George Hotz: Comma.ai, OpenPilot, and Autonomous Vehicles

George Hotz is the founder of Comma.ai, a machine learning based vehicle automation company. He is an outspoken personality in the field of AI and technology in general. He first gained recognition for being the first person to carrier-unlock an iPhone, and since then has done quite a few interesting things at the intersection of hardware and software. This conversation is part of the Artificial Intelligence podcast. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter https://twitter.com/lexfridman, LinkedIn https://www.linkedin.com/in/lexfridman/, Facebook https://www.facebook.com/lexfridman, Medium https://medium.com/@lexfridman, or YouTube https://www.youtube.com/lexfridman where you can watch the video versions of these conversations. If you enjoy the podcast, please rate it 5 stars on iTunes or support it on Patreon https://www.patreon.com/lexfridman.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Lex Fridman?

  The following is a conversation with George Hots. He's the founder of Kama AI, a machine learning-based vehicle automation company. He is most certainly an outspoken personality in the field of AI and technology in general. He first gained recognition for being the first person to carry or unlock an iPhone. And since then, he's done quite a few interesting things at the intersection of hardware and software. This is the Artificial Intelligence Podcast. If you enjoy it, subscribe on YouTube, give it five stars on iTunes, support it on Patreon, or simply connect with me on Twitter at LexFriedman, spelled F-R-I-D-M-A-N. And I'd like to give a special thank you to Jennifer from Canada for her support of the podcast on Patreon.

• 0:45Lex Fridman?

  Merci beaucoup, Jennifer. She's been a friend and an engineering colleague for many years since I was in grad school. Your support means a lot and inspires me to keep this series going. And now, here's my conversation with George Hots.

• 1:19Lex Fridman?

  Do you think we're living in a simulation?

• 1:23George Hotz?

  Yes, but it may be unfalsifiable.

• 1:26Lex Fridman?

  What do you mean by unfalsifiable?

• 1:29George Hotz?

  So if the simulation is designed in such a way that they did like a formal proof to show that no information can get in and out, and if their hardware is designed for anything in the simulation to always keep the hardware in spec, it may be impossible to prove whether we're in a simulation or not.

• 1:49Lex Fridman?

  So they've designed it such that it's a closed system, you can't get outside the system?

• 1:53George Hotz?

  Well, maybe it's one of three worlds. We're either in a simulation which can be exploited, we're in a simulation which not only can't be exploited, but like, the same thing's true about VMs. A really well-designed VM, you can't even detect if you're in a VM or not.

• 2:07Lex Fridman?

  That's brilliant. So we're, it's, yeah, so the simulation is running on a virtual machine.

• 2:13George Hotz?

  Yeah, but now in reality, all VMs have ways to attack.

• 2:15Lex Fridman?

  That's the point. I mean, is it, you've done quite a bit of hacking yourself. So you should know that really any complicated system will have ways in and out.

• 2:27George Hotz?

  So this isn't necessarily true going forward. I spent my time away from Kama, I learned COCK. It's a dependently typed, like, it's a language for writing math proofs in. And if you write code that compiles in a language like that, it is correct by definition. The types check its correctness. So it's possible that the simulation is written in a language like this, in which case.

• 2:56Lex Fridman?

  Yeah, but that can't be sufficiently expressive, a language like that.

• 3:00George Hotz?

  Oh, it can.

• 3:01Lex Fridman?

  It can be?

• 3:01George Hotz?

  Oh, yeah.

• 3:02Lex Fridman?

  Okay, well, so, all right, so.

• 3:05George Hotz?

  The simulation doesn't have to be Turing complete if it has a scheduled end date. Looks like it does, actually, with entropy.

• 3:11Lex Fridman?

  I mean, I don't think that a simulation that results in something as complicated as the universe would have a form of proof of correctness, right? It's possible, of course.

• 3:26George Hotz?

  We have no idea how good their tooling is. And we have no idea how complicated the universe computer really is. It may be quite simple.

• 3:34Lex Fridman?

  It's just very large, right?

• 3:36George Hotz?

  It's very, it's definitely very large.

• 3:38Lex Fridman?

  But the fundamental rules might be super simple.

• 3:41George Hotz?

  Yeah, Conway's gonna live kind of stuff.

• 3:42Lex Fridman?

  Right.

• 3:44Lex Fridman?

  So if you could hack, so imagine a simulation that is hackable, if you could hack it, what would you change about the, like, how would you approach hacking a simulation?

• 3:58George Hotz?

  The reason I gave that talk.

• 4:00Lex Fridman?

  By the way, I'm not familiar with the talk you gave. I just read that you talked about escaping the simulation or something like that. So maybe you can tell me a little bit about the theme and the message there, too.

• 4:11George Hotz?

  It wasn't a very practical talk about how to actually escape a simulation. It was more about a way of restructuring an us versus them narrative. If.

• 4:24George Hotz?

  We continue on the path we're going with technology. I think we're in big trouble, like as a species and not just as a species, but even as me as an individual member of the species. So if we could change rhetoric to be more like to think upwards, like to think about that we're in a simulation and how we could get out. Already we'd be on the right path. What you actually do, once you do that, well, I assume I would have acquired way more intelligence in the process of doing that. So I'll just ask that.

• 4:56Lex Fridman?

  So the thinking upwards, what kind of ideas, what kind of breakthrough ideas do you think thinking in that way could inspire? And why did you say upwards?

• 5:06George Hotz?

  Upwards.

• 5:06Lex Fridman?

  Into space? Are you thinking sort of exploration in all forms?

• 5:10George Hotz?

  The space narrative that held for the modernist generation doesn't hold as well for the postmodern generation.

• 5:21Lex Fridman?

  What's the space narrative? Are we talking about the same space? The three dimensional space?

• 5:23George Hotz?

  No, no, space, like going up to space, like building like Elon Musk. Like we're going to build rockets, we're going to go to Mars, we're going to colonize the universe.

• 5:30Lex Fridman?

  And the narrative you're referring, I was born in the Soviet Union. You're referring to the race to space.

• 5:34George Hotz?

  The race to space, yes.

• 5:35Lex Fridman?

  Explore, okay.

• 5:36George Hotz?

  That was a great modernist narrative. It doesn't seem to hold the same weight in today's culture. I'm hoping for good postmodern narratives that replace it.

• 5:48Lex Fridman?

  So think, let's think, so you work a lot with AI. So AI is one formulation of that narrative. There could be also, I don't know how much you do in VR and AR.

• 5:58George Hotz?

  Yeah.

• 5:59Lex Fridman?

  That's another, I know less about it, but every time I play with it in our research, it's fascinating, that virtual world. Are you, are you interested in the virtual world?

• 6:08George Hotz?

  I would like to move to virtual reality.

• 6:11Lex Fridman?

  In terms of your work?

• 6:12George Hotz?

  No, I would like to physically move there. The apartment I can rent in the cloud is way better than the apartment I can rent in the real world.

• 6:19Lex Fridman?

  Well, it's all relative, isn't it? Because others will have very nice apartments too, so you'll be inferior in the virtual world.

• 6:25George Hotz?

  But that's not how I view the world, right? I don't view the world, I mean, it's a very like, like, almost zero sum-ish way to view the world. Say like, my great apartment isn't great because my neighbor has one too. No, my great apartment is great because like, look at this dishwasher, man.

• 6:40Lex Fridman?

  Yeah.

• 6:40George Hotz?

  You just touch the dish and it's washed, right? And that is great in and of itself if I have the only apartment or if everybody had the apartment. I don't care.

• 6:48Lex Fridman?

  So you have fundamental gratitude. The, the world first learned of, uh, Giohot, George Hotz, in August 2007. Maybe before then, but certainly in August 2007, when you were the first person to unlock, carry unlock an iPhone. How did you get into hacking? What was the first system you discovered vulnerabilities for and broke into?

• 7:12George Hotz?

  So, that was really kind of the first thing. I had a, I had a book in, in 2006 called Grey Hat Hacking. And I guess I realized that if you acquired these sort of powers, you could control the world. But I didn't really know that much about computers back then. I started with electronics. The first iPhone hack was physical.

• 7:40Lex Fridman?

  Hardware.

• 7:41George Hotz?

  You had to open it up and pull an address line high. And it was because I didn't really know about software exploitation. I learned that all in the next few years and I got very good at it. But back then, I knew about like, how memory chips are connected to processors and stuff.

• 7:55Lex Fridman?

  You knew about software and programming, you just didn't, didn't know. Oh, really? So you, your view of the world and computers was, uh, physical, was, was hardware?

• 8:05George Hotz?

  Actually, if you read the code that I released with that in August 2007, it's atrocious.

• 8:12Lex Fridman?

  What language was it?

• 8:13George Hotz?

  C.

• 8:13Lex Fridman?

  C, nice.

• 8:14George Hotz?

  And in a broken sort of state machine-esque C. I, I didn't know how to program.

• 8:19Lex Fridman?

  Yeah. So how did you learn to program? What was your journey? Because, I mean, we'll talk about it. You've livestreamed some of your programming. This chaotic, beautiful mess. How did you arrive at that?

• 8:32George Hotz?

  Years and years of practice. I interned at Google after, the summer after, the iPhone unlock. And I did a contract for them where I built hardware for Street View. And I wrote a software library to interact with it. And it was terrible code. And for the first time, I got feedback from people who I respected saying, like, "Don't write code like this." Now, of course, just getting that feedback is not enough. The way that I really got good was I wanted to write this thing that could emulate and then visualize ARM binaries, because I wanted to hack the iPhone better.

• 9:16George Hotz?

  And I didn't like that I couldn't see what the... I couldn't single step through the processor, because I had no debugger on there, especially for the low-level things like the boot rom and the bootloader. So I tried to build this tool to do it. And I built the tool once, and it was terrible. I built the tool second times, it was terrible. I built the tool a third time. By the time I was at Facebook, it was kind of okay. And then I built the tool a fourth time, when I was a Google intern again in 2014. And that was the first time I was like, "This is finally usable."

• 9:42Lex Fridman?

  How do you pronounce this? Kira?

• 9:43George Hotz?

  Kira, yeah.

• 9:44Lex Fridman?

  So it's essentially the most efficient way to visualize the change of state of the computer as the program is running? That's what you mean by debugger?

• 9:55George Hotz?

  Yeah, it's a timeless debugger. So you can rewind just as easily as going forward. Think about if you're using GDB, you have to put a watch on a variable if you want to see if that variable changes. In Kira, you can just click on that variable, and then it shows every single time when that variable was changed or accessed. Think about it like Git for your computer's run lock.

• 10:16Lex Fridman?

  So there's like a... A deep log of uh of the state of the computer as the program runs and you can rewind. Uh why isn't that, or maybe it is, maybe you can educate me, why isn't that kind of debugging used more often?

• 10:30George Hotz?

  Uh, cause the tooling's bad. Uh, well, two things. One, if you're trying to debug Chrome, Chrome is a 200 megabyte binary that runs slowly on desktops. So that's gonna be really hard to use for that. But it's really good to use for like CTFs and for bootroms and for small parts of code. So it's it's hard if you're trying to debug like massive systems.

• 10:52Lex Fridman?

  What's a CTF and what's a bootrom?

• 10:54George Hotz?

  The bootrom is the first code that executes the minute you give power to your iPhone.

• 10:58Lex Fridman?

  Okay.

• 10:59George Hotz?

  And CTF were these competitions that I played, capture the flag.

• 11:03Lex Fridman?

  Capture the flag, I was gonna ask you about that. What are those? Look, I watched a couple of videos on YouTube, those look fascinating. What have you learned about maybe at the high level of vulnerability of systems from these competitions?

• 11:15George Hotz?

  The, like, I feel like, like, in the heyday of CTFs, you had all of the best security people in the world challenging each other.

• 11:26Lex Fridman?

  Yeah.

• 11:26George Hotz?

  And coming up with new toy exploitable things over here and then everybody, okay, who can break it? And when you, when you break it, you get like, there's like a file on the server called flag. And then there's a program running, listening on a socket that's vulnerable. So you write an exploit, you get a shell, and then you cat flag, and then you type the flag into like a web-based scoreboard and you get points.

• 11:45Lex Fridman?

  So the goal is essentially to find an exploit in a system that allows you to run shell, to run arbitrary code on that system?

• 11:54George Hotz?

  That's one of the categories. That's like the poneable category.

• 11:58Lex Fridman?

  Um, poneable?

• 12:00George Hotz?

  Yeah, poneable. It's like, uh, you know, you pone the program, you, uh, it's a program.

• 12:05Lex Fridman?

  Uh, uh, yeah, you know, first of all, I apologize, I'm gonna, I'm gonna say it's because I'm Russian, but maybe you can help educate me, um.

• 12:16George Hotz?

  Some video game like misspelled own way back in the day?

• 12:19Lex Fridman?

  Yeah, and there's just, uh, I wonder if there's a definition, I'll have to go to Urban Dictionary for it.

• 12:24George Hotz?

  Be interesting to see what it says.

• 12:26Lex Fridman?

  Okay, so what was the heyday of CTF, uh, by the way, but was it, what decade are we talking about?

• 12:31George Hotz?

  I think like, I mean, maybe I'm biased because it's the era that, that, that I played. But like 2011 to 2015.

• 12:43George Hotz?

  Because the modern CTF scene is similar to the modern competitive programming scene. You have people who like do drills. You have people who practice. And then, once you've done that, you've turned it less into a game of generic computer skill and more into a game of, okay, you memorize, you drill on these five categories.

• 13:02George Hotz?

  And then, before that, it wasn't, it didn't have like as much attention as it had.

• 13:09George Hotz?

  I don't know, they were like, I won $30,000 once in Korea for one of these competitions.

• 13:12Lex Fridman?

  Holy crap.

• 13:13George Hotz?

  Yeah, they were, they were.

• 13:14Lex Fridman?

  So that means, I mean, money is money, but that means there was probably good people there.

• 13:18George Hotz?

  Exactly, yeah.

• 13:20Lex Fridman?

  Are the challenges human constructed? Or are they grounded in some real flaws in real systems?

• 13:27George Hotz?

  Usually, they're human constructed, but they're usually inspired by real flaws.

• 13:32Lex Fridman?

  What kind of systems, I imagine, is really focused on mobile? Like, what has vulnerabilities these days? Is it primarily mobile systems like Android?

• 13:41George Hotz?

  Oh, everything does still. Yeah, of course. The price has kind of gone up because less and less people can find them. And what's happened in security is now if you want to like jailbreak an iPhone, you don't need one exploit anymore, you need nine.

• 13:54Lex Fridman?

  Nine chained together, what do you mean? Yeah, wow, okay, so it's really, what's the, what's the benefit? Speaking higher level philosophically about hacking. I mean, it sounds from everything I've seen about you, you just love the challenge and you don't want to do anything. You don't want to bring that exploit out into the world and doing the actual,

• 14:17Lex Fridman?

  let it run wild. You just want to solve it and then you go on to the next thing.

• 14:21George Hotz?

  Oh, yeah, I mean, doing criminal stuff's not really worth it. And I'll actually use the same argument for why I don't do defense for why I don't do crime. If you want to defend a system, say the system has 10 holes, right? If you find nine of those holes as a defender, you still lose because the attacker gets in through the last one. If you're an attacker, you only have to find one out of the 10. But if you're a criminal, if you log on with a VPN nine out of the 10 times, but one time you forget, you're done.

• 14:54Lex Fridman?

  Because you're caught, okay.

• 14:55George Hotz?

  Because you only have to mess up once to be caught as a criminal?

• 14:59Lex Fridman?

  Yeah.

• 14:59George Hotz?

  That's why I'm not a criminal.

• 15:02Lex Fridman?

  But okay, let me, uh, cause I was having a discussion with somebody, uh, just at a, at a high level about nuclear weapons, actually, why we're having blown ourselves up yet. And my feeling is all the smart people in the world, if you look at the distribution of, of smart people, smart people are generally good. And then this other person, I was talking to Sean Carroll, the physicist, and he was saying, no, good and bad people are evenly distributed amongst everybody. My sense was good hackers are in general, good people, and they don't want to mess with the world. What's your sense?

