- Hop in a driverless taxi with the Google-owned self-driving company Waymo opening up its robo-taxis to all users in that city. - The technology, of course, is hugely expensive. - Alphabet's autonomous vehicle unit, Waymo, completing another major round of funding. - Tahedra has served as co-CEO at Waymo since 2021.

- We are on a mission to build the world's most experienced driver and make it safe and easy for people and things to get where they're going. - This is not a move fast and break things. This is a move with great, great iterative pace and focus on safety.

- Please join me in welcoming Tahedra from Waymo to the stage. - Hi. - Thank you for coming. - Hi, thank you. - It's a pleasure. - Thanks for coming. - Thank you so much. - Hi. - How are you? Good to see you. - Hi, David. Thank you.

- Tahedra, thanks so much for being here. - Yeah, thank you. - Tahedra's been co-CEO of Waymo since 2021, where she leads the company's business operations and expansion in autonomous driving. Recently, Waymo has been reported to be completing-- I don't know if Waymo put this out or if it was in the news, but over 100,000 paid trips per week now.

Is that public? Is that like a true thing? - It's completely true. And it's over 100,000. - Over 100,000. - Per week, yeah. - And we can now see Waymo's autonomous ride-hailing services in Phoenix, LA, San Francisco in the Bay Area, Austin. And I think on X, you announced $5 billion in additional committed funding from Alphabet-- I don't know if that was recent-- to support Waymo's expansion.

- It was recent, yes. - Congratulations. - Thank you. - So thanks for joining us. The Waymo story started a long time ago as a Google X project. When did you join, and how did you kind of become co-CEO? - Yeah, thank you. So Waymo started as the Google Self-Driving Car Project back in 2009.

And actually, my co-CEO was one of those founding team members, Dmitry Dolgov. And in 2016, at the end of 2016, we spun out and sort of became Waymo. And that's when I joined. At that point, it had been determined that the technology was on a good enough path to start to think about commercialization and scale.

And so the fun thing about joining a moonshot is I came from a big tech company. So I'm like, oh, product market fit. Like, it's all there. I'm going to join, and we're going to scale. And the reality is it was an opportunity to learn how much this is a process of discovery, how much this is a process of figuring out not only the technology roadmap, but also the external environment.

And at that time, I had been head of policy at a lot of companies. And I joined under the idea that if the technology works, the market opportunity would come from being able to open markets. And so that was super terrifying to me, because if you're at most large tech companies doing policy work, you're actually trying to avoid costs, or you're trying to defeat legislation that's harmful.

But this was really the opportunity of, like, how do you usher in this amazing technology? And so that's when I joined. And then I was asked by my prior CEO to help launch the first market, which was Phoenix. Back then, it was Chandler. And we launched that back in 2017.

And we've been-- since 2018, a lot of people don't realize, we've had a 24/7 ride-hailing service around the clock. It was in 2020, during COVID, though, that we removed the driver from behind the wheel. So we have the Waymo driver that's driving the vehicles at all time. But there's no human behind the wheel.

And we've been offering that service now. We're almost at four years, 24/7, not only in Phoenix, which is a 315-square-mile territory, but also in San Francisco and now here in LA. So in that sort of ushering in from market one to this role, I moved through the COO role.

And that's when we actually removed the driver. And we're the first company that's done that. And so it was daunting. And I think that's one thing I love about Waymo. We call our team Waymonauts, is you just have to, every time, get to this point of determining what's good enough, what's safe enough, and then make that leap.

And then learn, because you can't learn until you get to that next point. And so, of course, we have rigorous safety culture. But it was in that time that I was COO. And then in April of '21, I moved into the co-CEO role with Dmitry. So now that it's scaling, 100,000 paid rides a week, maybe we can talk a little bit.

I mean, Jason's spent a lot of time on this. Well, I think maybe starting with security would be an interesting-- safety, rather, would be a really good idea. You've done, it seems like, millions of rides now. No fatalities, correct? That's right. But there have been accidents. So when you look at the accidents that have occurred, how often does an accident occur?

