Waymo's Dmitri Dolgov: 20 Million Rides and the Road to Full Autonomy

We have an unbelievable treat next: a founder who's touching a ton of lives and who's been at it for a very long time. How many people here have been in a Waymo?

Whoo. Whoo. Okay. Well, that's a relief. This chef eats his own food.

I do. I do a lot of it.

All right. And how many people love the Waymo experience?

Whoo. Rock on. I'm a daily active now. It's incredible.

Excellent.

Excellent. Thank you.

We have here the creator, a man who has been at this mission for, get this, founders who've been in AI since 2022, almost 20 years building in the autonomous vehicle challenge. And he has not only been at this, he's been at it in the great times and the tough times. He's been persistent, and he has created something that is unlike anything else on Earth. Truly exceptional. Please join me in welcoming Dmitri Dolgov.

Thanks. Great to be here.

All right, Dmitri. So we've got about 25 minutes together. The goal is to understand a little bit about you, what makes you tick, what has made you persist since the early days of the DARPA Challenge 21 years ago, all the way through Waymo, from the early days to today and the future. Let's start with you, and then we'll get into technology very quickly. Sound good?

Sounds good.

So, Dmitri, you are known by your team as technically brilliant, incredibly intense, but also very kind and humble. You were born in the Soviet Union, raised in the States, and then chose to go back to one of the most prestigious, intense physics programs on the planet in Moscow. How did those first few years of your life shape you, and how did they shape your character?

My parents went to the same school, so that, to a large degree, drove my decision to go back and go to high school. I actually traveled around quite a bit. I spent a year in Japan, then went to high school in the States, and came back to college to do math and physics in Russia. That was the same school that my parents went to, and I grew up hearing stories about what it was like to be at that place. I really wanted to go back.

And I think, in terms of how it shaped me, it really set the foundation—the technical foundation. In those early days in college, one of the most important things is acquiring the ability to learn and independently explore. I think that really helped me in my future career.

Now, you did this very intense program at Moscow Institute of Physics and Technology, and then you decided to keep going on the AI path. You earned your PhD, also in AI.

Mm-hmm.

And then you pretty quickly were attracted to autonomous vehicles. In 2005—

Yeah.

—you were a part of the DARPA Challenge. Can you tell us about those early days? What drew you to autonomy?

That was a light-switch moment for me. In the early days, when I went to college and then to grad school, it was more about learning the fundamentals. I didn't have a clear picture or idea at all of what I wanted to do afterward. Then I think the timing was incredibly lucky: when I was finishing up grad school, the Grand Challenge and then the Urban Challenge—the one that I took part in—were happening, and it clicked.

The technology was incredibly interesting. The mission was so powerful that nothing else came close, and there was a real product there. You could be hands-on and experience it yourself. It really checked all the boxes for me. As you said, that's been 20-some years ago. Who's counting? I've never looked back, and that's what I've been doing since.

Amazing. So Waymo started out of a project at Stanford Automotive Lab. There were 2 sides of this building. There was the autonomy side, and then there was a solar-car side. Fun fact: I was an idealist. I worked on the solar car. I got that bet very wrong. You bet on autonomy. Tell us about the first few years of Waymo, from 2009 through the formative years.

We started in 2009. That was at the time the Google self-driving car project. The first couple of years were all about learning the problem space and understanding what it meant to try to put an autonomous vehicle on public roads.

In service of those goals of learning and understanding the problem space, we created a couple of goals for ourselves. One was to drive 100,000 miles total in full autonomy, which at the time was unheard of. The second one was to drive 10 routes, each one 100 miles long. They were all over the Bay Area, chosen to be very difficult, and we had to do each one from beginning to end in full autonomy.

There was still a person behind the wheel who could take control, but the challenge was to complete each one without an intervention. It was a small team of us—about a dozen people. It was the early, crazy startup days: everybody working 24/7, writing code and building hardware during the day, then doing some testing at night. It took us about 18 months to complete both of those challenges.

Incredible. It seemed impossible at the time.

Yeah.

Now you guys are on hundreds of millions of miles.

Absolutely.

Okay, so the early Waymo days: extreme challenge, starting to achieve. Over the next few years, you developed a reputation on your team for grinding really hard. You were sleeping at the office. Tell us about Dmitri in the first few years of Waymo and how you formed your leadership style.

I have to say, those early days were probably the most fun I've ever had in my professional life. It was the momentum and the pace of the early startup days, when you are making so much progress every hour of every day.

And you're doing everything. You're working on setting up the hardware in the cars, configuring and calibrating the sensors and your pose-estimation system, and writing software during the day. It's everything: the core of the software, the algorithms that drive the car, all of the tools and UIs, and the user experience in the car.