• 15:38George Hotz?

  I'm not even sure about that. Like, I have a nice life. Crime wouldn't get me anything. But if you're good and you have these skills, you probably have a nice life too, right?

• 15:55Lex Fridman?

  Right, you can use the farther things, but is there an ethical, is there some, is there a little voice in your head that says, uh, well, yeah, if you could hack something to where you could hurt people? And you could earn a lot of money doing it though, not hurt physically perhaps, but disrupt their life in some kind of way. Isn't there a little voice that says?

• 16:19George Hotz?

  Um, well, two things. One, I don't really care about money. So like the money wouldn't be an incentive. The thrill might be an incentive, but when I was 19, I read Crime and Punishment.

• 16:30Lex Fridman?

  Right.

• 16:30George Hotz?

  And that was another, that was another great one that talked me out of ever really doing crime. Um, cause it's like, that's gonna be me. I'd get away with it, but it would just run through my head, even if I got away with it, you know? And then you do crime for long enough, you'll never get away with it.

• 16:45Lex Fridman?

  That's right, in the end. That's a good reason to be good.

• 16:49George Hotz?

  I wouldn't say I'm good, I would just say I'm not bad.

• 16:51Lex Fridman?

  You're a talented programmer and a hacker in a good, positive sense of the word. You've, uh, played around, found vulnerabilities in various systems. What have you learned broadly about the design of systems and so on from that, from that whole process?

• 17:09George Hotz?

  You learn to not take things for what people say they are, but you look at things for what they actually are.

• 17:23Lex Fridman?

  Yeah.

• 17:24George Hotz?

  I understand that's what you tell me it is, but what does it do?

• 17:27Lex Fridman?

  Right. And you have nice visualization tools to really know what it's really doing.

• 17:33George Hotz?

  Oh, I wish. I'm a better programmer now than I was in 2014. I said, Kira, that was the first tool that I wrote that was usable. I wouldn't say the code was great. I still wouldn't say my code is great.

• 17:44Lex Fridman?

  So how was your evolution as a programmer, except practice? You want, you started with C. At which point did you pick up Python? Because you're pretty big in Python now.

• 17:53George Hotz?

  Now, yeah, in, uh, in college, uh, I went to Carnegie Mellon when I was 22. Um, I went back, I'm like, all right, I'm gonna take all your hardest CS courses. And we'll see how I do, right? Like, did I miss anything by not having a, a real undergraduate education? Took, uh, operating systems, compilers, AI, and they're like, uh, freshman Weeder Math course.

• 18:15George Hotz?

  And.

• 18:17Lex Fridman?

  Operating systems, some of those, some of those classes you mentioned, uh, pretty tough, actually.

• 18:20George Hotz?

  They're great. At least one of the 2012, circa 2012, operating systems and compilers, uh, were two of the, they were the best classes I've ever taken in my life. Because you write an operating system and you write a compiler. I wrote my operating system in C and I wrote my compiler in Haskell, but somehow, somehow I picked up Python, uh, that semester as well. I started using it for the CTFs, actually. That's when I really started to get into CTFs. And, uh, CTFs, you're all, it's a race against the clock, so I can't write things in C.

• 18:51Lex Fridman?

  Oh, there's a clock component, so you really want to use the programming language just so you can be fastest.

• 18:55George Hotz?

  Forty-eight hours, uh, pone as many of these challenges as you can.

• 18:57Lex Fridman?

  Pone.

• 18:58George Hotz?

  Yeah. You got like a hundred points a challenge, whatever team gets the most.

• 19:02Lex Fridman?

  You were, uh, both at Facebook and Google for a brief stint.

• 19:06George Hotz?

  Yeah.

• 19:07Lex Fridman?

  With, uh, Project Zero, actually, at Google for five months, where you developed, Kira. What was Project Zero about in general? What, uh, I'm just curious about the security efforts in these companies.

• 19:21George Hotz?

  Well, Project Zero started the same time I, I, I went there. What, what, what years were you there?

• 19:27Lex Fridman?

  2015.

• 19:28George Hotz?

  2015, so that was right at the beginning of, of, of Project, it's small. It's Google's offensive security team.

• 19:38George Hotz?

  I'll try to give, I'll try to give the best public facing explanation that I can. So

• 19:45George Hotz?

  the idea is basically these vulnerabilities exist in the world. Nation states have them. Some high powered bad actors have them.

• 19:56George Hotz?

  Sometime

• 19:59George Hotz?

  people will find these vulnerabilities and submit them in bug bounties to the companies. But a lot of the companies don't really care. They don't even fix the bug. There's no, it doesn't hurt for there to be a vulnerability. So Project Zero is like, we're gonna do it different. We're going to announce a vulnerability and we're going to give them 90 days to fix it. And then whether they fix it or not, we're gonna drop the, drop the zero day.

• 20:19Lex Fridman?

  Oh, wow.

• 20:20George Hotz?

  We're gonna drop the weapon.

• 20:21Lex Fridman?

  That's so cool. That is so cool. I love that deadlines. Oh, that's so cool.

• 20:26George Hotz?

  Give them real deadlines.

• 20:27Lex Fridman?

  Yeah.

• 20:28George Hotz?

  And I think it's done a lot for moving the industry forward.

• 20:32Lex Fridman?

  I watched your coding sessions on the streamed online. You code things up, the basic projects, usually from scratch. I would say sort of as a programmer myself, just watching you, that you type really fast and your brain works in both brilliant and chaotic ways. I don't know if that's always true, but certainly for the live streams. So it's, it's interesting to me because I'm more, I'm much slower and systematic and careful. And you just move, I mean, probably in order of magnitude faster. So I'm curious, is there a method to your madness? Is this, or is this just who you are?

• 21:09George Hotz?

  There's pros and cons. There's pros and cons to my programming style. And I'm aware of them. Like, if you ask me to like, like, get something up and working quickly with like an API that's kind of undocumented, I will do this super fast because I will throw things at it until it works. If you ask me to take a vector and rotate it 90 degrees and then flip it over the X Y plane.

• 21:35George Hotz?

  I'll spam program for two hours and won't get it.

• 21:38Lex Fridman?

  Oh, because it's something that you could do with a sheet of paper, think through, design, and then just, you really just throw stuff at the wall and you get so good at it that it usually works.

• 21:51George Hotz?

  I should become better at the other kind as well. Sometimes I'll do things methodically. It's nowhere near as entertaining on the Twitch streams. I do exaggerate it a bit on the Twitch streams as well. The Twitch streams, I mean, what do you want to see a gamer? You want to see actions per minute, right? I'll show you APM for programming, too.

• 22:04Lex Fridman?

  Yeah, I recommend people go to it. I think I watched, I watched probably several hours of you. Like, I've actually left you programming in the background while I was programming because you made me, it was like watching a really good gamer. It's like energizes you because you're like moving so fast. It's so, it's awesome. It's inspiring. It made me jealous that like, because my own programming is inadequate in terms of speed. Because I was like.

• 22:33George Hotz?

  So I'm twice as frantic on the live streams as I am when I code without them.

• 22:39Lex Fridman?

  It's super entertaining. So I wasn't even paying attention to what you were coding, which is great. It's just watching you switch windows and Vim, I guess, is the most.

• 22:47George Hotz?

  Yeah, there's Vim in screen. I've developed the workflow at Facebook and stuck with it.

• 22:52Lex Fridman?

  How do you learn new programming tools, ideas, techniques these days? What's your methodology for learning new things?

• 22:58George Hotz?

  So I wrote for comma, the distributed file systems out in the world are extremely complex. Like, if you want to install something like Ceph, Ceph is, I think, the open infrastructure distributed file system. Or there's newer ones like Seaweed FS, but these are all like 10,000-plus line projects. I think some of them are even 100,000 line. And just configuring them was a nightmare. So I wrote, uh, I wrote one. It's 200 lines and it's, it uses like NGINX to the line servers and has this little master server that I wrote in Go. And the way I

• 23:39Lex Fridman?

  In Go, wow.

• 23:40George Hotz?

  This, if I would say that I'm proud per line of any code I wrote, maybe there's some exploits that I think are beautiful. And then this, this is 200 lines and just the way that I thought about it, I think was very good. And the reason it's very good is because that was the fourth version of it that I wrote and I had three versions that I threw away.

• 23:55Lex Fridman?

  You mentioned, did you say Go?

• 23:57George Hotz?

  I wrote in Go, yeah.

• 23:58Lex Fridman?

  In Go.

• 23:58George Hotz?

  So I

• 23:59Lex Fridman?

  Is that a functional language? I forget what Go is.

• 24:01George Hotz?

  Go is Google's language.

• 24:03Lex Fridman?

  Right.

• 24:04George Hotz?

  Um, it's not functional. It's some, it's like, in a way it's C++ but easier. It's, it's strongly typed. It has a nice ecosystem around it. When I first looked at it, I was like, this is like Python, but it takes twice as long to do anything.

• 24:21Lex Fridman?

  Yeah.

• 24:22George Hotz?

  Now that I've, OpenPilot is migrating to C, but it still has large Python components. I now understand why Python doesn't work for large code bases and why you want something like Go.

• 24:32Lex Fridman?

  Interesting. So why, why doesn't Python work for, so even most, uh, speaking for myself at least, like we do a lot of stuff, basically demo level work with autonomous vehicles and most of the work is Python.

• 24:44George Hotz?

  Yeah.

• 24:45Lex Fridman?

  Why doesn't Python work for large code bases?

• 24:48George Hotz?

  Because, well, lack of type checking is, is, is a big

• 24:54Lex Fridman?

  So errors creep in.

• 24:55George Hotz?

  Yeah. And like, you don't know, the, the compiler can tell you like nothing, right? So everything is either, uh, you know, like, like syntax errors, fine. But if you misspell a variable in Python, compiler won't catch that. There's like linters that can catch it some of the time. There's no types is really the biggest downside. And then, well, Python's slow, but that's not related to it. Well, maybe it's kind of related to it, it's lack of.

• 25:21Lex Fridman?

  So what's, what's in your toolbox these days? Is it Python or what else? Go?

• 25:24George Hotz?

  I need to move to something else. My, my adventure into dependently typed languages. I love these languages. They just have like syntax from the eighties.

• 25:34Lex Fridman?

  What do you think about JavaScript? ES6, like the modern TypeScript?

• 25:40George Hotz?

  JavaScript is The whole ecosystem is unbelievably confusing.

• 25:45Lex Fridman?

  Right.

• 25:45George Hotz?

  NPM updates a package from 022 to 025 and that breaks your Babel linter, which translates your ES5 into ES6, which doesn't run on. So why do I have to compile my JavaScript again?

• 25:58Lex Fridman?

  It may be the future, though. You think about, I mean, I've embraced JavaScript recently just because, just like I've continually embraced PHP. It seems that these worst possible languages live on for the longest. Like cockroaches never die.

• 26:13George Hotz?

  Yeah. Well, it's in the browser and it's fast.

• 26:17Lex Fridman?

  It's fast.

• 26:18George Hotz?

  Yeah.

• 26:18Lex Fridman?

  It's in the browser and compute might stay, become, you know, the browser. It's unclear what the role of the browser is in terms of distributed computation in the future. So.

• 26:30George Hotz?

  JavaScript is definitely here to stay.

• 26:31Lex Fridman?

  Yeah. It's interesting if autonomous vehicles will run on JavaScript one day. I mean, you have to consider these possibilities.

• 26:38George Hotz?

  All our debug tools are JavaScript. We actually just open source them. We have a tool explorer, which you can annotate your disengagements. And we have a tool Cabana, which lets you analyze the CAN traffic from the car.

• 26:49Lex Fridman?

  So basically anytime you're visualizing something about the log, you're using JavaScript.

• 26:54George Hotz?

  Well, the web is the best UI toolkit by far.

• 26:56Lex Fridman?

  Yep.

• 26:57George Hotz?

  So, and then, you know what? You're coding in JavaScript. We have a React guy. He's good.

• 27:00Lex Fridman?

  React, nice. Let's get into it. So let's talk autonomous vehicles. You founded Comma AI. Let's, at a high level, how did you get into the world of vehicle automation? Can you also just, for people who don't know, tell the story of Comma AI?

• 27:17George Hotz?

  Sure. So I was working at this AI startup and a friend approached me and he's like, dude, I don't know where this is going, but the coolest applied AI problem today is self-driving cars. I'm like, well, absolutely. Do you want to meet with Elon Musk? And he's looking for somebody to build a vision system for autopilot. This is when they were still on AP1. They were still using Mobileye. Elon, back then, was looking for a replacement. And he brought me in and we talked about a contract where I would deliver something that meets Mobileye level performance. I would get paid $12 million if I could deliver it tomorrow and I would lose $1 million

• 28:01George Hotz?

  for every month I didn't deliver. So I was like, okay, this is a great deal. This is a super exciting challenge. You know what? Even if it takes me 10 months, I get $2 million. It's good. Maybe I can finish up in five. Maybe I don't finish it at all and I get paid nothing and I'll work for 12 months for free.

• 28:17Lex Fridman?

  So maybe just take a pause on that. I'm also curious about this because I've been working robotics for a long time. I'm curious to see a person like you just step in and sort of somewhat naive, but brilliant, right? So that's the best place to be because you basically full steam take on a problem. How confident, how from that time, because you know a lot more now, at that time, how hard do you think it is to solve all autonomous driving?

• 28:42George Hotz?

  I remember I suggested to Elon in the meeting, I'm putting a GPU behind each camera to keep the compute local. This is an incredibly stupid idea. I leave the meeting 10 minutes later and I'm like, I could have spent a little bit of time thinking about this problem.

• 28:59Lex Fridman?

  Why is it a stupid idea?

• 29:00George Hotz?

  Oh, just send all your cameras to one big GPU. You're much better off doing that.

• 29:04Lex Fridman?

  Oh, sorry. You said behind every camera, have a GPU.

• 29:07George Hotz?

  Have a small GPU. I was like, oh, I'll put the first few layers of my com there. Like, why'd I say that?

• 29:12Lex Fridman?

  That's possible.

• 29:13George Hotz?

  It's possible, but it's a bad idea.

• 29:15Lex Fridman?

  It's not obviously a bad idea.

• 29:17George Hotz?

  Pretty obviously bad. But whether it's actually a bad idea or not, I left that meeting with Elon, like beating myself up. I'm like, why did I say something stupid?

• 29:23Lex Fridman?

  Yeah, you haven't, like you haven't at least like thought through every aspect fully, yeah.

• 29:28George Hotz?

  He's very sharp too. Like usually in life, I get away with saying stupid things and then kind of course, oh, right away he called me out about it. And like usually in life, I get away with saying stupid things and then like people will, you know, a lot of times people don't even notice. And I'll like correct it and bring the conversation back. But with Elon, it was like, nope. Like, okay, well, that's not at all why the contract fell through. I was much more prepared the second time I met him.

• 29:52Lex Fridman?

  Yeah. But in general, how hard did you think it is? Like 12 months is a tough timeline.

• 30:00George Hotz?

  Oh, I just thought I'd clone Mobileye IQ3. I didn't think I'd solve level five self-driving or anything.

• 30:04Lex Fridman?

  So the goal there was to do lane keeping, good lane keeping.

• 30:09George Hotz?

  I saw, my friend showed me the outputs from a Mobileye and the outputs from a Mobileye was just basically two lanes and a position of a lead car. I'm like, I can, I can gather a data set and train this net in weeks. And I did.

• 30:21Lex Fridman?

  Well, first time I tried the implementation of Mobileye in a Tesla, I was really surprised how good it is. It's quite incredibly good because I thought it's just because I've done a lot of computer vision, I thought it'd be a lot harder to create a system that that's stable. So I was personally surprised, you know, have to admit it because I was kind of skeptical before trying it because I thought it would go in and out a lot more, it would get disengaged a lot more and it's pretty robust. So what, how, how, how hard is the problem when you, when you tackled it?