And in what percentage of the time is it just somebody running into the Waymo versus the Waymo running into somebody else? Yeah. So we've done 2 million paid rides. So when we think about-- we just released this week our safety hub, so you're sort of tapping into an issue that's really important.

What we've been able to determine as of June of this year, with over 22 million, what we refer to as RO miles-- those are miles that are fully autonomous-- we've been able to have 83% fewer airbag deployments, which of course is important because those are high impact, and 73% fewer injury-causing crashes.

Now, that's what we're focused on. Over a human. Over a traditional human. Meaning the counterfactual in that area at that same point in time-- Exactly. --via insurance claims and other data. Exactly. We're comparing in San Francisco, not on highways. There's an apples-to-apples comparison. We're comparing in San Francisco or in Phoenix.

OK, great. Because this is the next piece, which is, you're not yet on highways. That's right. We're testing on highways in Phoenix and San Francisco. With safety drivers. With employees and a human behind the wheel. Right. At times. And maybe you could explain to us why you're taking that approach.

Because in my experience driving a Tesla since the beginning with full self-driving and autopilot before that, it seems like it's better on the highway than it is on local streets, where you have people jumping out and weird things that occur on the highway. You very rarely have to run into a weird thing.

It's pretty ABC. So I would think it was the opposite. But is it because speed kills? What's the thinking there? So I think, just for everyone to orient, when you think about the levels of autonomy, we are only focused on level four and above. Level four and above means you do not need a driver's license, nor do you need to sit behind the wheel.

So when you're focused on tackling the hardest challenge first, it's really important for the driver, the Waymo driver, to ingest dense urban environments. Because that is where the driver is going to learn the most. Sitting on freeways, seeing the same thing every day can be predictable. It's not entirely predictable, but it's not additive to what the machine learns.

That's a great answer. So we're trying to learn as much as we can learn. And we've always, from the beginning, been focused on freeways. It's just that freeways isn't actually how the machine learns the fastest in the most sophisticated environments. But because you brought up sort of a level three or two or one system, I think it's also important that for systems where you need someone to sit behind the wheel, you can have a completely different approach to this technology.

But we're really-- I don't know how many of you here have had the chance to take a Waymo ride here in LA. It actually includes this campus where we're sitting right now. But you can't sit behind the wheel. So there's no ambiguity. You sit in the passenger seat, or you sit in the back seat.

So let's talk about the LiDAR versus camera bet. Elon's been pretty clear, hey, you don't need LiDAR. It's a waste of money. The LiDAR on Waymos, I believe, is still in the $10,000 to $20,000 incremental cost. Am I ballpark correct? I'm not sure. Oh, OK. So with that-- As you scale, costs go down.

Yeah, of course. So who's right? I mean, are you using the LiDAR, and is that a critical piece of this? Or do you think eventually, hey, the cameras are seeing in 360 degrees. They have higher fidelity than human eyes. And if humans do a great job, and you just have to beat the humans at driving, why even have the LiDAR?

What do you think? So obviously, we have conviction that we need LiDAR, radar, and cameras. We have all three. It's given us the opportunity to scale at this point. We're the only company that's doing level 4 driving on public roads 24 hours a day. It's very different to get to the place of a demo.

We learned that ourselves back in 2015, and '16, and '17. We had an autopilot highway product. We allowed employees, at the time Google employees, to ride in that product. We told them to pay attention. We told them, you have to be alert and be prepared to take over. What did they do?

We're human. We want to do other stuff when we're in the car. Like, I want to be on the phone. I want to get to that meeting. I need to send that last email. I need to check whether or not my son's doctor's appointment is tomorrow or next Thursday.

We're distracted. And so what happens, the thing's beeping at you. It's telling you to re-engage and pay attention. And what we learned is that that's not a way to improve road safety. And so the question is, what's the end goal? Is it to improve road safety or not? For us as a company, we're obsessed.