So you're doing everything, you're learning at an insane rate, and you're making progress at an insane rate. Those were the early days of Project Chauffeur. In those couple of years, we convinced ourselves that this was worth pursuing, so we doubled down and started actually building toward the future of a fully autonomous product.

Okay, so exciting first few years: intense, fast-paced, technically really difficult. Now take us to 2016–17 for a moment. This was a period when we actually had a hype cycle in AI. Turns out there have been a few of them. And autonomous vehicles—

Yeah.

—were at the center of that hype cycle. I remember so many companies going after this. And then there was a massive slump.

Mm-hmm.

When most people gave up or failed or fell apart, you guys persisted, and you were a leader in that persistence. For all the builders in this room, how did you navigate through the hard times?

First, a comment on what these cycles look like to me and how I've seen them. You said there have been many—some in AV, but more generally. Often, what leads to a cycle like this is some breakthrough that leads to very rapid progress in the early parts of the problem, and very rapid advancements still in the early part of the problem.

Right.

In AVs, the problem has always had this property: it's very easy to get started, but it's very difficult to take it all the way to a real product, full autonomy, and superhuman performance.

It's somewhat natural, given those ingredients, that whenever there's been a big breakthrough in technology—whether it's convolutional nets, transformers, or large language models—it has led to this cycle: “Okay, now the problem is going to be...” It reshapes the early part of the curve, but it doesn't change the long tail of it.

Mm-hmm.

For us, I think it was understanding that it's not going to be an easy problem, but it's a very important one.

Hmm.

Believing in the mission is important. Today, worldwide, somebody loses their life to a crash on our roads every 26 seconds. I guess it's the combination of knowing that the mission is really important and understanding what you're up against—not looking for easy wins, quick solutions, or silver bullets—that helps the team have the right stamina to go the distance.

Brilliant. So you guys were in this moment where it was really easy to get started. A lot of people got there.

Yeah.

But you guys actually persisted and got through to the other side with a truly magical experience. Pretty much every hand in this room went up.

A truly magical experience because of that persistence. Let’s talk about technology today. A lot of people are talking about world models. You have had all the components of world models for many years. How do you think about a world model, and what is Waymo’s version of a world model?

Yeah, there are a few terms that people use nowadays. People talk about world models, world-action models, omni models, and visual-language-action models. At the core of each, there’s an ingredient that is relevant and really important for Waymo and for what we’ve been building in our AI ecosystem. At the core of our AI ecosystem is what we call the Waymo Foundation Model, and it powers 3 main pillars of our AI and our technology: the driver, the simulator, and the critic.

Mm-hmm.

Right? Those are very related but distinct tasks. At the core of what our foundation model needs to be capable of are things like understanding how the world works—the physics and dynamics of the physical world.

Mm-hmm.

It needs to understand what it means to be a good driver and how the actions of that driver or AI agent affect other agents in the world.

Mm-hmm.

Then we need to instantiate those capabilities in the physical agent that we’re putting on the roads. In a way, that foundational model that we’ve been building over the years is a multimodal world-action-language model.

Right.

It’s multimodal in that it needs to be able to reason about not just images or video, but also other sensors like clay, lidars and radars. It’s a world-action model in that it really has to have a deep, precise understanding of the 3D spatial properties of the world, the dynamics, the physics, and the behavioral aspects of other agents like cars, pedestrians, cyclists, and so forth. We are not just passively modeling those worlds; we’re active participants in them.

Mm-hmm.

So not only does the world model have to be controllable, but we also need to have a deep understanding of what it means to be a good agent in that world. Finally, it’s aligned with language, and that allows us to pull in the general world knowledge of a VLM into our model—

Hmm.

—that is very, very useful in giving us a boost in understanding the semantics and the deep social aspects of driving. We’ve been working on productionizing that model for years, and it requires an extremely high degree of performance, accuracy, and realism in every aspect of what we just talked about.

Brilliant. So with this driver-simulator-critic architecture, there’s also been a lot of conversation about end-to-end architectures. Is that the appropriate dichotomy? How do we think about the approach to getting us to extremely performant, efficient autonomous vehicles that are—

Mm.

—totally generalizable?

Yeah. To be very clear, the world model that I just described—the Waymo Foundation Model—is an end-to-end model. When we talk about an end-to-end model, we typically mean that it’s one model that goes from sensors to decisions or actions. There are some very nice properties of such a model. One of the most important ones is that it learns the right, rich representations between different components of the system, like the encoder and the decoder, or the perception and planning parts of your system.

Mm.

