Uber 的 Robotaxi 路线图、人类驾驶终结与100亿美元机器人押注｜Dara Khosrowshahi（Uber CEO）

I'm gonna need to get my driver's license in a couple of years, but after hearing this, I have a question.

Well, you may not have to get your driver's license, but it's okay.

That was my entire question.

So much of driving now is the technology in the car and around the car. Technology's always going to be a part of the human experience. Humans will be demonstrably less safe than autonomous drivers, and then it's going to be up to regulators to decide what a human license looks like.

Mass production of these AVs is gonna take some time. They are very expensive today, much more expensive than a regular car. It's just happening faster than even I expected, and I'm an optimist as it relates to technology. Is society gonna adjust, and can it adjust fast enough?

1. Autonomy Redefines Human Driving

All right, let's go to the mics for your questions. I'm gonna start with this question on Slido: what's harder—convincing regulators that autonomous cars are safe or convincing passengers to get into one?

It's actually pretty easy to get passengers to get into an autonomous vehicle. What we're finding is, for example, in Atlanta or Austin, we let you know that you've been matched with an autonomous vehicle. Eighty percent of people say yes, and 20% say no, and the 80% who say yes love the experience. Regulators obviously move a little bit slower, but this is a technology that I think more people are excited about than not.

All right, we'll go to one of our teens. We have a group of 20 teens here in the room.

Very cool.

Please.

How you doing?

I'm doing good. I had a quick question about, in 15 or 20 years, when most of the cars on the road are self-driving: does it ever become illegal for human drivers because they become a safety hazard on the road? And how does that end up affecting the next generation of drivers? Do they have to learn how to drive? And what happens to sports like race-car driving? Is that gonna die out because nobody has to drive anymore? That's kind of what I'm asking about.

Well, I think race-car driving's always gonna be a thing because it's just so damn cool. Again, you may be augmented, right? You think about how much of driving now is the technology in the car and around the car, so technology's always going to be a part of the human experience.

I don't think it's gonna happen in the next 15 years. But certainly in the next 25 years, humans will be demonstrably less safe than autonomous drivers. Then it's gonna be up to regulators to decide what a human license looks like. Maybe the driving test is gonna be much more demanding than it is today. I like free will, so I'm like, "Hey, if you want to drive, that's not a problem." But I can definitely see a future where there are gonna be fewer and fewer drivers on the road, just like there are fewer people who know how to ride a horse.

Thank you for that question. I'll build on that. There's a lot of deaths in the Middle East from young teenagers with fast cars.

Yeah.

I could imagine a future in which an AI overlay allows you to drive your car as fast and crazy as you want, but it knows the car's limits and stops you from ever swerving off the road. It's like bumpers that you go up against, so that would be fascinating.

Yeah.

But that's just a comment. Let's go to the next question. Mic 1.

2. Flying Cars Need Vertiports

Just a question regarding flying cars. From a real estate perspective, what will the opportunity be? Will you be putting them on top of buildings or building heliports?

We're talking—again, this is more Joby—but there's a lot of discussion on vertiports, so to speak. We want this product to be a mass-market product. These vertiports will have to be designed to have multiple vehicles coming, landing, and taking off. The location of those vehicles is gonna be based on kind of very high traffic areas. I think, obviously, airports, city centers, or the sides of cities are going to be targets, so to speak.

We've got perfect data in these cities as to what the best places are for you to place vertiports, to be able to have the most business and reduce the most amount of traffic as well. So there's absolutely a real estate opportunity. If you're interested, give our Uber folks or the Joby folks a call.

Yeah. I think there's a huge opportunity as well for small islands that are hard to get to.

Totally.

Elshin, go to you, mic 2.

3. Supply Makes Marketplaces Work

Thank you very much. I'm building a solid-shape construction-materials marketplace. If you were starting a marketplace now, in the abundance era with AI, what would you focus on first? We're using Uber for our LTL. Thank you.

So, if we start a marketplace, what would you focus on first? Is that the question? Our marketplace has always been supply-led. One of the differences when I first came to Uber from Expedia is that Expedia was very focused on building out audience. Uber has always been focused on building out the liquidity, whether it's more cars in a market, more restaurants, more retailers, et cetera.