Our culture is about safety. This mission is about safety. And we want to expand the number of people who have access to mobility options. If you need a driver's license, then it's just like every other mobility platform. You have to be behind the wheel. Sorry, go ahead. I'm curious about when this flips from a technology problem and a go-to-market problem to one of public health.

Because that's a staggering fact. And we all know people that have lost somebody close to them, whether it's just a road accident or to drunk driving or to-- I don't know if you guys saw "Anybody Follows Hockey," but Johnny Hockey who was killed with his brother the day before his sister's wedding by a drunk drive.

This is so avoidable. So what does it take for a politician or a group of politicians-- and I don't even know at which level this decision would get made-- where they say, OK, we've seen enough. That should be the only solution. And are you pushing for that? So we are pushing to make sure people who put autonomous vehicles on the road have to demonstrate their safety case.

We think the worst thing that could happen is introducing a new technology that doesn't actually improve this problem. Because it'll kill the whole industry. Because it'll kill the whole industry. And the state of affairs, 40,000 Americans, 1.35 globally, dying annually from road crashes, is avoidable. And so the challenge that we have to tackle with a lot of humility is how prepared is the public-- humans can kill other people.

How prepared is the public to accept that this isn't going to be a panacea, and it's not going to be perfect? You didn't hear me say 100%, 100%, 100%, right? It was 70. There'll be a small error rate. There's going to be a small-- and so that's the work that we're doing, which is we have community partners, Mothers Against Drunk Driving, National Safety Council, a host of organizations who've been trying to tackle this issue, who we partnered with early because it's actually in both of our interests to figure out how to tell the story that if this technology can reduce 60% of the fatalities, then it's worth the exposure to the risks.

Are you on any sort of internal shock clock, meaning there must have been a moment where you convened a meeting, or you guys had a meeting, and you said, all right, we have enough data. Let's go. We're going to pull the guy out of the car. Or, hey, this is it.

We're ready. And then you show it to regulators, and they say, OK. Is there a set of these really important milestones that we're going to feel like goes to other cities? Is it only grid cities where this works? So where do you go from here? Yeah, I hope that what, over time, you start to experience is, one, because we are super focused on safety, you'll see us scaling.

And because we're building a driver and not a car, you'll see us scaling in a number of ways. So we have partnerships. Right now, we have a partnership with Uber in Phoenix. Why are we doing that? People are like, oh, my gosh. I got my Uber Eats in a Waymo.

Why are you all doing that? It's because we're building a driver. And so we're trying to figure out all of the ways in which this driver could be applied to future use cases. And so over time, you're certainly going to see us announce new places, but you're also going to see, like in San Francisco, we launched a 24-hour service in a limited territory.

And over time, we've expanded that territory. Is it a step function in complexity to go to cities that are more-- how would I call it-- more chaotically city planned? It is definitely an opportunity for the driver to learn more. And so coming to Los Angeles, for example, was more like being in Phoenix.

Going to Austin-- Grids. Exactly. Exactly. Going to Manhattan, we will certainly benefit from our work. And San Francisco, same thing with DC. So start with the grid cities and work your way into more complex cities. So can I follow up on that last point? I have a friend who's the third or fourth investor in Uber.

And he's very worried that once this product scales, there's not really going to be a need for Uber anymore. Why would you need Uber if you can just summon the robot car directly, or the robot driver, whatever you want to call it, right? Driver, car, whatever. I mean, doesn't this logically disrupt Uber once it's at scale?

I think-- Well, there are current partners. As I said, we have a partnership with-- right, I mean, we have a partnership. Your friend hasn't sold enough shares. Your friend's doing just fine. We have the partnership with Uber because we think it's important for us to create an AV ecosystem around this technology.

This isn't going to be we've created technology. We're going to figure out everything ourselves. That would be foolish and very expensive. And we'd never find ourselves profitable. But having companies that are experts in integration, having automotive partner-- we're not building the car-- having network partners, having fleet operating partners, public officials.