• 30:58George Hotz?

  So I think AP1 was great, like Elon talked about disengagements on the 405 down in LA with like the landmarks are kind of faded and the Mobileye system would drop out. Like I had something up and working that I would say was like the same quality in three months.

• 31:18Lex Fridman?

  Same quality, but how do you know? You, you say stuff like that confidently, but you can't, and I love it, but the question is you can't, you're kind of going by feel because you.

• 31:30George Hotz?

  You're going by feel, absolutely, absolutely. Like, like I would take, I hadn't, I borrowed my friend's Tesla. I would take AP1 out for a drive and then I would take my system out for a drive.

• 31:38Lex Fridman?

  And seems reasonably like the same. So the 405, how hard is it to create something that could actually be a product that's deployed? I mean, I've, I've read an article where Elon just responded, said something about you saying that to build autopilot is, is more complicated than a single George Hod's level job. How hard is that job to create something that would work across globally?

• 32:14George Hotz?

  Um, I don't think globally is the challenge, but Elon followed that up by saying it's going to take two years and a company of 10 people.

• 32:21Lex Fridman?

  Yeah.

• 32:22George Hotz?

  And here I am four years later with a company of 12 people and I think we still have another two to go.

• 32:26Lex Fridman?

  Two years? So yeah, so what do you think, what do you think about how Tesla's progressing with autopilot V2, V3?

• 32:35George Hotz?

  I think we've kept pace with them pretty well.

• 32:40George Hotz?

  I think navigating autopilot is terrible. We had some demo features internally of the same stuff. And we would test it and I'm like, I'm not shipping this even as like open source software to people.

• 32:51Lex Fridman?

  Why do you think it's terrible?

• 32:53George Hotz?

  Consumer Reports does a great job of describing it. Like when it makes a lane change, it does it worse than a human. You shouldn't ship things like autopilot, open pilot, they lane keep better than a human. If you turn it on for a stretch of highway, like an hour long, it's never going to touch a lane line. Human will touch probably a lane line twice.

• 33:15Lex Fridman?

  You just inspired me. I don't know if you're grounded in data on that.

• 33:18George Hotz?

  I read your paper.

• 33:19Lex Fridman?

  Okay, but no, but that's interesting.

• 33:23Lex Fridman?

  I wonder actually how often we touch lane lines in general, like a little bit, because it is.

• 33:29George Hotz?

  I could answer that question pretty easily with the common data set.

• 33:32Lex Fridman?

  Yeah, I'm curious.

• 33:33George Hotz?

  I've never answered it. I don't know. I just, too, is like my personal.

• 33:36Lex Fridman?

  It feels, it feels right. That's interesting because every time you touch a lane, that's a source of a little bit of stress. And kind of lane keeping is removing that stress. That's ultimately the big, the biggest value add, honestly, is just removing the stress of having to stay in lane. And I think, honestly, I don't think people fully realize, first of all, that that's a big value add. But also that that's all it is.

• 34:01George Hotz?

  And that, not only, I find it a huge value add. I drove down, when we moved to San Diego, I drove down in an enterprise or antecar. And I missed it. So I missed having the system so much. It's so much more tiring to drive without it. It's, it is that lane centering that's the key feature.

• 34:21Lex Fridman?

  Yeah.

• 34:23George Hotz?

  And in a way, it's the only feature that actually adds value to people's lives and autonomous vehicles today. Waymo does not add value to people's lives. It's a more expensive, slower Uber. Maybe someday it'll be this big cliff where it adds value, but I don't usually.

• 34:35Lex Fridman?

  It is fascinating. I haven't talked to, this is good, because I haven't. I have intuitively, but I think we're making it explicit now. I, I actually believe um that really good lane keeping is a reason to buy a car, will be a reason to buy a car, is a huge value add. I've never, until we just started talking about it, I haven't really quite realized that. That I, I've felt with Elon's chase of level four is not the correct chase. It was on, because you should just say Tesla has the best, as if from a Tesla perspective, say, Tesla has the best lane keeping. Kama AI should say Kama AI is the best lane keeping.

• 35:20Lex Fridman?

  And that is it.

• 35:22George Hotz?

  Yeah.

• 35:22Lex Fridman?

  Yeah. So do you think?

• 35:24George Hotz?

  Um, you have to do the longitudinal as well. You can't just lane keep. You have to do ACC, but ACC is much more forgiving than lane keep. Um, especially on the highway.

• 35:33Lex Fridman?

  By the way, are you um Kama AI's camera only, correct?

• 35:38George Hotz?

  Uh, no, we use the radar. Uh, we.

• 35:40Lex Fridman?

  From the car, you're able to get the, okay.

• 35:43George Hotz?

  Um, we can do a camera only now. It's gotten to the point, but we leave the radar there as like a, it's, it's fusion now.

• 35:49Lex Fridman?

  Okay, so let's maybe talk through some of the system specs on the hardware. What, what's, what's the hardware side of uh, what you're providing? What's the capabilities on the software side with?

• 36:01George Hotz?

  Sure.

• 36:01Lex Fridman?

  Open Pilot and so on?

• 36:03George Hotz?

  So Open Pilot

• 36:06George Hotz?

  as the, the, the box that we sell, that it runs on, it's a phone in a plastic case. Um, it's nothing special. We sell it without the software. So you're like, you know, you buy the phone, it's just easy. It'll be easy setup, but it's sold with no software.

• 36:20George Hotz?

  Open Pilot right now is about to be 0.6. When it gets to 1.0, I think we'll be ready for a consumer product. We're not going to add any new features. We're just going to make the lane keeping really, really good.

• 36:30Lex Fridman?

  Okay, I got it.

• 36:32George Hotz?

  So what do we have right now? It's a Snapdragon 820.

• 36:36George Hotz?

  Say Sony IMX 298 forward facing camera, driver monitoring camera, just a selfie cam on the phone. And a CAN transceiver, maybe there's a thing called pandas. And they talk over USB to the phone. And then they have three CAN buses that they talk to the car. One of those CAN buses is the radar CAN bus. One of them is the main car CAN bus. And the other one is the proxy camera CAN bus. We leave the existing camera in place. So we don't turn AEB off. Right now, we still turn AEB off. If you're using our longitudinal, but we're going to fix that before 1.0.

• 37:12Lex Fridman?

  Got it. Wow, that's cool. So and it's can both ways. So how are you able to control vehicles?

• 37:19George Hotz?

  So we proxy. The vehicles that we work with already have a lane keeping assist system. So lane keeping assist can mean a huge variety of things. It can mean it will apply a small torque to the wheel after you've already crossed a lane line by a foot, which is the system in the older Toyotas. Versus like, I think Tesla still calls it lane keeping assist, where it'll keep you perfectly in the center of the lane on the highway.

• 37:48Lex Fridman?

  You can control like you with a joystick, the cars. So these cars already have the capability of drive by wire. So is it?

• 37:57Lex Fridman?

  Is it trivial to convert a car that it operates with? It open pilot is able to control the steering?

• 38:04George Hotz?

  Oh, a new car or a car that we so we have support now for 45 different makes of cars.

• 38:09Lex Fridman?

  What are what are the cars in general?

• 38:11George Hotz?

  Mostly Hondas and Toyotas. We support almost every Honda and Toyota made this year.

• 38:18George Hotz?

  And then a bunch of GMs, a bunch of Subarus.

• 38:21Lex Fridman?

  It doesn't have to be like a Prius. It could be Corolla as well.

• 38:24George Hotz?

  The 2020 Corolla is the best car with open pilot. It just came out there. The actuator has less lag than the older Corolla.

• 38:32Lex Fridman?

  I think I started watching a video with your, I mean, the way you make videos is awesome. Literally at the dealership streaming.

• 38:41George Hotz?

  I had my friend the phone, I'm like, you want to stream for an hour?

• 38:44Lex Fridman?

  Yeah, and basically, like, if stuff goes a little wrong, you're just like, you just go with it. Yeah, I love it.

• 38:50George Hotz?

  Well, it's real.

• 38:51Lex Fridman?

  Yeah, it's real. That's, that's, it's, that's so beautiful and it's so in contrast to the way other companies would put together a video like that.

• 39:01George Hotz?

  Kind of why I like to do it like that.

• 39:02Lex Fridman?

  Good. I mean, if you become super rich one day and successful, I hope you keep it that way because I think that's actually what people love, that kind of genuine.

• 39:11George Hotz?

  Oh, it's all that has value to me.

• 39:12Lex Fridman?

  Yeah.

• 39:13George Hotz?

  Money has no, if I sell out to like make money, I sold out, it doesn't matter. What do I get? A yacht? I don't want a yacht.

• 39:21Lex Fridman?

  And I think Tesla's actually has a small inkling of that as well with autonomy day. They did reveal more than, I mean, of course, there's marketing communications, you could tell, but it's more than most companies would reveal, which is, I hope they go towards that direction more, other companies, GM, Ford.

• 39:39George Hotz?

  Oh, Tesla's gonna win level five. They really are.

• 39:43Lex Fridman?

  So let's talk about it. You think, you're focused on level two? Currently, currently?

• 39:49George Hotz?

  We're gonna be one to two years behind Tesla getting to level five.

• 39:53Lex Fridman?

  Okay.

• 39:54George Hotz?

  We're Android, right? We're Android.

• 39:56Lex Fridman?

  You're Android.

• 39:57George Hotz?

  I'm just saying, once Tesla gets it, we're one to two years behind. I'm not making any timeline on when Tesla's gonna get it.

• 40:02Lex Fridman?

  You did, that's brilliant.

• 40:03George Hotz?

  I'm sorry Tesla investors, if you think you're gonna have an autonomous robot taxi fleet by the end of the year.

• 40:08Lex Fridman?

  Yes.

• 40:09George Hotz?

  I'll bet against that.

• 40:11Lex Fridman?

  So what do you think about this? The most level four companies

• 40:18Lex Fridman?

  are kind of just doing their usual safety driver, doing full autonomy kind of testing, and then Tesla does, basically trying to go from lane keeping to full autonomy. What do you think about that approach? How successful would it be?

• 40:34George Hotz?

  It's a ton better approach. Because Tesla is gathering data on a scale that none of them are. They're putting real users behind the, behind the wheel of the cars. It's, I think, the only strategy that works. The incremental.

• 40:51Lex Fridman?

  Well, so there's a few components to Tesla approach that's that's more than just the incremental. What you spoke with is the one is the software. So over the air air software updates.

• 41:00George Hotz?

  Necessity. I mean, Waymo have cruise have those too. Those aren't.

• 41:04Lex Fridman?

  But.

• 41:04George Hotz?

  Those differentiate them from the automakers.

• 41:06Lex Fridman?

  Right. No lane keeping assist systems have no cars or lane keeping system have that except Tesla.

• 41:11George Hotz?

  Yeah.

• 41:12Lex Fridman?

  And the other one is the data, the other direction, which is the ability to query the data. I don't think they're actually collecting as much data as people think, but the ability to turn on collection and turn it off.

• 41:26Lex Fridman?

  So I'm both in the robotics world and the the psychology, human factors world. Many people believe that level two autonomy is problematic because of the human factor. Like the more the task is automated, the more there's a vigilance decrement. You start to fall asleep, you start to become complacent, start texting more and so on. Do you worry about that? Because if we're talking about transition from lane keeping to full autonomy, if you're spending 80% of the time not supervising the machine, do you worry about what that means?

• 42:02George Hotz?

  Two things.

• 42:02Lex Fridman?

  Safety of the drivers?

• 42:03George Hotz?

  One, we don't consider open pilot to be 1.0 until we have 100% driver monitoring. You you can cheat right now. Our driver monitoring system, there's a few ways to cheat it. They're pretty obvious. We're working on making that better. Before we ship a consumer product that can drive cars, I want to make sure that I have driver monitoring that you can't cheat.

• 42:22Lex Fridman?

  What's like a successful driver monitoring system look like? It's keep, is it all about just keeping your eyes on the road?

• 42:27George Hotz?

  Um, well, a few things. So that's what we went with at first for driver monitoring. I'm checking. I'm actually looking at where your head is looking. The camera's not that high resolution. Eyes are a little bit hard to get.

• 42:38Lex Fridman?

  Well, head is is big. I mean, that's.

• 42:39George Hotz?

  Head is head is good. And actually, a lot of it, just psychology wise, to have that monitor constantly there, it reminds you that you have to be paying attention. But we want to go further. We just hired someone full time to come on to do the driver monitoring. I want to detect phone in frame. And I want to make sure you're not sleeping.

• 42:59Lex Fridman?

  How much does the camera see of the body?

• 43:01George Hotz?

  This one, not enough.

• 43:04Lex Fridman?

  Not enough.

• 43:04George Hotz?

  The next one, everything.

• 43:07Lex Fridman?

  Well, it's interesting, fisheye, because we're doing just data collection, not real time. But fisheye is a beautiful, being able to capture the body. And the smartphone is really like the biggest problem.

• 43:19George Hotz?

  I'll show you, I can show you one of the pictures from our new system.

• 43:24Lex Fridman?

  Awesome. So you're basically saying the driver monitoring will be the answer to that.

• 43:29George Hotz?

  Um, I think the other point that you raised in your paper is good as well. You're not asking a human to supervise a machine without giving them the, they can take over at any time.

• 43:39Lex Fridman?

  Right.

• 43:40George Hotz?

  Our safety model, you can take over, we disengage on both the gas or the brake. We don't disengage on steering, I don't feel you have to. But we disengage on gas or brake, so it's very easy for you to take over. And it's very easy for you to reengage. That switching should be super cheap.

• 43:55Lex Fridman?

  Yep.

• 43:55George Hotz?

  The cars that require, even autopilot requires a double press. That's almost, I see, I don't like that.

• 44:00Lex Fridman?

  Yeah.

• 44:00George Hotz?

  And then the cancel, to cancel in autopilot, you either have to press cancel, which no one knows what that is. So they press the brake. But a lot of times you don't have to want to press the brake. You want to press the gas, so you should cancel on gas. Or wiggle the steering wheel, which is bad as well.

• 44:14Lex Fridman?

  Wow, that's brilliant. I haven't heard anyone articulate that point.

• 44:17George Hotz?

  Oh, this is all I think about.

• 44:19Lex Fridman?

  It's the, because I think, I think actually Tesla has done a better job than most automakers at making that frictionless. But you just described that it could be even better.

• 44:33George Hotz?

  I love Super Cruise as an experience, once it's engaged.

• 44:37Lex Fridman?

  Yeah.

• 44:37George Hotz?

  I don't know if you've used it, but getting the thing to try to engage.

• 44:40Lex Fridman?

  Yeah. Yeah, I've used the, I've driven Super Cruise a lot. So what's your thoughts on the Super Cruise system in general?

• 44:46George Hotz?

  You disengage Super Cruise and it falls back to ACC, so my car is like still accelerating. It feels weird. Otherwise, when you actually have Super Cruise engaged on the highway, it is phenomenal. We bought that Cadillac, we just sold it. But we bought it just to like experience this. And I wanted everyone in the office to be like, this is what we're striving to build. GM pioneering with the driver monitoring.

• 45:08Lex Fridman?

  You like their driver monitoring system?

• 45:11George Hotz?

  It has some bugs. If there's a sun shining back here, it'll be blind to you.

• 45:18George Hotz?

  But overall, mostly, yeah.

• 45:19Lex Fridman?

  That's so cool that you know all this stuff. That's, uh, I don't, uh, often talk to people that, because it's such a rare car, unfortunately, currently.

• 45:27George Hotz?

  We bought one explicitly for us. We, we lost like, like 25K in the deprecation, but I feel it's worth it.

• 45:33Lex Fridman?

  I was very pleasantly surprised that our GM system was so innovative. And really, uh, that wasn't advertised much, wasn't talked about much.

• 45:43George Hotz?

  Yeah.

• 45:44Lex Fridman?

  And I was nervous that it would die, that it would disappear.