That’s as opposed to something where that interface is engineered, which is not sufficient for a task like driving. Now, I do think there’s a false dichotomy there. There’s end-to-end or something else. In my mind, it’s always been the question of: It’s end-to-end, and then what else?

Mm.

What else do you need to build if you want to have a product that is fully autonomous, has a superhuman level of safety, and that you want to deploy at scale and drive hundreds of millions of miles?

Mm-hmm.

It turns out that the basic, vanilla end-to-end system is insufficient, right? There’s a massive difference between using end-to-end and purely relying on it. At Waymo, we’ve really gone beyond that basic, vanilla end-to-end approach, and we’ve augmented the learned representation with a structured, materialized intermediate representation.

What that allows us to do are a few very important things that you might not actually need if you’re building a different product—if you’re building a driver-assist system, a prototype, a demo, or a small-scale deployment. But again, those things are absolutely critical if you want to go all the way to a fully autonomous, safe system with superhuman performance.

Those include having extra validation at runtime of the agent that’s running on the car in the physical world. They include richer training and evaluation recipes that are very difficult or impractical to do in a pure, basic end-to-end system. This structured, materialized representation gives you a boost in things like closed-loop evaluation, closed-loop training, and rich reward functions for reinforcement learning. So that’s been our approach.

And all the human feedback that you get from support and drivers dropping in and all of that—it’s essential to have this type of architecture to do that.

Exactly. Exactly.

Makes perfect sense. So not only have you innovated on the software stack, but also the hardware stack. There’s a 6th generation now of the Waymo Driver, and you guys have always focused on being the driver. Tell us about the new 6th generation, and what was it like the first time you interfaced with it?

Yeah, the 6th generation is our most advanced hardware and sensor suite yet. The focus has been on performance, but also on simplification, drastic cost reduction, and high-volume production at scale. This is the driver that's powering our latest vehicle platform. That's the OHI.

Earlier this year, we started fully autonomous operations. It’s currently only open to employees, but it’s coming to all of our riders later this year. I had a chance to take a ride in one literally as soon as we started running fully autonomous operations.

I’ve spent a lot of my life in various generations of our cars. Every once in a while, there’s a new first moment, and that was definitely it. The whole car is designed around the rider experience. Even though the external footprint of the car is about the same as the IP, when you get inside, it feels like a living room. There’s so much space in the back.

We have new screens, and we have these doors that slide open and will open automatically when you approach the car. I had a blast, and I can’t wait to have this car in our fleet, open to everyone.

So you guys are going through a period of incredible scaling. For many years, you were purely in the lab and in R&D. It took roughly 16 years to get to 100 million miles and roughly 6 months to get to 200 million. Things continue to scale—

That’s it.

—really rapidly. 11 cities now, with many, many more on the horizon. Tell us, what is it like to scale a new city? And then tell us about your daily life with a Waymo. How do you use it as a creator?

There’s a lot. So, exponential scaling. First of all, absolutely—it’s been a phase transition for us in how we’re scaling. To give you a couple of additional data points, it took us 8 years from the day when we started our fully autonomous operations to the day when we had our service, our driver, providing rides to the public in 4 cities.

Earlier this year, just a few weeks ago, we launched 4 cities in 1 day. We’ve given over 20 million fully autonomous rides. 10 million of those happened in the last 7 months.

Amazing.

So that’s what exponential scaling looks like.

Amazing.

Launching new cities involves operational components. You have to collect the data, characterize the environment, and validate the driver. A significant part of it is starting the conversation with the local communities, because it’s a new thing and a new product. It’s on us to earn the trust of the people there.

More often than not today, we’re seeing that the driver is generalizing incredibly well, and it’s just a matter of high-fidelity, rigorous evaluation and validation before we deploy the fully autonomous product. Then we go from there.

Your daily life.

My daily life, yeah.

Yeah, as the creator—

It was a multipart question, yeah.

Of course, you are.

Waymo is how I get around nowadays. That’s how I got here today. It was a great ride from Palo Alto up to San Francisco on the freeways. My family uses it. I have 3 kids, and they love Waymo.

I think nowadays they get annoyed if, on a rare occasion, we have to be in a car driven by myself, my wife, or another human being. They’re like, “Okay, what’s going on here?”

I feel the same way at this point.

Yeah, they love it. It’s been part of their lives for the entirety of their lives. When we’re driving around, there are only 2 things that get call-outs from my kids nowadays: doggies and Waymos.

Nice. Probably similar amounts of cognition between those 2.

Okay, let’s talk about safety. One of the most meaningful, exciting parts of partnering with Waymo has been the fact that 1.19 million people a year die in road accidents worldwide. This is life or death. Not only does it touch everyone in this room, but everybody has some connection to someone who’s been impacted by this.