There's a lot that is built around us. And these are the early days, the early innings. And so that's part of it is learning. I'd say five years ago, people would say to me-- network operators would say, oh, this is exactly-- we know how to do this. Well, this is a constrained supply.

It's not a supply you just surge. This is a driver that you need to understand actually how to operate with. And so I think we're going through those steps to learn together. And so what I would say is there's room for a lot of people-- is the insurance lobby and the insurance industry a proponent?

Or would they try to constrain your progress? So one of the largest reinsurers in the world, Swiss Re, has actually been doing our safety-- Yeah, they would want to underwrite this. But I'm talking about more like the GEICOs of the world. Same. I think they're very eager to get the data sets that we're releasing.

We released 20 safety papers last year. They're all wanting to understand what the actual impact of this technology can be. Because one of the things I think that's hard for us to grasp as humans-- or at least I'll say for me. I don't like speaking for everybody. We drive in a week more than a human drives in a lifetime.

And you can say that as a company that runs fleets. But we're the only company where it's one driver. And that's not the way the world is organized right now. FedEx has a lot of drivers. All of you who have Teslas have a lot of drivers. Everyone has a lot of drivers.

And all of those drivers have different capabilities. This is one driver learning all the time in unison. And so what can we demonstrate that's possible? And it really goes back to your other question of what are policymakers five years from now really going to want to know? And how will they be able to derive from what we're doing today?

Where were you when you heard the news of Cruz dragging that woman in San Francisco 20 feet? And then where were you? And what was your reaction internally to when you found out Cruz covered it up? I don't remember where I was when I learned. But we're here to improve road safety.

So it was pretty upsetting, I think, to me and the whole org. I think as we learned about the lack of transparency, our view has just been, be transparent. It's not perfection. It's technology. Yeah, but you're considered their contemporary. And you have a different approach. You would never cover it up.

They obviously had something dysfunctional. Did that set the industry back significantly? How are regulators talking to you before and after that, differently, if at all? It definitely led to more scrutiny in San Francisco, because there was already an active, and continues to be, a fairly active voice against technology in San Francisco.

And so we needed to continue to engage, which we were already doing, provide more data, which we're, of course, willing to do. And I think over time, what we'll see is we have to get to a place where the amount of data that we're providing actually advances the safety concerns.

I think at that point, it was just providing a lot more engagement. If the mission is to reduce those deaths and to really scale this technology, Uber's doing a quarter billion rides a week. You just raised $5 billion. It's a big number. That gives you another 50,000 cars, which is a fraction of the rides in but one city.

It seems like you need a lot of partners in the car space to build this. So wouldn't the best strategy be to license this technology to every carmaker and then allow the Ubers, the Lyfts, the DoorDashes, and fleet managers to handle cleaning the puke in the back of the cars and putting the air in the tires?

Because I don't think Waymo wants to be in the back of that car cleaning up the puke and doing the air tire, right? There has to be somebody who manages all that. So maybe talk a little bit about that possibility of-- and even a more beautiful thing would be for you to open source it.

So has there been a discussion internally about open sourcing this technology? So I'll start at the beginning. I think what you just described is exactly the AV ecosystem that I'm talking about. And yes, out of necessity. And because we are-- I mean, whoever was first was going to need to create this.

It doesn't exist. So we're doing that. I think the second part of what you're talking about, let everyone do what they're good at. We're good at building this driver. There's, though, a period where we have to become really good at what does it mean to operate at scale. And that's where we are right now.

What does it mean to operate at scale? Technology can work and not be delightful, not be reusable, not be repeatable. And we don't just want people having tourism rides. We want people actually integrating this into their lives. And we just did a UX study. And we found that a third of people use the service in San Francisco to go to their doctor appointments.

We're like, that feels like real life. And 36% were using the vehicles to go to local businesses, which is also real life. And so these are the kinds of signals that we're trying to get. As far as personally owned cars, we've always said as part of our strategy from the beginning that we would start with ride hailing and local delivery and that eventually we would make our technology available to automotive companies so that at some point, you'll find Waymo cars sitting on showroom floors for sale.