• 45:48George Hotz?

  Well, I did, they put it on the wrong car. They should have put it on the Bolt and not some weird Cadillac that nobody bought.

• 45:53Lex Fridman?

  I think that's gonna be into, they're saying at least it's gonna be into their entire fleet. So what do you think about, as long as we're on the driver monitoring, uh, what do you think about Elon Musk's claim that driver monitoring is not needed?

• 46:08George Hotz?

  Normally, I love his claims. That one is stupid. That one is stupid and, you know, he's not gonna have his level five fleet by the end of the year. Hopefully, he's like, okay, I was wrong. I'm gonna add driver monitoring. Because when these systems get to the point that they're only messing up once every thousand miles, you absolutely need driver monitoring.

• 46:30Lex Fridman?

  So let me play, cause I agree with you, but let me play devil's advocate. One possibility is that without driver monitoring, people are able to monitor, uh, self-regulate, monitor themselves, you know, that, so your idea is...

• 46:47George Hotz?

  You've seen all the people sleeping in Teslas?

• 46:50Lex Fridman?

  Yeah, well, I'm a little skeptical of all the people sleeping in Teslas because, um,

• 46:58Lex Fridman?

  I, I've, I've stopped paying attention to that kind of stuff because I wanna see real data. There's too much glorified, it doesn't feel scientific to me. So I wanna know, you know, what, how many people are really sleeping in Teslas versus sleeping? I've, I was driving here, sleep deprived, in a car with no automation. I was falling asleep.

• 47:17George Hotz?

  I agree that it's hypey. It's just like, you know what? If Elon put driver monitoring, I, I rented a, my last autopilot experience was I rented a Model 3 in March and drove it around. The wheel thing is annoying. And the reason the wheel thing is annoying, we use the wheel thing as well, but we don't disengage on wheel. For Tesla, you have to touch the wheel just enough to trigger the torque sensor to tell it that you're there, but not enough as to disengage it. Which, don't use it for two things. Don't disengage on wheel. You don't have to.

• 47:48Lex Fridman?

  That whole experience. Wow, beautifully put. All those elements, even if you don't have driver monitoring, that whole experience needs to be better.

• 47:57George Hotz?

  Driver monitoring, I think would make, I mean, I think Supercruise is a better experience once it's engaged over autopilot. I think Supercruise's transition to engagement and disengagement are significantly worse.

• 48:11Lex Fridman?

  There's a tricky thing because if I were to criticize Supercruise, it's a little too crude and I think like six seconds or something, if you look off-road, it'll start warning you. It's some ridiculously long period of time. And just the way, I think it's basically, it's a binary.

• 48:32George Hotz?

  It should be adaptive.

• 48:33Lex Fridman?

  Yeah, it needs to learn more about you and needs to communicate what it sees about you more. I'm not, you know, Tesla shows what it sees about the external world. It would be nice if Supercruise would tell us what it sees about the internal world.

• 48:47George Hotz?

  It's even worse than that. You press the button to engage and it just says Supercruise unavailable. Why?

• 48:52Lex Fridman?

  Why? Yeah, that transparency is good.

• 48:57George Hotz?

  We've renamed the driver monitoring packet to driver state.

• 49:01Lex Fridman?

  Driver state.

• 49:02George Hotz?

  We have car state packet, which has the state of the car, and we have driver state packet, which has the state of the driver.

• 49:07Lex Fridman?

  So what are the?

• 49:08George Hotz?

  Estimate their BAC.

• 49:10Lex Fridman?

  What's BAC?

• 49:11George Hotz?

  Blood alcohol content?

• 49:13Lex Fridman?

  You think that's possible with computer vision?

• 49:15George Hotz?

  Absolutely.

• 49:19Lex Fridman?

  To me, it's an open question. I haven't looked into it too much. Actually, I quite seriously looked at the literature. It's not obvious to me that from the eyes and so on, you can tell.

• 49:27George Hotz?

  You might need stuff in the car as well. You might need how they're controlling the car, right? And that's fundamentally, at the end of the day, what you care about. But I think, especially when people are really drunk, they're not controlling the car nearly as smoothly as they would look at them walking, right? The car is like an extension of the body. So I think you could totally detect. And if you could fix people who are drunk, distracted, asleep, if you fix those three.

• 49:49Lex Fridman?

  Yeah, that's huge. So what are the current limitations of open pilot? What are the main problems that still need to be solved?

• 49:58George Hotz?

  We're hopefully fixing a few of them in zero six. We're not as good as autopilot at stop cars. Um. So if you're coming up to a red light at like fifty five. Um. So it's the radar stopped car problem, which is responsible for two autopilot accidents. Uh. It's hard to differentiate a stopped car from a uh like signpost.

• 50:20Lex Fridman?

  Yeah, static object.

• 50:21George Hotz?

  Um. So you have to fuse. You have to do this visually. There's no way from the radar data to tell the difference. Maybe you can make a map, but I don't really believe in mapping at all anymore. Um.

• 50:30Lex Fridman?

  Wait, wait, wait. What? You don't believe in mapping?

• 50:32George Hotz?

  No.

• 50:33Lex Fridman?

  So you basically, the open pilot solution is saying react to the environment as you see it, just like a human doing beings do.

• 50:40George Hotz?

  And then eventually, when you want to do navigate on uh open pilot, I'll train the net to look at ways. I'll run ways in the background. I'll train a comp set of ways.

• 50:48Lex Fridman?

  Are you using GPS at all?

• 50:51George Hotz?

  We use it to ground truth. We use it to very carefully ground truth the paths. Uh. We have a stack which can recover relative to ten centimeters over one minute. Um. And then we use that to ground truth exactly where the car went in that local part of the environment, but it's all local.

• 51:05Lex Fridman?

  How are you testing in general, just for yourself, like experiments and stuff? Right. Uh. Where are you? Where are you located?

• 51:11George Hotz?

  San Diego.

• 51:12Lex Fridman?

  San Diego.

• 51:12George Hotz?

  Yeah.

• 51:13Lex Fridman?

  Okay. Uh. What? So you basically drive around there, then collect some data and and watch performance?

• 51:19George Hotz?

  We have a simulator now, and we have our simulator is really cool. Our simulator is not uh, it's not like a unity based simulator. Our simulator lets us load in real state.

• 51:29Lex Fridman?

  What do you mean?

• 51:30George Hotz?

  We can load in a drive and simulate what the system would have done on the historical data.

• 51:36Lex Fridman?

  Ooh, nice.

• 51:39Lex Fridman?

  Interesting. So what? Yeah.

• 51:40George Hotz?

  Right now, we're only using it for testing, but as soon as we start using it for training, that's it.

• 51:45Lex Fridman?

  That's for testing. What's your feeling about the real world versus simulation? Do you like simulation for training if if this moves to training?

• 51:52George Hotz?

  So we have to distinguish two types of simulators, right? There's a simulator that like is completely fake. I could get my car to drive around in GTA. Um. I feel that this kind of simulator is useless. You're never, there there's so many. Uh. My analogy here is like, okay, fine. You're not solving the computer vision problem, but you're solving the computer graphics problem.

• 52:18Lex Fridman?

  Right. And you don't think you can get very far by creating ultra realistic. Graphics.

• 52:24George Hotz?

  No, because you can create ultra-realistic graphics of the road. Now create ultra-realistic behavioral models of the other cars. Oh, well, I'll just use my self-driving. No, you won't. You need real, you need actual human behavior. Because that's what you're trying to learn. The driving does not have a spec. The definition of driving is what humans do when they drive. Whatever Waymo does, I don't think it's driving.

• 52:49Lex Fridman?

  Right, well, I think actually Waymo and others, if there's any use for reinforcement learning, I've seen it used quite well, I study pedestrians a lot too, is try to train models from real data of how pedestrians move and try to use reinforcement learning models to make pedestrians move in human-like ways.

• 53:06George Hotz?

  By that point, you've already gone so many layers, you detected a pedestrian?

• 53:11Lex Fridman?

  Yeah.

• 53:12George Hotz?

  Did you, did you hand code the feature vector of their state?

• 53:16Lex Fridman?

  Right.

• 53:16George Hotz?

  Did you guys learn anything from computer vision before deep learning? Well, okay, you know, I feel like this is.

• 53:23Lex Fridman?

  So perception to you is, is the sticking point. This, I mean, what's, what's the hardest part of the stack here?

• 53:30George Hotz?

  There is no human understandable feature vector separating perception and planning. That's the best way I can, I can put that.

• 53:41Lex Fridman?

  There is no, so it's all together and it's, it's a, it's a joint problem.

• 53:46George Hotz?

  So you can take localization. Localization and planning, there is a human understandable feature vector between these two things. I mean, okay, so I have like three degrees position, three degrees orientation and those derivatives, maybe those second derivatives, right? That's human understandable, that's physical. The, between perception and planning, um, so like Waymo has a perception stack and then a planner. And one of the things Waymo does right is they have a simulator that can separate those two. They can like replay their perception data and test their system, which is what I'm talking about, about like the two different kinds of simulators. There's the kind that can work on real data and there's the kind that can't work on real data.

• 54:27George Hotz?

  Now, the problem is that I don't think you can hand code a feature vector, right? Like, like you have some list of like, oh, here's my list of cars in the scenes. Here's my list of pedestrians in the scene. This isn't what humans are doing.

• 54:39Lex Fridman?

  What are humans doing?

• 54:41George Hotz?

  Global.

• 54:43Lex Fridman?

  And you're saying that's too difficult to hand engineer?

• 54:48George Hotz?

  I'm saying that there is no state vector. Given a perfect, I could give you the best team of engineers in the world to build a perception system and the best team to build a planner. All you have to do is define the state vector that separates those two.

• 55:00Lex Fridman?

  I'm missing the state vector that separates those two. What do you mean?

• 55:05George Hotz?

  So what is the output of your perception system?

• 55:10Lex Fridman?

  Output of the perception system?

• 55:13Lex Fridman?

  It's a, there's, okay, well, there's several ways to do it. One is the slam component is localization. The other is drivable area, drivable space.

• 55:22George Hotz?

  Drivable space, yeah.

• 55:23Lex Fridman?

  And then there's the different objects in the scene.

• 55:27Lex Fridman?

  And different objects in the scene over time, maybe to give you input to then try to start modeling the trajectories of those objects.

• 55:38George Hotz?

  Sure.

• 55:38Lex Fridman?

  That's it.

• 55:39George Hotz?

  I can give you a concrete example of something you missed.

• 55:41Lex Fridman?

  What's that?

• 55:42George Hotz?

  So say there's a bush in the scene. Humans understand that when they see this bush, that there may or may not be a car behind that bush. Drivable area and a list of objects does not include that. Humans are doing this constantly at the simplest intersections. So now you have to talk about occluded area.

• 56:00Lex Fridman?

  Right.

• 56:01George Hotz?

  Right? But even that, what do you mean by occluded? Okay, so I can't see it. Well, if it's the other side of a house, I don't care. What's the likelihood that there's a car in that occluded area? Right? And if you say, okay, we'll add that, I can come up with 10 more examples that you can't add.

• 56:18Lex Fridman?

  Certainly occluded area would be something that simulator would have because it's simulating the entire, you know, occlusion is part of it.

• 56:27George Hotz?

  Occlusion is part of a vision stack. But what I'm saying is if you have a hand engineered, if your perception system output can be written in a spec document, it is incomplete.

• 56:39Lex Fridman?

  Yeah, I mean, certainly it's, it's hard to argue with that because in the end that's going to be true.

• 56:46George Hotz?

  Yes, and I'll tell you what the output of our perception system is.

• 56:49Lex Fridman?

  What's that?

• 56:49George Hotz?

  It's a 1,000, it's a 1,024 dimensional vector.

• 56:53Lex Fridman?

  Oh, you know that?

• 56:55George Hotz?

  No, no, it's the 1,024 dimensions of who knows what.

• 56:59Lex Fridman?

  Because it's operating on real data.

• 57:01George Hotz?

  Yeah. And that's the perception. That's the perception state, right? Think about, uh, think about an autoencoder for faces, right? If you have an autoencoder for faces and you, uh, say it has, uh, two fifty-six dimensions in the middle, and I'm taking a face over here and projecting it to a face over here.

• 57:18Lex Fridman?

  Yep.

• 57:19George Hotz?

  Can you hand-label all two fifty-six of those dimensions?

• 57:22Lex Fridman?

  Well, no, but, but those are generated automatically, but I...

• 57:25George Hotz?

  But even if you tried to do it by hand, could you come up with a spec for your, and between your encoder and your decoder?

• 57:33Lex Fridman?

  No, no, because that's not, it wasn't designed, but there...

• 57:37George Hotz?

  No, no, no, but if you could design it,

• 57:40George Hotz?

  if you could design a face reconstructor system, could you come up with a spec?

• 57:45Lex Fridman?

  No, but I think we're missing here a little bit, uh, I think the, the, you're just being very poetic about expressing a fundamental problem of simulators, uh, that they're going to be missing so much that, um, the feature vector would just look fundamentally different from, in the simulated world than, uh, the real world.

• 58:07George Hotz?

  I'm not making a claim about simulators. I'm making a claim about the spec division between perception and planning.

• 58:15Lex Fridman?

  And planning.

• 58:15George Hotz?

  Even in your system.

• 58:17Lex Fridman?

  Just in general.

• 58:18George Hotz?

  Right? Just in general. If you're trying to build a car that drives, if you're trying to hand-code the output of your perception system, like saying like, here's a list of all the cars in the scene. Here's a list of all the people. Here's a list of the included areas. Here's a vector of drivable areas. It's insufficient. And if you start to believe that, you realize that what Waymo and Cruise are doing is impossible.

• 58:37Lex Fridman?

  Currently, what we're doing is the perception problem is converting the scene into a chessboard. You, and then you reason some basic reasoning around that chessboard.

• 58:49George Hotz?

  Yeah.

• 58:49Lex Fridman?

  And you're saying that really, there's a lot missing there. First of all, why are we talking about this? Because isn't this a full autonomy? Is this something you think about?

• 59:01George Hotz?

  Oh, I want to win self-driving cars.

• 59:03Lex Fridman?

  So your definition of win includes?

• 59:08George Hotz?

  Level four, five.

• 59:09Lex Fridman?

  Level five.

• 59:10George Hotz?

  I don't think level four is a real thing. I want to build, I want to build the AlphaGo of driving.

• 59:17Lex Fridman?

  So, so AlphaGo is really end to end?

• 59:22George Hotz?

  Yeah.

• 59:23Lex Fridman?

  Is, uh, yeah, it's end to end. And do you think this whole problem, is that also kind of what you're getting at with the perception?

• 59:32George Hotz?

  Yes.

• 59:32Lex Fridman?

  And the planning? Is that this whole problem, the right way to do it is really to learn the entire thing?

• 59:38George Hotz?

  I'll argue that not only is it the right way, it's the only way that's going to exceed human performance.

• 59:44Lex Fridman?

  Well.

• 59:45George Hotz?

  Certainly true for Go. Everyone who tried to hand code Go things built human inferior things. And then someone came along and wrote some 10,000 line thing that doesn't know anything about Go that beat everybody. It's 10,000 lines.

• 59:57Lex Fridman?

  True. In that sense, the, the open question then that maybe I can ask you is, uh, driving is much harder than Go. The open question is how much harder? So how, because I think the Elon Musk approach here with planning and perception is similar to what you're describing, which is really turning into, uh, not some kind of modular thing, but really do formulate it as a learning problem and solve the learning problem with scale. So how many years, at a point, is how many years would it take to solve this problem? Or just how hard is this freaking problem?

• 1:00:38George Hotz?

  Well, the cool thing is,

• 1:00:42George Hotz?

  I think there's a lot of value that we can deliver along the way. I think that you can build.

• 1:00:51George Hotz?

  Lane keeping assist actually plus adaptive cruise control plus, okay, looking at ways extends to like all of driving.

• 1:01:02Lex Fridman?