If you build liquid supply, if you have product-market fit to some extent, the demand just shows up. Obviously, there's always a challenge in marketplaces: how do you build both sides? But to me, put real tools in place to become the easiest to use and the most liquid as it relates to supply. Then, if you do have product-market fit, the demand will show up.

Great point. Thank you.

Let's go to Burgundy on mic 1. Burgundy.

4. Autonomous Fleets Need Trust

Hi, my name is Burgundy, and I am definitely biased toward automation. But as a lawyer outside of this room, everybody is just freaking out. So my question is:

Given that we have this automation and sensors, and having seen the data on human error from speeding, intoxication, et cetera, what are some opportunities for insurance to cover this through, for example, product liability, so we can take the human out of the equation and yet still give people an assurance that they're gonna be covered if accidents do happen?

Yeah, I think it's a great question. The first thing I would say is that with Uber, the professional driver is safer than the human driver, and the data bears that out. But I do think that you're gonna have different layers of insurance as well. The autonomous provider will have the specific insurance related to the car, and they'll have to take responsibility for the quality of their driver.

Then we'll also have another insurance layer on everything else that happens, so that people are covered and can feel comfortable taking an AV to wherever they're going.

By the way, there's a hidden statistic that very few people know about: about half of the court cases in the U.S. are car-accident-related.

Wow.

It's a ridiculous number.

It's a very, very big industry.

What a benefit to society it would be to get rid of that. All right, go to Shilpi, is it?

Shilpi.

Yeah, it's a pleasure.

Hi. I'm Shilpi. I led robotic systems at Amazon Air hubs. I launched and led them, and now we're building continuous biosensing technology at Quinn Labs.

My question to you is: as you start moving toward multi-robot supersystems, with Joby aircraft working in coordination with ground vehicles—autonomous ground vehicles—I think the hardest part is the coordination and the continuous monitoring and upkeep of these multi-robot-type systems. We faced a version of this at Amazon as we were managing these large robot fleets.

I'm very, very interested in your vision of this infrastructure. What do you believe the bottleneck is: hardware, software, or regulation?

You've described exactly what we do. We have a very, very large marketplace team. We've got over 9 million drivers, probably 10 million drivers now globally, and in every city in which we're operating, we're constantly taking the state of that city in real time, or as real time as we can be, because we've got to recalculate the state of the city.

Every time you hail an Uber, we're not just dispatching you the closest car; we're also predicting what's gonna happen over the next 5 or 6 seconds and connecting you with the right driver at the right time in a way that's most efficient for the overall network. There may be another driver who's closer to you, but we think that Peter's gonna hail that driver, and we'll leave that driver free and hail another driver to you. Our systems have to deal with human beings who are highly unpredictable and highly erratic as well, in terms of their behaviors.

So sometimes if a dispatch isn't going well, we'll do another dispatch, et cetera. So I'm actually psyched for machines because they're much more predictable, and they usually say what they're going to do. One of the contracts we have with our human drivers is that when we dispatch a ride to them, they can accept that ride or not accept that ride.

With machines, again, they're not only going to be predictable, but they'll have acceptance rates that we think are much higher than humans as well. So, of course, there will be a coordination problem. How do we stitch together one ride with another ride and make sure that that experience is perfect? But this is exactly what we're doing at scale: 40 million trips every single day, with markets that are very, very different from each other in terms of behavior. I'm very much looking forward to that future that you described.

Nice. Let's go to Mark on mic 2 here.

5. Affordable Living Stabilizes Society

Thank you, Peter. Dara, my moonshot is what I call affordable living as a service: housing, food, energy, connectivity, and mobility for $250 a month. It won the XPRIZE Visioneering Contest in October, and so it's going to be a $50 million prize to deliver this.

Cool.

$250 a month? Wow.

$250 a month. And I think people in this room know it's possible, but delivering it sooner rather than later, as a stability lever in this moment of transition, is going to prevent us from having chaos. So I want to challenge you to be a funder and a competitor in this prize.