Oh, wow. So we could go buy one and then put it into whatever fleet we want. Yes. What does it cost to make one of these cars today? And where do you see it in five years? And just, I guess, on that question, where's the tipping point in unit economics?

So if you look on the ride hailing business, based on-- I know a lot of people drive during rush hour. And then they don't need rides the rest of the time. So there's a lot of excess inventory sitting around. So if you look at the utilization of a car over its lifetime and the cost of the car over the lifetime, what's the breaking point where the unit economics make sense?

Yeah. So your question of cost of car, we just don't break it out since we're not reported separately for Alphabet. We're just bundled up. The rumor is $1.50 right now. It's close. OK. $1.25? Not commenting on rumors. Not commenting on rumors. OK. OK. Wait, if that's just the system for self-driving.

The car plus the system. The car plus. But what if-- $125 to $150 is the rumored cost for the car today. It's actually like-- You're going to comment on the rumor? I love this. I mean, that's obviously a lot more than a Tesla, right? So-- Yeah, Tesla's $40,000, $50,000, yeah.

So do we know ballparks here? What can we say? So keep going. It's so fun. We're poker players, so we're going to read you and bait you. And it's not working. We didn't know you were a Jedi. So sorry, let me just ask. Is it about reinventing the system or getting the next gen of the system before the unit economics break?

Or is it about volume? What's the breaking point of where the unit economics become profitable, become positive? Yeah, not going to get into that other than to say to you, it is the thing I spend all of my time focused on. And we are on-- But there's a path there.

There's completely a path there. Well, without getting into costs, so is one of the plans here to sell this to all the car makers so that they can compete with Tesla? Because most of the car companies do not have the technical capabilities to develop their own self-driving. It's just impossible.

So are you guys going to be like an OEM to all the car makers so that they can provide self-driving? Yeah, we've said sort of publicly, and we've had partnerships where we've made the technology available. It's a question of when will a level four sort of product be available and interesting to consumers at scale?

That's part of the question. But absolutely, this is part of our strategy. A friend of mine-- Can I get that? Yeah, of course. I'm sorry. A friend of mine who couldn't make it here is an angel investor in Uber. And he-- no. If you look at the business outside of the United States-- actually, if you look at Uber's business, there's a lot of opportunity in other markets.

And sometimes, in other markets, they will prove probably more able and open-minded to taking the public health priority around these things. So why such a heavy reliance on America, considering the scale and the imprimatur of Google, could probably pull you into any number of markets anywhere around the world?

Yeah. So we really think about it as this sort of learning journey. But it's also just where we started, right? I mean, part of the conversation we get to have now is what would-- if the driver was at this level of maturity, where would we have started? That said, we have global aspirations.

We're always thinking about the-- I mean, doesn't it just take one geography to prove this point? Or do you view it the problem that way? Like, it's like, we're training, we're training, we're training. And then at some point, there'll be some city or county-- Which point? I just want to make sure I understand.

--this idea where you can prove that it's just so reliable and valuable. And that you're saving just-- I really care about that idea. Yeah. To who? Who would it prove the point to? People that would make the rules, I guess. Yeah. I think-- I hope. I don't know. I think there's a lot of noise in the system.

There's a lot of confusion about full autonomy versus partial autonomy. There's not a lot of legislative activity that's informed by experts. So I don't know, right? Like, certainly, it's my hope. And I think one of the things, when you think about certain countries around the world, there is more of a nation-state, top-down approach to some regulations.

And so that could be easier. Singapore? Yeah. You prove it, it's codified into law, you launch. That hasn't been the primary way we've thought about market entry or opportunity, but it's certainly one of the areas that we focus on. But as long as some markets adopt, the rest of the markets will see and they'll fall behind.

I mean, you just have to penetrate some markets. Yeah. Google had amazing success with Android and open source. So back to that open source question, you must have had this discussion many times internally. Why not open source the technology and then provide the premium Android to people who want to pay?

