00:45:01 · Dmitri Dolgov
[00:45:01] S0: Let me answer it maybe at different levels. There is the most specific version of generalizability, where we think about the Waymo Driver generalizing to different cities, different countries, different vehicles, different sensor configuration. That is done. It generalizes extremely well. We're in 15 cities across The US right now. We're operating the fleet on the fifth generation and the sixth generation of the driver. It has different sensors, different vehicles, the JLR base and the Ojai, and it's the same driver. That will continue. We are focusing on ride hailing today. There's you know, north of 13,000,000,000,000 miles being driven worldwide, predicted to be about 20 by the end of the decade, so why wouldn't you want to have the benefits of autonomous vehicles in all of the driving modalities, whether it's ride hailing, whether it's personally owned vehicles, whether it's trucking and deliveries. So that will, you know, same product, slight modifications, will generalize extremely well, something that, you know, we're working towards. Then there's the question of, you know, will it generalize to other versions of physical AI agents? And there I think my answer is a little bit more nuanced. The perceptual parts I would expect to generalize really well. Sensing and understanding the environment. I think the fundamentals that we talked about, like training recipes, like how do you properly do, you know, close loop eval and RL in those domains, the fundamental infrastructure that supports all of that, that will generalize pretty well. The specifics of, you know, how do you build a simulator for a humanoid robot versus how you build it for a car, there's gonna be a lot of work there. And the actual generative part, if your, you know, physical kind of embodiment of your agent is very different, that planning generative part, that will require some adaptation as well.
00:47:16 · Ruchi Sanghvi, Dmitri Dolgov
[00:47:16] S1: I have one on behalf of my son, who is here on the first row. He's nine years old. He's a huge fan of robotics and cars as well. And he wanted to know whether he would still need to learn to drive by the time he's 16 or whether he'll be able to afford an autonomous vehicle by the time he's 16. [00:47:36] S0: Would you like to drive? Yes. Yeah. So then you can. Right? I I think it's gonna be a [00:47:43] S1: little Maybe of questions for me. Can I buy him an autonomous vehicle by the time he's 16? [00:47:49] S0: I get the question from my wife quite a bit. Like, when can I autonomously send my kids in Waymo without needing to be in the car with them? So I think we're in 15 cities today. We're doing about half a million trips every week. Driving about 5,000,000 fully autonomous miles, and it's scaling exponentially and very fast. So last year we were operating fully autonomously in five cities. Right now we're in 15 in The US, expanding internationally. The driver is generalizing pretty well. So I think it's not that far where you'll be in a place where you, you know, any major city across the globe, you get in there and you take a Waymo and it takes you anywhere you want to go. So I have three small kids. We are definitely kind of a Waymo primary family. Which is good. Okay. [00:48:46] S1: But that's not the same as owning one. When can I buy one?
00:48:46 · Ruchi Sanghvi, Dmitri Dolgov, Aditya Agarwal
[00:48:46] S1: But that's not the same as owning one. When can I buy one? [00:48:49] S0: So I think that will as I mentioned, I don't have a timeline for you yet, but this is just a different product. So we're starting with ride hailing because there's more of a commercial market, and we get higher utilization of the vehicles. But as that expands, if you want to have your own personal vehicle or in your Waymo that people like to keep their stuff in them and you don't want to share, yeah, absolutely. Why not? [00:49:15] S2: Very cool. I mentioned this to you earlier, Dmitri, but I do think that there are not too many moments that I can recall that I've had the same kind of almost like visceral, almost like a religious experience when you first get into Waymo and you're sitting there and the thing is driving itself, right? It's one of those moments where you just have a magical experience and it kind of reminds Can [00:49:38] S1: I just add to that? So I had the privilege of putting Neal Stephenson, the author of Snow Crash, and Ken Liu, the translator of The Three Body Problem in their first Waymo's in San Francisco. [00:49:51] S0: And Wow.
00:49:51 · Dmitri Dolgov, Ruchi Sanghvi, Aditya Agarwal
[00:49:51] S0: And Wow. [00:49:52] S1: I know. So that, I was like, oh my god, I cannot believe I get to give you that experience. And I will forward you the email that I got from Ken Liu. Despite being, like, you know, one of the foremost sci fi authors of our generation and our time, he said this was an absolute magical experience. My Waymo was, like, you know, debating whether to pull up next to another car and no language, just a hand wave gesture from the driver and the Waymo pulled up having understood that. And he was like, this was an absolutely magical experience. I had to share that with you. [00:50:25] S2: Thank you. Yeah. And we're all here to build great things, but I think it's just sometimes worth having these experiences to remind you how magical technology can be. So thank you for everything you and the team have built, not only I think at Waymo, but I suspect that the core underpinnings of what you have built as a system, as I think a company, will end up kind of affecting all of us in many ways in many different companies in the future. So thank you for everything. It's a pleasure and privilege to have you here. [00:50:57] S0: Thanks for having me.