  Yeah, most of driving, right?

• 1:01:04George Hotz?

  Oh, your adaptive cruise control treats red lights like cars. Okay.

• 1:01:07Lex Fridman?

  So let's jump around. You mentioned that you didn't like navigate an autopilot.

• 1:01:11George Hotz?

  Yeah.

• 1:01:12Lex Fridman?

  What advice, how would you make it better? Do you think as a feature that if it's done really well, it's a good feature?

• 1:01:18George Hotz?

  I think that it's too reliant on like hand-coded hacks for like, how does navigate an autopilot do a lane change? It actually does the same lane change every time and it feels mechanical. Humans do different lane changes. Humans sometime will do a slow one, sometimes do a fast one. Navigate an autopilot, at least every time I use it, it did the identical lane change.

• 1:01:39Lex Fridman?

  How do you learn? I mean, this is a fundamental thing, actually.

• 1:01:42George Hotz?

  Yeah.

• 1:01:43Lex Fridman?

  Is the braking and then accelerating, something that's still, Tesla probably does it better than most cars, but it still doesn't do a great job of creating a comfortable, natural experience. And navigate an autopilot is just lane change is an extension of that. So how do you learn to do a natural lane change?

• 1:02:05George Hotz?

  So we have it and I can talk about how it works. So I feel that we have the solution for lateral. We don't yet have the solution for longitudinal. There's a few reasons longitudinal is harder than lateral. The lane change component, the way that we train on it very simply is like, our model has an input for whether it's doing a lane change or not. And then when we train the end-to-end model, we hand-label all the lane changes because you have to. I've struggled a long time about not wanting to do that, but I think you have to. Because you or

• 1:02:41Lex Fridman?

  For the training data.

• 1:02:41George Hotz?

  For the training data, right? We actually, we have an automatic ground truther, which automatically labels all the lane changes.

• 1:02:47Lex Fridman?

  Was that possible?

• 1:02:48George Hotz?

  To automatically label lane changes?

• 1:02:49Lex Fridman?

  Yeah.

• 1:02:49George Hotz?

  Yeah, detect the lane and I see when it crosses it, right? And I don't have to get that that high percent accuracy, but it's like 95, good enough.

• 1:02:54Lex Fridman?

  Okay.

• 1:02:55George Hotz?

  Now I set the bit when it's doing the lane change in the end-to-end learning. And then I set it to zero when it's not doing a lane change. So now if I want it to do a lane change at test time, I just put the bit to a one and it'll do a lane change.

• 1:03:08Lex Fridman?

  Yeah, but so if you look at the space of lane change, you know, some percentage, not 100% that we make as humans is not a pleasant experience because we mess some part of it up.

• 1:03:19George Hotz?

  Yeah.

• 1:03:19Lex Fridman?

  It's nerve wracking to change the look, you have to see, the desire to accelerate. How do we label the ones that are natural and feel good? You know, that's the, because that's your ultimate criticism. The current navigate autopilot just doesn't feel good.

• 1:03:33George Hotz?

  Well, the current navigate autopilot is a hand coded policy written by an engineer in a room who probably went out and tested it a few times on the 280.

• 1:03:41Lex Fridman?

  Probably a more, a better version of that, but yes.

• 1:03:45George Hotz?

  That's how we would have written it to come AI.

• 1:03:47Lex Fridman?

  Yeah.

• 1:03:47George Hotz?

  Maybe Tesla, they tested it in.

• 1:03:49Lex Fridman?

  There might have been two engineers.

• 1:03:51George Hotz?

  Two engineers, yeah. No, but so if you learn the lane change, if you learn how to do a lane change from data, just like, just like you have a label that says lane change and then you put it in when you want it to do the lane change, it'll automatically do the lane change that's appropriate for the situation. Now, to get at the problem of some humans do bad lane changes. We haven't worked too much on this problem yet. It's not that much of a problem in practice. My theory is that all good drivers are good in the same way and all bad drivers are bad in different ways. And we've, we've seen some data to back this up.

• 1:04:27Lex Fridman?

  Well, beautifully put. So you just basically, if that's true.

• 1:04:31George Hotz?

  Yeah.

• 1:04:32Lex Fridman?

  Hypothesis.

• 1:04:32George Hotz?

  Yeah.

• 1:04:33Lex Fridman?

  Then your task is to discover the good drivers.

• 1:04:36George Hotz?

  The good drivers stand out because they're in one cluster and the bad drivers are scattered all over the place and your net learns the cluster.

• 1:04:43Lex Fridman?

  Yeah, that's, uh, so you just learn from the good drivers and they're easy to cluster.

• 1:04:49George Hotz?

  In fact, we learn from all of them and the net automatically learns the policy that's like the majority, but we'll eventually probably have to build it up.

• 1:04:55Lex Fridman?

  If that theory is true, I hope it's true because the, the counter theory is, um, there is many clusters.

• 1:05:06Lex Fridman?

  Maybe arbitrarily many clusters of good drivers. Because if there's one cluster of good drivers, you can at least discover a set of policies. You can learn a set of policies, which would be good universally.

• 1:05:17George Hotz?

  Yeah.

• 1:05:18Lex Fridman?

  That would be a nice, that would be nice if it's true. And you're saying that there is some evidence that.

• 1:05:23George Hotz?

  Let's say lane changes can be clustered into four clusters.

• 1:05:26Lex Fridman?

  Right.

• 1:05:26George Hotz?

  Right?

• 1:05:27Lex Fridman?

  There's this finite level.

• 1:05:28George Hotz?

  I would argue that all four of those are good clusters.

• 1:05:31Lex Fridman?

  Right.

• 1:05:31George Hotz?

  All the things that are random are noise and probably bad. And which one of the four you pick, or maybe it's 10, or maybe it's 20.

• 1:05:38Lex Fridman?

  You can learn that.

• 1:05:39George Hotz?

  It's context dependent. It depends on the scene.

• 1:05:43Lex Fridman?

  And the hope is it's not too dependent on the driver?

• 1:05:47George Hotz?

  Yeah, the hope is that it all washes out. The hope is that there's, that the distribution's not bimodal. The hope is that it's a nice Gaussian.

• 1:05:54Lex Fridman?

  Right. So what advice would you give to Tesla? How to fix, how to improve, navigate an autopilot? That's the lessons that you've learned from Kama AI.

• 1:06:04George Hotz?

  The only real advice I would give to Tesla is please put driver monitoring in your cars. With respect to improving.

• 1:06:11Lex Fridman?

  You can't do that anymore. I said I didn't interrupt, but you know, there's a practical nature of many of hundreds of thousands of cars being produced that don't have a good driver facing camera.

• 1:06:22George Hotz?

  The Model 3 has a selfie cam. Is it not good enough? Did they not put IRL LEDs for night?

• 1:06:27Lex Fridman?

  That's a good question, but I do know that it's fisheye and it's relatively low resolution. So it's really not designed, it wasn't.

• 1:06:34George Hotz?

  It wasn't designed for driver monitoring.

• 1:06:35Lex Fridman?

  You can hope that you can kind of scrape up and have something from it.

• 1:06:40George Hotz?

  Yeah.

• 1:06:40Lex Fridman?

  But.

• 1:06:42George Hotz?

  Why didn't they put it in today? Put it in today.

• 1:06:44Lex Fridman?

  Put it in today.

• 1:06:46George Hotz?

  Every time I've heard Karpathy talk about the problem and talking about like software 2.0 and how the machine learning is gobbling up everything, I think this is absolutely the right strategy. I think that he didn't write and navigate on autopilot. I think somebody else did and kind of hacked it on top of that stuff. I think when Karpathy says, wait a second, why did we hand code this lane change policy with all these magic numbers? We're going to learn it from data. They'll fix it. They already know what to do there.

• 1:07:07Lex Fridman?

  Well, that's, that's Andre's job is to turn everything into a learning problem and collect a huge amount of data. The, the reality is though, not every problem can be turned into a learning problem in the short term. In the end, everything will be a learning problem. The reality is. Like if you want to build L5 vehicles today, it will likely involve no learning. And that's the reality is so at which point does learning start? It's the crutch statement that LIDAR is a crutch. At which point will learning get up to part of human performance? It's over human performance on ImageNet, classification on driving is the question still.

• 1:07:50George Hotz?

  It is a question. I'll say this, I'm here to play for 10 years. I'm not here to try to, I'm here to play for 10 years and make money along the way. I'm not here to try to promise people that I'm going to have my L5 taxi network up and working in two years.

• 1:08:04Lex Fridman?

  Do you think that was a mistake?

• 1:08:06George Hotz?

  Yes.

• 1:08:07Lex Fridman?

  What do you think was the motivation behind saying that? Other companies are also promising L5 vehicles with very different approaches in 2020, 2021, 2022.

• 1:08:18George Hotz?

  If anybody would like to bet me that those things do not pan out, I will, I will bet you even money, even money. I'll bet you as much as you want.

• 1:08:27Lex Fridman?

  So are you worried about what's going to happen? Cause you're not in full agreement on that. What's going to happen when 2022, 21 come around and nobody has fleets of autonomous vehicles?

• 1:08:39George Hotz?

  Well, you can look at the history. If you go back five years ago, they were all promised by 2018 and 2017.

• 1:08:46Lex Fridman?

  But they weren't that strong of promises. I mean, Ford really declared pretty, I think not many have declared as, as like definitively as they have now with these dates.

• 1:08:59George Hotz?

  Well, okay, so let's separate L4 and L5. Do I think that it's possible for Waymo to continue to kind of like, like hack on their system until it gets to level four in Chandler, Arizona? Yes.

• 1:09:11Lex Fridman?

  No safety driver?

• 1:09:13George Hotz?

  Chandler, Arizona, yeah.

• 1:09:15Lex Fridman?

  By, sorry, which year are we talking about?

• 1:09:19George Hotz?

  Oh, I even think that's possible by like 2020, 2021. But level four Chandler, Arizona, not level five, New York City.

• 1:09:27Lex Fridman?

  Level four, meaning some very defined streets it works on really well.

• 1:09:34George Hotz?

  Very defined streets and then practically these streets are pretty empty. If most of the streets are covered in Waymos, Waymo can kind of change the definition of what driving is. Right? If your self-driving network is the majority of cars in an area, they only need to be safe with respect to each other and all the humans will need to learn to adapt to them. Now go drive in downtown New York.

• 1:09:57Lex Fridman?

  Oh, yeah, that's.

• 1:09:58George Hotz?

  I mean, already you can talk about autonomy and like, like on farms, it already works great. Because you can really just follow the GPS line.

• 1:10:07Lex Fridman?

  So what does success look like for Calm AI? What, what are the milestones like where you can sit back with some champagne and say, we did it, boys and girls?

• 1:10:19George Hotz?

  Um, well, it's never over.

• 1:10:22Lex Fridman?

  Yeah, but you must drink champagne. So what is a good, what are some wins?

• 1:10:29George Hotz?

  Um, a big milestone that we're hoping for

• 1:10:34George Hotz?

  by mid next year is profitability of the company.

• 1:10:40George Hotz?

  And we're gonna have to revisit the idea of selling a consumer product. But it's not going to be like the comma one. When we do it, it's going to be perfect. OpenPilot has gotten so much better in the last two years. We're gonna have a few, a few features, we're gonna have a hundred percent driver monitoring. We're gonna disable no safety features in the car. Actually, I think it'd be really cool what we're doing right now. Our project this week is we're analyzing the data set and looking for all the AEB triggers from the manufacturer's systems.

• 1:11:12George Hotz?

  We have better data set on that than the manufacturers. How much does, how many, does Toyota have 10 million miles of real world driving to know how many times their AEB triggered?

• 1:11:21Lex Fridman?

  So let me give you, because you asked, right? Financial advice.

• 1:11:26George Hotz?

  Yeah.

• 1:11:27Lex Fridman?

  Because I work with a lot of automakers and one possible source of money for you, which I'll be excited to see you take on is basically selling the data.

• 1:11:41Lex Fridman?

  So, uh, which is something that most people and not selling in a way where here, here automaker, but creating, we've done this actually at MIT, not for money purposes, but you could do it for significant money purposes and make the world a better place by creating a consortia where automakers would pay in and then they get to have free access to the data. And I, I think a lot of people are really hungry for that and would pay significant amount of money for it.

• 1:12:10George Hotz?

  Here's the problem with that. I like this idea all in theory. It'd be very easy for me to give them access to my servers and we already have all open source tools to access this data. It's in a great format. We have a great pipeline, but they're gonna put me in the room with some business development guy. And I'm gonna have to talk to this guy and he's not gonna know most of the words I'm saying. I'm not willing to tolerate that.

• 1:12:33Lex Fridman?

  Okay, Mick Jagger.

• 1:12:35George Hotz?

  No, no, no, no.

• 1:12:36Lex Fridman?

  I think I agree with you. I'm the same way, but you just tell them the terms and there's no discussion needed.

• 1:12:41George Hotz?

  If I could, if I could just tell them the terms.

• 1:12:44Lex Fridman?

  Yeah.

• 1:12:45George Hotz?

  Then like, all right, uh, who wants access to my data? I will sell it to you for, uh, let's say, you want, you want a subscription? I'll sell it to you for a hundred K a month?

• 1:12:57George Hotz?

  Anyone?

• 1:12:57Lex Fridman?

  Hundred K a month?

• 1:12:58George Hotz?

  Hundred K a month? I'll give you access to this data subscription?

• 1:13:01Lex Fridman?

  Yeah.

• 1:13:02George Hotz?

  Um, yeah, I think that's kind of fair. Came up with that number off the top of my head. If somebody sends me like a three line email where it's like, we would like to pay a hundred K a month to get access to your data. We would agree to like reasonable privacy terms of, of, of, of the people who are in the data set. I would be happy to do it, but that's not going to be the email. The email is going to be, hey, do you have some time in the next month where we can sit down and we can, I don't have time for that. We're moving too fast.

• 1:13:24Lex Fridman?

  Yeah. You could politely respond to that email by not saying, I don't have any time for your bullshit. Uh, you say, oh, well, unfortunately, these are the terms. And so this is, we try to, we, we brought the cost down for you in order to minimize the friction, the communication.

• 1:13:38George Hotz?

  Yeah, absolutely.

• 1:13:39Lex Fridman?

  Here's the, whatever it is, one, two million years, dollars a year and you have access.

• 1:13:45George Hotz?

  And it's not like I get that email from like, but okay, am I going to reach out? Am I going to hire a business development person who's going to reach out to the automakers? No way.

• 1:13:53Lex Fridman?

  Yeah, okay, I got you.

• 1:13:54George Hotz?

  If they reached into me, I'm not going to ignore the email. I'll come back with something like, yeah, if you're willing to pay a hundred K a month for access to the data, I'm happy to to set that up. That's worth my engineering time.

• 1:14:04Lex Fridman?

  That's actually quite insightful of you. You're right. You're probably because many of the automakers are quite a bit old school.

• 1:14:10George Hotz?

  Yeah.

• 1:14:10Lex Fridman?

  There will be need to reach out and they want it, but they don't need to be some some communication. You're right.

• 1:14:16George Hotz?

  Mobileye circa 2015 had the lowest R and D spend of any chipmaker. Like per per and you look at all the people who work for them and it's all business development people because the car companies are impossible to work with.

• 1:14:31Lex Fridman?

  Yeah, so you're you have no patience for that and you're you're legit android, huh?

• 1:14:36George Hotz?

  I have something to do, right? Like, like, it's not like, it's not like, I don't like, I don't mean to like be a dick and say like, I don't have patience for that, but it's like that stuff doesn't help us with our goal of winning self-driving cars. If I want money in the short term,

• 1:14:50George Hotz?

  if I showed off like the actual, like the learning tech that we have, it's it's somewhat sad, like it's years and years ahead of everybody else's. Not maybe not Tesla's. I think Tesla has similar stuff to us, actually.

• 1:15:01Lex Fridman?

  Yeah.

• 1:15:02George Hotz?