And is that something on the roadmap? Because I hear there's a big discussion internally about that. To OEMs, to car OEMs? Just open sourcing the maps, the software, to make everybody contribute to it, to just get this out quicker. So where are you at with those internal discussions at Waymo?

Give away the whole business? Yeah, exactly. No, well, I mean, Android, it's worked, right? It's like $20 billion in. It's like, let's just give it away. But Friedberg, if you look at Android, it's only gotten stronger, right? It's completely different. I think that-- That was defensive on search, but-- It's OK, bad note, your answer.

No, I like it when you guys answer each other's questions, because I get to learn what you actually think. We want to see your answer. We fight a lot, it's fine. No, it's fun. It's actually fun for me. So feel free to keep-- So on the question of open sourcing, that isn't something that we're spending a lot of time talking about right now.

Let's talk about level six, which I believe is inclement weather, and really-- Level five. Oh, level five, rather, sorry. And I know when I'm in my Tesla, it gives me a little warning. Hey, inclement weather, and the FSD might be degraded, extra attention time. How close are you to feeling comfortable with these things driving in snow, ice, rain?

Or when that happens, do fleets just pull over and game over? You can't use this technology. Yeah, we've been improving weather. We've tested across 25 cities specifically for that. So we had massive storms in San Francisco last year, and we didn't take down the fleet. Our fleet was able to handle all of the rain.

So rain and flooding, pretty good. Rain and flooding, pretty good. Because of LIDAR and radar, maybe. You said it. OK, fog, snow, ice. So we've tested in snow and ice. We don't have any deployments there, but that's an area that we're learning. Fog, we've been able to master in San Francisco.

So that 280 craziness. No problem. Same thing with those sandstorms in Phoenix. That was an early learning. Those are actually very challenging. And then there are swarms of birds in Phoenix. Like, really. And heat, right? You have a machine-- Phoenix sounds brutal. It's like Revelations. There's a lot. There's a lot.

And we're the only company that can pick you up autonomously at the airport, right? And so like, curbside hiccup is the whole thing. What has been the hardest challenge for you as CEO? I think-- I mean, we've heard all these rumors about cultural issues at Google or whatnot. But I get the sense that you're much more separated in terms of building a much more technical and specific culture.

But what have been the hardest challenges for you? I think one of the hardest challenges was making-- we made such a big decision in going fully autonomous, which is sort of the reason we're here, during COVID. You mean getting that person out? The person from behind the wheel, scaling this, and so then bringing the culture back together to now scale.

So we just needed-- like, we needed a minute to lock back in. That's one of the things. And the other one, I would say, is what you just said, which is it's a brilliant team. And the team works really hard. But like, bad things are going to happen in the real world with human drivers around our cars.

And the team takes that so seriously. And so that's-- it's like having worked at-- I mean, I worked at AOL, and you were there, too. But like, other kinds of tech companies-- I know every kind of company is stressful. It's a different kind of stress when people's lives are at stake.

They internalize it. Completely. And they personalize it. It's their work. And so that's-- it's just a different level of stress and accountability and a sense of responsibility. But do you have to manage people differently because of that pressure, or no? We do when things happen in industry. We need people to sort of stay buckled in and believing that this is something that is worth it.

It's audacious. It's hard. But it's worth it. And it's bigger. Most people come because they have lost someone. Like, they're very mission-oriented. I'm going to be very honest with you. It's an incredible framing you gave. And I totally missed it for so many years, which is when you said, we're building one driver.

I thought for whatever reason you were building an autonomous driving system, which I interpreted very technologically for a very long time. But you're absolutely right. You have one goal. It's very clarifying, actually. The world's greatest driver. One-- the world's greatest driver. The world's most trusted driver. The world's most trusted driver.

Takedra, thank you for-- That's an incredible thing. Yeah, I mean, it was incredible. Thank you so much. Thank you. That was awesome. Thank you so much.