? Are the unit economics profitable? I mean, obviously it's generating revenue yeah yeah, yeah for the production absolutely yeah yeah, so um yeah no no, And that's important of course, because you know it we lose money on every linear but we make it up in volume. Yeah, you make it up in volume. Yeah, exactly, exactly, Which, just to be clear, that is the strategy of like a lot of these growth companies and technology of these A lighter companies and av companies and everything is just obviously that's sustainable only for so long as long as you can have venture money funding it. Yeah, when the venture money isn't there, So when you're going to be profitable in theory, if you know, if you said, hell or high water, we're going to become profitable this year, you can you could find I con find a way to do that. It wouldn't be the right move for the business, but you know, there's also ways that you can love her. You mean, basically, stop stop spending money on R and D essentially and just make and sell the thing that you already know how to make and sell. Yeah, yeah, exactly. I could stop all investments for our future exactly, you know, but but that would be dumb as well. So you know, we've been in a good position, but it's it's just to be clear, it is not easy, Like this stuff is really freaking hard. In a minute, the lightning round, including the finer points of dropping out of college and how to hack both hardware and, maybe more impressively, the Bay Area real estate market. Now let's get back to the show. I want to finish. I appreciate your time. Almost done. I want to finish with Lightning round A bunch of quick questions. Absolutely, what's the most useful thing you learned in your twelve weeks of college? That it's not for me? Good? Who should drop out of college? Anybody that wants to build a great business and not being massive amounts of student debt. Okay, it's a ringing endorsement of dropping out. I like it's it's it's a spicy take, absolutely, and it's not for everyone. It's it's not, but but the reality is is that it is for a lot of people, and I think that's something that needs to be a more celebrated path. Who's your favorite inventor? Oh? Probably maybe Isaac Newton. Uh it's interesting to think about him as an inventor, Like, well, in your mind, what did he invent? Calculus? Yes? Sorry, the field of optics? You know from optics? Yea, optics and of course, uh I read that when you were a kid and your parents wouldn't let you have a have a cell phone. You hacked your Nintendo Wei to make it a phone? First of all, is that true? And second of all, if so, do you still hack stuff around the house now? So source it's it's partially true. It was my Nintendo DS, so that's that's all it is true. I did also hacking Nintendo, but as a difference, yes, DS makes more sense. Yeah, yeah, yeah exactly. Then a guy actually care about has a has a microphone at a speaker, had you know a void capabilities there too, so you know it always do fun stuff like that. Any example of a big or small hack in your adult life, actually talk about hacks. There was literally something like a content called like a hacker house, you know that you'd have that you were, you'd have a bunch of entrepreneurs. Let me the same plant watch Silicon Valley. Yes exactly. So when when I first came to Silicon Valley, I actually ran a number of hacker houses there too, which was talk about a hack. You know, no one would fund my business or do I have anything to do with this early on because they're like what the heck? You know, like that sounds crazy and all this stuff. So I actually ended up self funding it by renting out a bunch of these crazy mansions, you know, take out debt to go do that, and then would sublet each of the individual rooms to other entrepreneurs, you know, for thousands of dollars per month um, you know, to be able to actually then generate revenue, generate income that it would would fund towards the business. Is that the way you funded Luminar at the beginning? That was a huge part of it. Yeah, yeah, that was that was that was That was it. Austin Russell is the co founder and CEO of Luminar. Today's episode was produced by Edith Russelo, edited by Robert Smith and Sarah Nis, and engineered by Amanda k. Walk. I'm Jacob Goldstein. I am not on TikTok, but I am on Twitter at Jacob Goldstein. You can also email us at problem at Pushkin dot fm. We'll be back next week with another episode of the show.

What's Your Problem?

Every week on What’s Your Problem, entrepreneurs and engineers talk about the future they’re trying  
Social links
Follow podcast
Recent clips
Browse 143 clip(s)