  I think Tesla has similar stuff, but when you compare it to like what the Toyota Research Institute has, you're not even close to what we have.

• 1:15:10Lex Fridman?

  No comment. Well, but I also can't, I have to take your comments. I intuitively believe you, but I have to take it with a grain of salt because, I mean, you are an inspiration because you basically don't care about a lot of things that other companies care about. You don't try to bullshit in a sense like make up stuff so to drive up valuation. You're really very real and you're trying to solve the problem and admire that a lot. What I don't necessarily fully can't trust you on, with all due respect, is how good it is, right? I can only, but I also know how bad others are.

• 1:15:48George Hotz?

  And so I'll say, I'll say two things about don't trust but verify, right? I'll say two things about that. One is try, uh, get in a 2020 Corolla and try open pilot 0.6 when it comes out next month. I think already you'll look at this and you'll be like,

• 1:16:05Lex Fridman?

  Damn.

• 1:16:06George Hotz?

  This is already really good. And then I could be doing that all with hand labelers and all with, with like, like the same approach that like Mobileye uses. When we release a model that no longer has the lanes in it, that only outputs a path.

• 1:16:21George Hotz?

  Then think about how we did that machine learning. And then right away, when you see, and that's going to be an open pilot, that's going to be an open pilot before 1.0. When you see that model, you'll know that everything I'm saying is true because how else did I get that model?

• 1:16:33Lex Fridman?

  Good.

• 1:16:33Speaker 02

  You know what I'm saying is true about the simulator.

• 1:16:35Lex Fridman?

  Yeah, yeah, yeah. This is super exciting. That's super exciting. And.

• 1:16:39George Hotz?

  But like, you know, I listened to your talk with Kyle and Kyle was originally building the aftermarket system. And he gave up on it because of technical challenges.

• 1:16:50Lex Fridman?

  Yeah.

• 1:16:51George Hotz?

  Because of the fact that he's going to have to support 20 to 50 cars. We support 45. Because what is he going to do when the manufacturer ABS system triggers? We have alerts and warnings to deal with all of that and all the cars. And how is he going to formally verify it? Well, I got 10 million miles of data. It's probably better. It's probably better verified than the spec.

• 1:17:09Lex Fridman?

  Yeah, I'm glad you're here talking to me. This is, I'll remember this day. Because it's interesting. If you look at Kyle's from Cruise, I'm sure they have a large number of business development folks. And you work with, he's working with GM, he could work with Argo AI, he worked with Ford. It's interesting because chances that you fail, business-wise, like bankrupt, are pretty high.

• 1:17:36George Hotz?

  Yeah.

• 1:17:37Lex Fridman?

  And, and yet it's the Android model is you're actually taking on the problem. So that's really inspiring. I mean.

• 1:17:44George Hotz?

  Well, I have a, I have a long-term way for comment to make money too.

• 1:17:47Lex Fridman?

  And one of the nice things when you really take on the problem, which is my hope for Autopilot, for example, is things you don't expect, ways to make money or create value that you don't expect will pop up.

• 1:18:00George Hotz?

  Oh, I've known how to do it since kind of 2017 is the first time I sat it.

• 1:18:05Lex Fridman?

  Which part? Do you know how to do which part?

• 1:18:06George Hotz?

  Our long-term plan is to be a car insurance company.

• 1:18:09Lex Fridman?

  Insurance. Yeah, I love it. Yep, yep.

• 1:18:11George Hotz?

  I make driving twice as safe. Not only that, I have the best data set to know who statistically is the safest drivers. And, oh, oh, we see you, we see you driving unsafely. We're not going to insure you. And that that causes a like bifurcation in the market because the only people who can't get common insurance are the bad drivers. Geico can insure them. Their premiums are crazy high. Our premiums are crazy low. We'll win car insurance, take over that whole market.

• 1:18:34Lex Fridman?

  Okay, so.

• 1:18:36George Hotz?

  If we win, if we win. But that's I'm saying like, how do you turn comma into a $10 billion company? It's that.

• 1:18:41Lex Fridman?

  That's right. So you, Elon Musk, who else? Who else is thinking like this and working like this in your view? Who are the competitors? Are there people seriously? I don't think anyone that I'm aware of is seriously taking on lane keeping. You know, like Toukwa is a huge business that turns eventually into full autonomy that then creates.

• 1:19:06Lex Fridman?

  Yeah, like that creates other businesses on top of it and so on, thinks insurance, thinks all kinds of ideas like that. Do you know who anyone else thinking like this?

• 1:19:17George Hotz?

  Not really.

• 1:19:18Lex Fridman?

  That's interesting. I mean, my sense is everybody turns to that in like four or five years. Like Ford, once the autonomy doesn't fall through. But at this time.

• 1:19:29George Hotz?

  Elon's the iOS. By the way, he paved the way for for all of us.

• 1:19:33Lex Fridman?

  It's the iOS, true.

• 1:19:34George Hotz?

  I would not be doing comma AI today if it was not for those conversations with Elon. And if it were not for him saying like, I think he said like, well, obviously, we're not going to use LiDAR. We use cameras. Humans use cameras.

• 1:19:47Lex Fridman?

  So what do you think about that? How important is LiDAR? Everybody else is on L5 is using LiDAR. What are your thoughts on his provocative statement that LiDAR is a crutch?

• 1:19:57George Hotz?

  See, sometimes he'll say dumb things like the driver monitoring thing, but sometimes he'll say absolutely, completely, 100% obviously true things.

• 1:20:04Lex Fridman?

  Yeah.

• 1:20:04George Hotz?

  Of course, LiDAR is a crutch. It's not even a good crutch. You're not even using it. Oh, they're using it for localization.

• 1:20:12Lex Fridman?

  Yeah.

• 1:20:13George Hotz?

  Which isn't good in the first place. If you have to localize your car to centimeters in order to drive. Like

• 1:20:19Lex Fridman?

  Yeah, they're not

• 1:20:20George Hotz?

  That's not driving.

• 1:20:20Lex Fridman?

  Currently not doing much machine learning of LIDAR data. Meaning like to help you in the task of general task of perception.

• 1:20:29George Hotz?

  The main goal of those LIDARs on those cars, I think, is actually localization more than perception. Or at least that's what they use them for.

• 1:20:36Lex Fridman?

  Yeah, that's true.

• 1:20:37George Hotz?

  If you want to localize to centimeters, you can't use GPS. The fanciest GPS in the world can't do it, especially if you're under tree cover and stuff.

• 1:20:43Lex Fridman?

  Yeah.

• 1:20:43George Hotz?

  LIDAR, you can do this pretty easily.

• 1:20:44Lex Fridman?

  So you really, they're not taking on, I mean, in some research, they're using it for perception, but and they're certainly not, which is sad, they're not fusing it well with vision.

• 1:20:55George Hotz?

  They do use it for perception. I'm not saying they don't use it for perception, but the thing that they have vision-based and radar-based perception systems as well. You could remove the LIDAR and, and, and keep around a lot of the dynamic object perception. You want to get centimeter accurate localization? Good luck doing that with anything else.

• 1:21:15Lex Fridman?

  So what should Cruise, Waymo do? Like, what would be your advice to them now?

• 1:21:22George Hotz?

  Oh, I don't know.

• 1:21:22Lex Fridman?

  I mean, Waymo is actually, they're, I mean, they're doing, they're serious. Waymo out of the ball of them are quite so serious about the long game. If L5 is a lot, is, requires 50 years. I think Waymo will be the only one left standing at the end with the, given the financial backing that they have.

• 1:21:42George Hotz?

  Their Google box. Um, I'll say nice things about both Waymo and Cruise.

• 1:21:48Lex Fridman?

  Yeah, yeah, yeah. Let's do it. Nice is good.

• 1:21:52George Hotz?

  Waymo is by far the furthest along with the technology. Waymo has a three to five year lead on all the competitors. Um, if that, if the Waymo looking stack works, maybe three year lead. If the Waymo looking stack works, they have a three year lead. Now, I argue that Waymo has spent too much money to recapitalize, to, to, to, to gain back their losses in those three years. Also, self-driving cars have no network effect like that.

• 1:22:19Lex Fridman?

  Yeah.

• 1:22:20George Hotz?

  Uber has a network effect. You have a market. You have drivers and you have riders. Self-driving cars, you have capital and you have riders. There's no network effect. If I want to blanket a new city in self-driving cars, I buy the off-the-shelf Chinese knockoff self-driving cars and I buy enough of them in the city. I can't do that with drivers. And that's why Uber has a first mover advantage that no self-driving car company will.

• 1:22:40Lex Fridman?

  Can you just tangle that a little bit? Uber, you're not talking about Uber, the autonomous vehicle Uber. You're talking about the Uber car.

• 1:22:47George Hotz?

  Okay.

• 1:22:47Lex Fridman?

  Yeah.

• 1:22:48George Hotz?

  I'm Uber. I open for business in Austin, Texas, let's say. I need to attract both sides of the market. I need to both get drivers on my platform and riders on my platform. And I need to keep them both sufficiently happy, right? Riders aren't going to use it if it takes more than five minutes for an Uber to show up. Drivers aren't going to use it if they have to sit around all day and there's no riders. So you have to carefully balance a market. And whenever you have to carefully balance a market, there's a great first mover advantage. Because there's a switching cost for everybody, right? The drivers and the riders would have to switch at the same time. Let's even say that, you know, um, let's say, uh, Luber shows up. And Luber somehow, you know, agrees to do things at a, at a bigger, uh, you know, you know,

• 1:23:31George Hotz?

  we're just gonna, we've done it more efficiently, right? Luber is, only takes 5% of a cut instead of the 10% that Uber takes. No one is going to switch because the switching cost is higher than that 5%. So you actually can, in markets like that, you have a first mover advantage.

• 1:23:45Lex Fridman?

  Yeah.

• 1:23:46George Hotz?

  Autonomous vehicles of the level five variety have no first mover advantage. If the technology becomes commoditized, say I want to go to a new city. Look at the scooters. It's going to look a lot more like scooters. Every person with a checkbook can blanket a city in scooters. And that's why you have 10 different scooter companies. Which one's going to win? It's a race to the bottom. It's a terrible market to be in because there's no market for scooters.

• 1:24:11George Hotz?

  And The scooters don't get a say in whether they want to be bought and deployed to a city or not.

• 1:24:15Lex Fridman?

  Right, so the, yeah.

• 1:24:16George Hotz?

  We're going to entice the scooters with subsidies and deals and

• 1:24:20Lex Fridman?

  So whenever you have to invest that capital, it doesn't

• 1:24:23George Hotz?

  It doesn't come back.

• 1:24:24Lex Fridman?

  Yeah. That, that can't be your main criticism of the Waymo approach.

• 1:24:28George Hotz?

  Oh, I'm saying even if it does technically work.

• 1:24:31Lex Fridman?

  Even if it does technically work, that's a problem.

• 1:24:33George Hotz?

  Yeah.

• 1:24:34Lex Fridman?

  Uh, I don't, I don't know, I, if I were to say, I, I would, I would say You're already there. I haven't even thought about that, but I would say the bigger challenge is the technical approach.

• 1:24:46George Hotz?

  So Waymo's Cruise is

• 1:24:48Lex Fridman?

  And not just the technical approach, but of creating value. I still don't understand how you beat Uber, the the human driven cars. In terms of financially, it doesn't, it doesn't make sense to me that people want to get in an autonomous vehicle. I don't understand how you make money. In the long term, yes, like real long term, but it just feels like there's too much capital investment needed.

• 1:25:16George Hotz?

  Oh, and they're going to be worse than Uber's because they're going to they're going to stop for every little, you know, thing everywhere. Um, I'll say a nice thing about Cruise. That was my nice thing about Waymo. They're three years ahead of

• 1:25:25Lex Fridman?

  What was the nice? Oh, because they're

• 1:25:26George Hotz?

  That they're three years technically ahead of everybody. Their tech stack is is great. My nice thing about Cruise is GM buying them was a great move for GM.

• 1:25:37George Hotz?

  For $1 billion, GM bought an insurance policy against Waymo. They put Cruise is three years behind Waymo. That means Google will get a monopoly on the technology for at most three years.

• 1:25:53Lex Fridman?

  And technology works. You might not even be right about the three years. It might be less.

• 1:25:58George Hotz?

  Might be less. Cruise actually might not be that far behind. I don't know how much Waymo has waffled around or how much of it actually is just that long tail.

• 1:26:06Lex Fridman?

  Yeah, okay. If that's the best you could say in terms of nice things, that's more of a nice thing for GM, that that's the smart insurance policy.

• 1:26:15George Hotz?

  It's a smart insurance policy. I mean, I think that's how I can't see Cruise working out any other. For Cruise to leapfrog Waymo would really surprise me.

• 1:26:26Lex Fridman?

  Yeah. So let's talk about the the underlying assumption of everything.

• 1:26:30George Hotz?

  We're not going to leapfrog Tesla.

• 1:26:34George Hotz?

  Tesla would have to seriously mess up for us.

• 1:26:35Lex Fridman?

  Because you're, okay, so the way you leapfrog, right? Is you come up with an idea or you take a direction, perhaps secretly, that the other people aren't taking. And so Cruise, Waymo, even Aurora.

• 1:26:54George Hotz?

  I don't know Aurora, Zooks is the same stack as well. They're all the same codebase even. They're all the same DARPA Urban Challenge codebase.

• 1:27:01Lex Fridman?

  So the question is, do you think there's a room for brilliance and innovation there that will change everything?

• 1:27:08Lex Fridman?

  Like, say, okay, so I'll give you examples. It could be if revolution in mapping, for example, that allow you to map things, do HD maps of the whole world, all weather conditions somehow really well, or

• 1:27:28Lex Fridman?

  revolution in simulation

• 1:27:31Lex Fridman?

  to where the, all the way you said before becomes incorrect. That kind of thing. Any room for breakthrough innovation?

• 1:27:41George Hotz?

  What I said before about, oh, they actually get the whole thing. Well, I'll say this about, we divide driving into three problems. And I actually haven't solved the third yet, but I have an idea how to do it. So there's the static. The static driving problem is assuming you are the only car on the road.

• 1:27:55Lex Fridman?

  Right.

• 1:27:56George Hotz?

  Right? And this problem can be solved 100% with mapping and localization. This is why farms work the way they do. If all you have to deal with is the static problem and you can statically schedule your machines, right? It's the same as like statically scheduling processes. You can statically schedule your tractors to never hit each other on their paths. Right? Because they, you know, the speed they go at. So, so that's the static driving problem. Maps only helps you with the static driving problem.

• 1:28:20Lex Fridman?

  Yeah. The question about static driving.

• 1:28:23George Hotz?

  Yeah.

• 1:28:23Lex Fridman?

  You've just made it sound like it's really easy.

• 1:28:25George Hotz?

  Static driving is really easy.

• 1:28:28Lex Fridman?

  How easy. How, well, because the whole drifting out of lane. When Tesla drifts out of lane, it's failing on the fundamental static driving problem.

• 1:28:38George Hotz?

  Tesla is drifting out of lane? The static driving problem is not easy for the world. The static driving problem is easy for one route.

• 1:28:48Lex Fridman?

  One route in one weather condition with one state of lane markings and like no deterioration, no cracks in the road.

• 1:28:57George Hotz?

  No, I'm assuming you have a perfect localizer. So that's solved for the weather condition and the lane marking condition.

• 1:29:02Lex Fridman?

  But that's the problem is how could you, how do you have a perfect?

• 1:29:04George Hotz?

  You can build, perfect localizers are not that hard to build.

• 1:29:07Lex Fridman?

  Okay, come on now with, with, with LiDAR.

• 1:29:09George Hotz?

  With LiDAR, yeah.

• 1:29:10Lex Fridman?

  With LiDAR, okay.

• 1:29:11George Hotz?

  With LiDAR, yeah, but you use LiDAR, right? Like use LiDAR, build a perfect localizer. Building a perfect localizer without LiDAR,

• 1:29:19George Hotz?

  it's gonna be, it's gonna be hard. You can get 10 centimeters without LiDAR, you can get 1 centimeter with LiDAR.