You have been focused on safety from the very beginning, and it’s actually pretty hard. In a Silicon Valley where it’s “move fast and break things and see what happens,” you guys have been incredibly patient with safety. Can you tell us about a story that made it very real to you and how you keep that safety culture at Waymo?

The numbers you mentioned are what drives all of us at Waymo, and the status quo is not okay. We’ve grown this over time, but challenging the status quo is really important to everyone at our company.

You’re absolutely right that how you go about building a system like this is different from what you might do in other areas, fields, and industries, where safety has to be the nonnegotiable foundation. You have to build that into everything you do from day 1: your model architecture, your training and evaluation recipes, and the mindset of the team.

It can be very tempting to focus on capability first and get to 90% very quickly. But how you go about the first 90% is a totally different problem from how you go about getting to your next nines. Keeping that in mind and focusing on safety as the nonnegotiable fundamental layer from day 1 is super important.

Today, we’re driving more than 4 million miles in full autonomy per week. You see a lot of events from the field, and today we have data from more than 170 million fully autonomous miles, where we see that the Waymo Driver is more than 13 times safer than a human driver when it comes to serious-injury-causing collisions in the cities where we operate. You see that sort of superhuman safety behavior manifesting itself on the roads daily.

Mm.

Right? I saw an example recently of a person—I think it was a young woman—on an electric scooter on a road. She lost control, tripped, and fell right in front of the Waymo. The Waymo Driver showed superhuman accuracy and reaction time and was able to swerve and brake, and everybody walked away.

Things like this are very rewarding to me personally, and to the whole team, in terms of actually having a real impact on the safety of our roads. At the scale we’re operating, that 13× reduction means that we’re preventing a serious injury every 8 days. That impact will just grow as we scale up.

Wow. We’re going to open the room to audience questions in just a moment. But before we jump in, I heard a story about lidar detecting—or radar detecting—the footsteps of somebody behind a bus. Did that happen, and how does that work?

Yeah, this was one of those moments where I was positively surprised by the emerging capability of our system.

The situation was in San Francisco, I think. The Waymo Driver was at an intersection. There was a bus that crossed, and we were sitting there waiting at a red light. The bus crossed and stopped, partially blocking the intersection. Then our light turned green, the Waymo Driver started to proceed, and as it was proceeding, it detected a pedestrian on the other side of the bus.

You can’t see through the bus—not with lidar, radar, or cameras. The windows are reflecting the people inside the bus. The Waymo Driver started to react defensively, and sure enough, a pedestrian emerged from behind the bus. We were able to nudge around them, and everybody went on their way.

When I saw that, it blew my mind. I wasn’t sure what was going on. I guess, as capable and superhuman as the Waymo Driver is, it doesn’t see through solid objects. What turned out to be happening was that our lidar was bouncing the signal under the bus and got a little bit of a sparse return from the movement of the person’s feet under the bus. That was enough for the Waymo AI to not only detect that there was a pedestrian there, but also make a prediction about what was going to happen in the future and keep everyone safe.

Mind-blowing. Pretty unbelievable. We’ve got time for 1 question from the group. Jim, sorry, no free codes—not at this one. Yes, Jim, please.

Thank you. Does this work?

Yes.

I was just saying congratulations on all you’ve achieved. It’s really mind-blowing. If you think about the next 5 to 10 years, really focusing on the business model, what are the milestones? What happens in major cities? What’s going to be different from where we are today? Just walk us through your vision of the future.

We’re heads-down in execution mode. We’ve transitioned from intentional, sequential de-risking of the Driver and key parts of the business to rapid, parallel global commercialization.

That means deploying the Waymo Driver in more places across the United States. Today, we’re in 11 cities, operating fully autonomously and serving our riders. We’re going to expand in those existing places, and we’re going to add new geographies and new cities.

We’re also expanding internationally. We’ve announced that this year we plan to offer a service in London and Tokyo. You will see us accelerating that deployment, all in service of our mission.

Good news to our team in London.

Yeah.

We covered a lot, Dmitri, from the very early days, when you could get a lot of distance with not a lot of technology, to persisting through extremely hard times in autonomous vehicles and getting that extra mile. We talked about world models, driver-simulator-critic architecture, the hardware, the 6th-generation hardware, safety, and scaling.

Most of all, I hope that we learned a little bit more about Dmitri, the man who’s brought the magic that is Waymo to so many of us. As I’ve gotten to know you more and more, I’m constantly struck not only by your brilliance, persistence, and performance, but also by your humility. It says a lot about accomplishing this much. Thank you, Dmitri. Please join me in thanking Dmitri for all he does.

Thank you. Thank you. Thank you.

And the many lives saved.

Thank you.