• 1:29:23Lex Fridman?

  I'm not even concerned about the 1 or 10 centimeters. I'm concerned if every once in a while, you're just way off.

• 1:29:29George Hotz?

  Yeah, so this is why you have to carefully make sure you're always tracking your position. You want to use LiDAR camera fusion, but you can get the reliability of that system up to

• 1:29:43George Hotz?

  100,000 miles and then you write some fallback condition where it's not that bad if you're way off, right? I think that you can get it to the point, it's like Azel D, that you're, you're never in a case where you're way off and you don't know it.

• 1:29:54Lex Fridman?

  Yeah, okay, so this is brilliant. So that's the static.

• 1:29:57George Hotz?

  Static.

• 1:29:58Lex Fridman?

  We can, especially with LiDAR and good HD maps, you can solve that problem.

• 1:30:03George Hotz?

  Easy.

• 1:30:05Lex Fridman?

  Easy.

• 1:30:05George Hotz?

  No, it's, oh, the static, the static problem's so easy.

• 1:30:08Lex Fridman?

  It's very typical for you to say something is easy. I got it. It's not as challenging as the other ones. Okay.

• 1:30:13George Hotz?

  Well, it's, it's, okay, maybe it's obvious how to solve it. The third one's the hardest. Where do we get, and a lot of people don't even think about the third one and even see it as different from the second one. So the second one is dynamic. The second one is like, say there's a, an obvious example is like a car stopped at a red light, right? You can't have that car in your map.

• 1:30:28Lex Fridman?

  Yeah.

• 1:30:29George Hotz?

  Because you don't know whether that car is going to be there or not. So you have to detect that car in real time and then you have to, you know, do the appropriate action, right?

• 1:30:38George Hotz?

  Also, that car is not a fixed object. That car may move and you have to predict what that car will do. Right? So this is the dynamic problem.

• 1:30:47Lex Fridman?

  Yeah.

• 1:30:47George Hotz?

  So you have to deal with this. This involves, again, like, you're gonna need models of other people's behavior.

• 1:30:55Lex Fridman?

  Do you, are you including in that, I don't wanna step on, on the third one.

• 1:30:58George Hotz?

  Oh.

• 1:30:59Lex Fridman?

  But are you including in that your influence on people's?

• 1:31:03George Hotz?

  Ah, that's the third one.

• 1:31:04Lex Fridman?

  Okay.

• 1:31:05George Hotz?

  That's the third one. We call it the counterfactual.

• 1:31:08Lex Fridman?

  Yeah.

• 1:31:08George Hotz?

  And that.

• 1:31:09Lex Fridman?

  I just talked to Judea Pro, who's obsessed with counterfactuals.

• 1:31:12George Hotz?

  Oh, yeah, yeah, I've read his books.

• 1:31:15George Hotz?

  So the static and the dynamic.

• 1:31:17Lex Fridman?

  Yeah.

• 1:31:18George Hotz?

  Our, our approach right now for lateral will scale completely to the static and dynamic. The counterfactual, the only way I have to do it yet, the, the, the thing that I wanna do once we have all of these cars is I wanna do reinforcement learning on the world. I'm always gonna turn the exploiter up to max. I'm not gonna have them explore. But the only real way to get at the counterfactual is to do reinforcement learning because the other agents are humans.

• 1:31:44Lex Fridman?

  So that's fascinating that you break it down like that. I agree completely.

• 1:31:48George Hotz?

  I've spent my life thinking about this problem.

• 1:31:50Lex Fridman?

  Beautiful. There's so, and part of it, because you're slightly insane. Because.

• 1:31:57George Hotz?

  Not my life, just the last four years.

• 1:31:59Lex Fridman?

  No, no, you, you have like, some, some non-zero percent of your brain has a madman in it, which is, that's a really good feature. But the, there's a safety component to it that I think, sort of with counterfactuals and so on, that would just freak people out. How do you even start to think about, just in general? I mean, you've, you've had some friction with NHTSA and so on. I am frankly exhausted by safety engineers. The, the prioritization on.

• 1:32:35Lex Fridman?

  Safety over innovation to a degree where it kills, in my view, kills safety in the long term. So the counterfactual thing, they just, just actually exploring this world of how do you interact with dynamic objects and so on, how do you, how do you think about safety?

• 1:32:51George Hotz?

  You can do reinforcement learning without ever exploring. And I said that like, so you can think about your, in like reinforcement learning, it's usually called like a temperature parameter. And your temperature parameter is how often you deviate from the argmax. I could always set that to zero and still learn. And I feel that you'd always want that set to zero on your actual system.

• 1:33:10Lex Fridman?

  Gotcha. But the problem is you first don't know very much and so you're going to make mistakes. So the learning, the exploration happens through mistakes.

• 1:33:18George Hotz?

  Already, yeah, but, okay. So the consequences of a mistake.

• 1:33:22Lex Fridman?

  Yeah.

• 1:33:22George Hotz?

  Open pilot and autopilot are making mistakes left and right. We have, we have, we have 700 daily active users, 1,000 weekly active users. Open pilot makes tens of thousands of mistakes a week. These mistakes have zero consequences. These mistakes are, oh, it, it, I wanted to take this exit and it went straight. So I'm just going to carefully touch the wheel.

• 1:33:45Lex Fridman?

  The humans catch them.

• 1:33:45George Hotz?

  The humans catch them. And the human disengagement is labeling that reinforcement learning in a completely consequence free way.

• 1:33:53Lex Fridman?

  So driver monitoring is the way you ensure they keep.

• 1:33:56George Hotz?

  Yes.

• 1:33:56Lex Fridman?

  They keep paying attention. How's your messaging? Say I gave you a billion dollars, you would be scaling and now.

• 1:34:04George Hotz?

  Oh, if I could, I couldn't scale with any amount of money. I'd raise money if I could, if I had a way to scale it.

• 1:34:08Lex Fridman?

  Yeah, you're not, now I'm not focused on.

• 1:34:09George Hotz?

  I don't, I don't know how to do, oh, like, I guess I could sell it to more people, but I want to make the system better.

• 1:34:13Lex Fridman?

  Better, better.

• 1:34:13George Hotz?

  And I don't know how to, I mean.

• 1:34:15Lex Fridman?

  But what's the messaging here? I got a chance to talk to Elon and, and he, he basically said that the human factor doesn't matter. You know, the human doesn't matter because the system will perform, there'll be sort of a, sorry to use the term, but like a singular, like a point where it gets just much better. And so the human, it won't really matter. But it seems like that human catching. The system, when it gets into trouble, is like the thing which will make something like reinforcement learning work. So how do you, how do you think messaging for Tesla, for you, for the industry in general should change?

• 1:34:55George Hotz?

  I think our messaging is pretty clear. At least like, our messaging wasn't that clear in the beginning, and I do kind of fault myself for that. We are proud right now to be a level two system. We are proud to be level two. If we talk about level four, it's not with the current hardware, it's not going to be just a magical OTA upgrade, it's going to be new hardware, it's going to be very carefully thought out. Right now, we are proud to be level two, and we have a rigorous safety model. I mean, not like, like, okay, rigorous, who knows what that means. But we at least have a safety model, and we make it explicit, it's in safety.md in OpenPilot. And it says, seriously though.

• 1:35:30Lex Fridman?

  Safety.md.

• 1:35:31George Hotz?

  Safety.md.

• 1:35:32Lex Fridman?

  This is brilliant, this is so Android.

• 1:35:35George Hotz?

  So, well, this is, this is the, this is the safety model, and I like to have conversations like, if, like, you know, sometimes people will come to you and they're like, your system's not safe. Okay. Have you read my safety docs? Would you like to have an intelligent conversation about this? And the answer is always no, they just like scream about, it runs Python! Okay, what, so you're saying that, that because Python's not real time? Python not being real time never causes disengagements. Disengagements are caused by, you know, the model is QM. But, safety.md says the following. First and foremost, the driver must be paying attention at all times. I don't, I do, I still consider the software to be alpha software until we can actually enforce that statement, but I feel it's very well communicated to our users.

• 1:36:19George Hotz?

  Two more things. One is, the user must be able to easily take control of the vehicle at all times. So if you step on the gas or brake with OpenPilot, it gives full manual control back to the user, or press the cancel button.

• 1:36:35George Hotz?

  Step two, the car will never react so quickly. We define so quickly to be about one second. That you can't react in time. And we do this by enforcing torque limits, braking limits, and acceleration limits. So we have.

• 1:36:50George Hotz?

  Like, our torque limit's way lower than Tesla's. This is another potential. If I could tweak autopilot, I would lower their torque limit and I would add driver monitoring. Because autopilot can jerk the wheel hard. Open pilot can't. It's we we we limit. And all this code is open source, readable. And I believe now it's all Misra C compliant.

• 1:37:11Lex Fridman?

  What's that mean?

• 1:37:13George Hotz?

  Misra is like the automotive coding standard. At first, I, you know, I've come to respect. I've been reading like the standards lately and I've come to respect them. They're actually written by very smart people.

• 1:37:24Lex Fridman?

  Yeah, they're brilliant people, actually. They have a lot of experience. They're sometimes a little too cautious, but in this case, it pays off.

• 1:37:33George Hotz?

  Misra is written by like computer scientists and you can tell by the language they use. You can tell by the language they use. They they talk about like whether certain. Conditions in Misra are decidable or undecided. And I you mean like the halting problem? And yes, all right, you've you've earned my respect. I will read carefully what you have to say and we want to make our code compliant with that.

• 1:37:52Lex Fridman?

  All right, so you're proud level two, beautiful. So you were the founder and I think CEO of Com AI, then you were the head of research. What the heck are you now? What's your connection to Com AI?

• 1:38:03George Hotz?

  I'm the president, but I'm one of those like unelected, unelected presidents of like, like a small dictatorship country, not one of those like elected presidents.

• 1:38:11Lex Fridman?

  Oh, so you're like Putin when he was like the, yeah, I got you.

• 1:38:16Lex Fridman?

  So there's a what's the governance structure? What's the what's the future of Com AI finance? I mean.

• 1:38:23Lex Fridman?

  Yeah, as a business, do you want, are you just focused on getting things right now? Making some small amount of money in the meantime. And then when it works, it works and you scale.

• 1:38:34George Hotz?

  Our burn rate is about 200 K a month and our revenue is about 100 K a month. So we need to 4X our revenue. But we haven't like tried very hard at that yet.

• 1:38:44Lex Fridman?

  And the revenue is basically selling stuff online.

• 1:38:46George Hotz?

  Yeah, we sell stuff shop.com dot AI.

• 1:38:48Lex Fridman?

  Is there other? Well, OK, so you'll you'll have to figure out.

• 1:38:51George Hotz?

  That's our that's our only see, but to me, that's like respectable revenues. We make it by selling products to consumers. We're honest and transparent about what they are.

• 1:39:02Lex Fridman?

  Most actually level four companies, right? Cause you could easily start blowing up like smoke, like overselling the hype and feeding into getting some fundraisers. Oh, you're the guy, you're a genius because you hacked the iPhone.

• 1:39:18George Hotz?

  Oh, I hate that. I hate that. Yeah, I can trade my social capital for more money. I did it once. I almost regret it doing it the first time.

• 1:39:26Lex Fridman?

  Well, on a small tangent, what's your, you seem to not like fame and yet you're also drawn to fame. What's, where have you, where are you, where are you on that currently? Have you had some introspection, some soul searching?

• 1:39:43George Hotz?

  Yeah, I, I actually, I've come to a pretty stable position on that. Like after the first time, I realized that I don't want attention from the masses. I want attention from people who I respect.

• 1:39:56Lex Fridman?

  Who do you respect?

• 1:39:58George Hotz?

  I can give a list of people.

• 1:40:00Lex Fridman?

  So are these like Elon Musk type characters?

• 1:40:03George Hotz?

  Yeah.

• 1:40:05George Hotz?

  Or actually, you know what? I'll make it more broad than that. I won't make it about a person. I respect skill. I respect people who have skills, right? And I would like to like be, I'm not gonna say famous, but be like known among more people who have like real skills.

• 1:40:23Lex Fridman?

  Who in cars do you, do you think have skill, not do you respect?

• 1:40:31George Hotz?

  Oh, Kyle Vogt has skill. A lot of people at Waymo have skill and I respect them. I, I, I respect them as engineers. Like I can think, I mean, I think about all the times in my life where I've been like dead set on approaches and they turn out to be wrong.

• 1:40:45Lex Fridman?

  Yeah.

• 1:40:46George Hotz?

  So I mean, this might, I might be wrong. I, I accept that. I accept that there's a decent chance that I'm, I'm wrong.

• 1:40:53Lex Fridman?

  And actually, I mean, having talked to Chris Armson, Sterling Anderson, those, those guys, I mean, I deeply respect Chris. I just admire the guy. He's legit. When you drive a car through the desert, when everybody thinks it's impossible, that's, that's legit.

• 1:41:08George Hotz?

  And then I also really respect the people who are like writing the infrastructure of the world, like the Linus Torvalds and the Chris Latner.

• 1:41:14Lex Fridman?

  They're doing the real work.

• 1:41:15George Hotz?

  I know. They're doing the real work.

• 1:41:18Lex Fridman?

  Having talked to Chris, like Chris Latner, you realize, especially when they're humble, it's like you realize, oh, you guys, we're just using your.

• 1:41:26George Hotz?

  Oh, yeah.

• 1:41:26Lex Fridman?

  All the hard work that you did. Yeah, that's incredible. What do you think?

• 1:41:32Lex Fridman?

  Mr. Anthony Lowndowsky. What do you, he's a, he's another mad genius.

• 1:41:38George Hotz?

  Sharp guy. Oh, yeah.

• 1:41:39Lex Fridman?

  What, um, do you think he might long term become a competitor?

• 1:41:44George Hotz?

  Oh, Tacoma? Well, so I think that he has the other right approach. I think that right now there's two right approaches. One is what we're doing and one is what he's doing.

• 1:41:54Lex Fridman?

  Can you describe, I think it's called pronto AI, he started, you think, do you know what the approach is? I actually don't know.

• 1:41:59George Hotz?

  Embark is also doing the same sort of thing. The idea is almost that you want to, so if you're, I can't partner with Honda and Toyota. Honda and Toyota are, uh, like 400,000 person companies. It's not even a company at that point. Like, I don't think of it like, I don't personify it. I think of it like an object, but a trucker drives for a fleet. Maybe that has like, some truckers are independent. Some truckers drive for fleets with 100 trucks. There are tons of independent trucking companies out there. Start a trucking company and drive your costs down. Or figure out how to drive down the cost of trucking. Another company that I really respect is Nato. Actually, I respect their business model.

• 1:42:44George Hotz?

  Nato sells a driver monitoring camera and they sell it to fleet owners. If I, if I owned a fleet of cars and I could pay, uh, you know, 40 bucks a month to monitor my employees. This is gonna, it like reduces accidents 18%. It's, it's so like that in the space, that is like the business model that I like most respect. Cause they're creating value today.

• 1:43:11Lex Fridman?

  Yeah, which is, uh, that's a huge one is how do we create value today with some of this? And the lane keeping thing is huge. And, uh, it sounds like you're creeping in or full steam ahead on the driver monitoring too. Which I think actually where the short term value, if you can get it right. I still, um, I'm not a huge fan of the statement that everything has to have driver monitoring. Boy, I agree with that completely, but, um, that statement usually misses the point that to get the experience of it right is not trivial.

• 1:43:38George Hotz?

  Oh, no, not at all. In fact, like, so right now we have, uh, I think the timeout depends on speed of the car.

• 1:43:46George Hotz?

  But we want to depend on like the scene state. If you're on like an empty highway, it's very different if you don't pay attention than if like you're like coming up to a traffic light.

• 1:43:58Lex Fridman?

  And long term, it should probably learn from the driver because that's to do. I watched a lot of video. We've built a smartphone detector just to analyze how people are using smartphones. And people are using it very differently. There's this, um. So texting styles.

• 1:44:14George Hotz?

  We haven't watched nearly enough of the videos. We have, I got millions of miles of people driving cars.

• 1:44:19Lex Fridman?

  In this moment, I spent a large fraction of my time just watching videos because it's never fails to, to learn. Like it never, I've never failed from a video watching session to learn something I didn't know before. In fact, I usually, um, like when I eat lunch, I'll sit, especially when the weather is good and just watch pedestrians. Uh, with an eye to understand like from a computer vision eye, just to see. Can this model, can you predict what are the decisions made? And there's so many things that we don't understand.

• 1:44:49George Hotz?

  This is what I mean about the state vector.

• 1:44:51Lex Fridman?

  Yeah, it's, uh, I'm trying to always think like, cause I'm understanding in my human brain. How do we convert that into? How, how hard is the learning problem here? I guess is the fundamental question. So something that's, uh, from a hacking perspective, this is always comes up, especially with folks. Well, first, the most popular question is, uh, the trolley problem, right? So that's not a sort of a serious, uh, problem. There are some ethical questions, I think, that arise. Uh, maybe you wanna, do you, do you think there's any ethical, serious ethical questions?

• 1:45:27George Hotz?

  We have, we have a solution to the trolley problem at Com AI. Well, so there is actually an alert in our code, ethical dilemma detected. It's not triggered yet. We don't, we don't know how yet to detect the ethical dilemmas, but we're a level two system. So we're going to disengage and leave that decision to the human.

• 1:45:41Lex Fridman?

  You're such a troll.

• 1:45:43George Hotz?

  No, but the trolley problem deserves to be trolled.

• 1:45:45Lex Fridman?

  Yeah, you're right. That's a beautiful answer, actually.

• 1:45:48George Hotz?

  I know, I gave it to someone who was like, like, sometimes people ask, like you asked about the trolley problem. Like you can have a kind of discussion about it, like when you get someone who's like really like earnest about it. Because it's the kind of thing where. If you ask a bunch of people in an office whether we should use a sequel stack or no sequel stack, if they're not that technical, they have no opinion. But if you ask them what color they want to paint the office, everyone has an opinion on that. And that's why the trolley problem is.

• 1:46:12Lex Fridman?

  I mean, that's a beautiful answer. Yeah, we're able to detect the problem and we're able to pass it on to the human.

• 1:46:18George Hotz?

  There you go.

• 1:46:18Lex Fridman?

  Wow, I've never, never heard anyone say it. This is your nice escape route. Okay, but.

• 1:46:23George Hotz?

  Proud level two.

• 1:46:24Lex Fridman?

  I'm proud level two. I love it. So the other thing that people, you know, have some concern about with AI in general is hacking. So how hard is it, do you think, to hack an autonomous vehicle, either through physical access or through the more sort of popular now, these adversarial examples on the sensors?

• 1:46:44George Hotz?

  Okay, the adversarial examples, one. You want to see some adversarial examples that affect humans?

• 1:46:49Lex Fridman?

  Hmm.

• 1:46:50George Hotz?

  Right? Oh, well, there used to be a stop sign here, but I put a black bag over the stop sign and then people ran it.

• 1:46:57Lex Fridman?

  Right.

• 1:46:58George Hotz?

  Adversarial.

• 1:46:59Lex Fridman?

  Yeah.

• 1:47:00George Hotz?

  Right? Like, like, like, there's tons of human adversarial examples, too. The question in general about like, security, if you saw, something just came out today and like, there are always such hypey headlines about like, how Navigate on Autopilot was fooled by a GPS spoof to take an exit.

• 1:47:16Lex Fridman?

  Right.

• 1:47:17George Hotz?

  At least that's all they could do was take an exit. If your car is relying on GPS in order to have a safe driving policy, you're doing something wrong. If you're relying, and this is why V2V is such a terrible idea. V2V now relies on both parties getting communication right. This is not even, so I think of, uh, security, uh, security is like a special case of safety.

• 1:47:44Lex Fridman?

  Yeah.

• 1:47:45George Hotz?

  Safety is like, we put a little, you know, piece of caution tape around the hole so that people won't walk into it by accident. Security is like put a 10-foot fence around the hole so you actually physically cannot climb into it with barbed wire on the top and stuff, right? So like, if you're designing systems that are like unreliable, they're definitely not secure. Your car should always do something safe using its local sensors.

• 1:48:09Lex Fridman?

  Right.

• 1:48:10George Hotz?

  And then the local sensor should be hardwired. And then, could somebody hack into your CAN bus and turn your steering wheel and your brakes? Yes, but they could do it before AI too, so.

• 1:48:19Lex Fridman?

  Let's think out of the box on some things. So do you think teleoperation has a role in any of this? So remotely stepping in and controlling the cars?

• 1:48:30George Hotz?

  No, I think that if safety, if the safety operation by design requires a constant link to the cars, I think it doesn't work.

• 1:48:44Lex Fridman?

  So that's the same argument you're using for V2I, V2V?

• 1:48:47George Hotz?

  Well, there's a lot of non-safety critical stuff you can do with V2I. I like V2I. I like V2I way more than V2V. Because V2I is already like, I already have internet in the car, right? There's a lot of great stuff you can do with V2I. Like, for example, you can, well, I already have, Waze is V2I, right? Waze can route me around traffic jams. That's a great example of V2I. And then, okay, the car automatically talks to that same service, like.

• 1:49:12Lex Fridman?

  So it's improving the experience, but it's not a fundamental fallback for safety?

• 1:49:15George Hotz?

  No, if any of your, if any of your, if any of your things that require wireless communication are more than QM, like have an ASL rating, you shouldn't.

• 1:49:27Lex Fridman?

  You previously said that life is work. And that you don't do anything to relax. So how do you think about hard work? What is it, what do you think it takes to accomplish great things? And there's a lot of people saying that there needs to be some balance. You know, you need to, in order to accomplish great things, you need to take some time off. You need to reflect and so on.

• 1:49:51Lex Fridman?

  And then some people are just insanely working, burning the candle on both ends. How do you think about that?

• 1:49:57George Hotz?

  I think I was trolling in the Syrah interview when I said that. Off camera, right before, I smoked a little bit of weed, like, you know, come on. This is a joke, right? Like, I do nothing to relax. Look where I am. I'm at a party, right?

• 1:50:09Lex Fridman?

  Yeah, that's true.

• 1:50:11George Hotz?

  So no, no, of course, I don't. When I say that life is work, though, I mean that like, I think that what gives my life meaning is work. I don't mean that every minute of the day you should be working. I actually think this is not the best way to maximize results. I think that if you're working 12 hours a day, you should be working smarter and not harder.

• 1:50:31Lex Fridman?

  Well, so it gives, work gives you meaning. For some people, other sources of meaning is personal relationships, like family and so on. You've also, in that interview with Suraj, or the trolling, mentioned that one of the things you look forward to in the future is AI girlfriends.

• 1:50:49George Hotz?

  Yes.

• 1:50:50Lex Fridman?

  So, that's a topic that I'm very much fascinated by, not necessarily girlfriends, but just forming a deep connection with AI.

• 1:50:58George Hotz?

  Yeah.

• 1:50:59Lex Fridman?

  What kind of system do you imagine when you say, AI girlfriend, whether you were trolling or not?

• 1:51:04George Hotz?

  No, that one I'm very serious about. And I'm serious about that on both a shallow level and a deep level. I think that VR brothels are coming soon and are gonna be really cool. It's not cheating if it's a robot. I see the slogan already.

• 1:51:19George Hotz?

  But...

• 1:51:20Lex Fridman?

  There's, I don't know if you've watched, I just watched the Black Mirror episode.

• 1:51:24George Hotz?

  I watched the last one, yeah.

• 1:51:25Lex Fridman?

  Yeah, yeah.

• 1:51:27George Hotz?

  Oh, the Ashley 2 one? Or the...

• 1:51:31Lex Fridman?

  No, where there's two friends were having sex with each other in...

• 1:51:36George Hotz?

  Oh, in the VR game?

• 1:51:37Lex Fridman?

  In the VR game. It's the two guys, but one of them was a female, yeah.

• 1:51:43George Hotz?

  Yeah.

• 1:51:43Lex Fridman?

  Which is another mind blowing concept, that in VR, you don't have to be the form. You can be two animals having sex. It's weird.

• 1:51:54George Hotz?

  I mean, I'll see how nice of the software maps the nerve endings, right?

• 1:51:56Lex Fridman?

  Yeah, it's weird. I mean, yeah, they sweep a lot of the fascinating, really difficult, technical challenges under the rug, like assuming it's possible to do the mapping of the nerve endings, then...

• 1:52:07George Hotz?

  I wish, yeah, I saw that, the way they did it with the little, like, stim unit on the head, that'd be amazing.

• 1:52:13George Hotz?

  So, well, no, no, on a shallow level, like, you could set up, like, almost a brothel with, like, real dolls and Oculus Quests, write some good software. I think it'd be a cool novelty experience. But no, on a deeper, like, emotional level...

• 1:52:29George Hotz?

  I mean, yeah, I would really like to to to fall in love with with with a machine.

• 1:52:34Lex Fridman?

  Do you see yourself having a a long term relationship of the kind monogamous relationship that we have now with a robot, with a AI system even, not even just a robot?

• 1:52:49George Hotz?

  So I think about maybe my ideal future. When I was 15, I read Ali Yarukowski's early writings on the singularity and like that AI is going to surpass human intelligence massively.

• 1:53:10George Hotz?

  He made some Moore's Law based predictions that I mostly agree with. And then I really struggled for the next couple of years of my life. Like, why should I even bother to learn anything? It's all gonna be meaningless when the machines show up.

• 1:53:22Lex Fridman?

  Right.

• 1:53:23George Hotz?

  Um, maybe, maybe when I was that young, I was still a little bit more pure and really like clung to that. And then I'm like, well, the machines ain't here yet, you know, and I seem to be pretty good at this stuff. Let's uh, let's try my best, you know, like, what's the worst that happens? But the best possible future I see is me sort of merging with the machine. And the way that I personify this is in a long term monogamous relationship with a machine.

• 1:53:49Lex Fridman?

  Oh, you don't think there's a room for another human in your life if you really truly merge with another machine?

• 1:53:55George Hotz?

  I mean, I see merging. I see like the best interface to my brain is like the same relationship interface to merge with an AI, right? What does that merging feel like? I've seen, I've seen couples who've been together for a long time and like, I almost think of them as one person, like couples who spend all their time together and.

• 1:54:18Lex Fridman?

  That's fascinating. You're actually putting, what does that merging actually looks like? It's not just a nice channel. Like, a lot of people imagine it's just an efficient link, search link to Wikipedia or something.

• 1:54:30George Hotz?

  I don't believe in that.

• 1:54:31Lex Fridman?

  But it's more, you're saying that there's the same kind of, the same kind of relationship you have with another human, that's a deep relationship, that's what merging looks like. That's, that's pretty uh.

• 1:54:40George Hotz?

  I don't believe that link is possible. Um, I think that that link, so you're like, oh, I'm gonna download Wikipedia right to my brain.

• 1:54:46Lex Fridman?

  Yeah.

• 1:54:46George Hotz?

  My reading speed is not limited by my eyes.

• 1:54:49Lex Fridman?

  Yeah.

• 1:54:49George Hotz?

  My reading speed is limited by my inner processing loop.

• 1:54:52Lex Fridman?

  Yep.

• 1:54:53George Hotz?

  And to like bootstrap that sounds kind of unclear how to do it and horrifying. But if I am with somebody, and I'll use a somebody, who is making a super sophisticated model of me, and then running simulations on that model, I'm not going to get into the question whether the simulations are conscious or not. I don't really want to know what it's doing. But using those simulations to play out hypothetical futures for me, deciding what things to say to me to guide me along a path and that's how I envision it.

• 1:55:25Lex Fridman?

  So on that path to AI of superhuman level intelligence, you've mentioned that you believe in the singularity, that singularity is coming.

• 1:55:34George Hotz?

  Yeah.

• 1:55:35Lex Fridman?

  Again, could be trolling, could be not, could be part. All trolling has truth in it.

• 1:55:39George Hotz?

  I don't know what that means anymore. What is the singularity?

• 1:55:42Lex Fridman?

  Yeah, so that's, that's really the question. How, how many years do you think before the singularity, what form do you think it will take? Does that mean fundamental shifts in capabilities of AI? Does it mean some other kind of ideas?

• 1:55:54George Hotz?

  I'm a, maybe this is just my roots, but so I can buy a human being's worth of compute for like a million bucks today. It's about one TPU pod V3. I want like, I think they claim a hundred paid flops. That's being generous. I think humans are actually more like 20. So that's like five humans. That's pretty good. Google needs to sell their TPUs.

• 1:56:13George Hotz?

  But I could buy, I could buy, I could buy GPUs. I could buy a stack of like, I'd buy 1080 Ti's, build data center full of them. And for a million bucks, I can get a human worth of compute. But when you look at the total number of flops in the world, when you look at human flops, which goes up very, very slowly with the population, and machine flops, which goes up exponentially, but it's still nowhere near. I think that's the key thing to talk about when the singularity happened. When most flops in the world are silicon and not biological, that's kind of the crossing point. Like they are now the dominant species on the planet.

• 1:56:51Lex Fridman?

  And just looking at how technology is progressing, when do you think that could possibly happen? Do you think it would happen in your lifetime?

• 1:56:58George Hotz?

  Oh yeah, definitely in my lifetime. I've done the math. I like 2038 because it's the Unix timestamp rollover.

• 1:57:06Lex Fridman?

  Yeah, beautifully put. So you've,

• 1:57:11Lex Fridman?

  you've said that the meaning of life is to win. If you look five years into the future, what does winning look like?

• 1:57:19George Hotz?

  So.

• 1:57:24George Hotz?

  There's a lot of, I can go into like technical depth to what I mean by that, to win. It may not mean, I was criticized for that in the comments. Like, doesn't this guy want to like save the penguins in Antarctica? Or like, look, man, you know, listen to what I'm saying. I'm not talking about like I have a yacht or something.

• 1:57:45George Hotz?

  I am an agent. I am put into this world. And I don't really know what my purpose is. But if you're a reinforcement, if you're, if you're an intelligent agent and you're put into a world, what is the ideal thing to do? Well, the ideal thing mathematically, you can go back to like Schmidt-Huber theories about this, is to build a compressive model of the world, to build a maximally compressive, to explore the world such that your exploration function maximizes the derivative of compression of the past. Schmidt-Huber has a paper about this. And like, I took that kind of as like a personal goal function. So when I mean to win, I mean like, maybe, maybe this is religious, but like, I think that in the future, I might be given a real purpose.

• 1:58:29George Hotz?

  Or I may decide this purpose myself. And then at that point, now I know what the game is and I know how to win. I think right now, I'm still just trying to figure out what the game is. But once I know.

• 1:58:39Lex Fridman?

  So you have, you have imperfect information. You have a lot of uncertainty about the reward function and you're discovering it.

• 1:58:46George Hotz?

  Exactly.

• 1:58:46Lex Fridman?

  But the purpose is.

• 1:58:47George Hotz?

  That's a better way to put it.

• 1:58:48Lex Fridman?

  So the purpose is to maximize it while you have it, a lot of uncertainty around it. And you're both reducing the uncertainty and maximizing at the same time. And so that's at the technical level.

• 1:59:00George Hotz?

  What is the, if you believe in the universal prior, what is the universal reward function? That's the better way to put it.

• 1:59:07Lex Fridman?

  So that win is interesting. I think I speak for everyone in saying that I wonder what that reward function is for you. And I look forward to seeing that in five years, in 10 years. I think a lot of people, including myself, are cheering you on, man. So I'm, I'm happy you exist and I wish you the best of luck. Thanks for talking today, man.

• 1:59:31George Hotz?

  Thank you. This is a lot of fun.