Well, thank you. One thing that's very important to us is making sure that as we build out Uber in cities, our service isn't just available in the middle of cities where wealthy people live, but is available all throughout the cities, in the outer boroughs, et cetera. So, happy to talk in terms of the provision of transportation to a lot of these transportation deserts where you don't have mass transit available.

One of the promises of autonomous is going to be safer—check—but then eventually, autonomous will make transportation cheaper as well. We want to make sure that that cheap, affordable, safe transportation is available to everybody, regardless of where you live.

Dara, the concern right now is there's a lot of fear mounting about the future.

Sure.

Uncertainty—

Yeah.

Last night, we launched a Future Vision XPRIZE asking teams to create films that envision a very positive vision of the future, so we have something to look forward to.

This idea that Mark and his team have had is about universal basic services. So if a family doesn't know whether they have a roof over their head or where food is coming from, or anything—I mean, you had a disruptive childhood growing up in Iran—you can't think of anything else. If there is a delivery of basic fundamentals—food, water, energy, health, those sorts of things—then at least you can now think about dreaming about, “What do I do next with these incredible tools?” So that's the direction we want to head.

That's great.

Peter, to your point also about the bet, I think there's a possibility to decouple housing from land to get to that, too. So I'd love for Uber to be thinking about that.

Wonderful.

Thank you.

All right. Thank you. Thank you. Let's go to Charles on Zoom. Charles.

6. China Shapes Autonomous Expansion

Dara, this is going to be maybe a difficult question, but hopefully you can address it. Obviously, in China, Uber's business didn't work out, and now China's expanding into the Middle East and other emerging markets. I know you're in 70 countries, and their estimate is that 70% of their cost goes away with the autonomous car—the cost of the driver.

So what did you learn from your experience in China as an organization, and how do you plan to address the Middle East and other emerging markets if they flood the market with cheap autonomous vehicles?

Thanks.

When I joined Uber, Travis and Emil and the team had already decided to exit China. China is an unbelievably competitive market. I competed in China with Expedia and, to use the technical term, we got our asses kicked. So I can't say I disagree with that decision. I thought it was a very good decision, and I think it was absolutely the right thing to do.

We have a big stake in DiDi, which is the largest Chinese rideshare player. We haven't monetized that yet, but we're looking forward to that at some point. We work with Chinese partners. We work with Chinese OEMs. For example, BYD is a very big partner, in that they provide affordable electric EVs all over the world.

We still believe in EVs. They're great cars. Obviously, they're better for the environment, et cetera. In the same way, we're also working with Chinese autonomous players. So whether it's Pony.ai or WeRide, and hopefully Baidu as well, we are working with these players in many markets outside of the US, obviously. But we believe that building bridges to China is the right way forward for us. Autonomous vehicles are going to be an industry that we think the Chinese have a lot to offer, and we're working with these Chinese partners today as we speak.

Yeah. Amazing. Andy.

Thank you. Thank you, Peter. Thank you, Dara. I'm Andy from Hong Kong. We run an autonomous insurance company, and we insure EVs, AVs, and robotics.

My question is: Working with underwriters, are there any specific challenges or desires that you would love to work with actuaries or underwriters on? That's number 1. And the second one is, would you consider embedding insurance into autonomous vehicles as well?

I think the second one, absolutely, we would. The model hasn't been figured out yet, right? This is a very, very young industry. Is it going to be the driver? Is it going to be the platform? Will it be some kind of combination? All of that is TBD, so we'd love to talk to you about that.

In terms of the challenges, as with any actuarial kind of business, you need the volume, right? Right now, while autonomous is a reality today and is growing very, very quickly, it is a very small industry. I think last year, all of the autonomous trips in the world represented less than 1% of our growth in volume—not just the volume that we had, but the amount of rides that we added last year.

We need the law of large numbers to start speaking to us, so as to figure out how to price the services that you provide.

Yeah. Thank you. We provide the data to actuaries.

Yeah. It's perfect.

I'd love to have a chat with you. Thank you.

Nice. Thank you.

Ilia, one of our teens. Good to see you.

I'm going to need to get my driver's license in a couple of years. But after hearing this, I have a question. If cars are going to drive themselves and drones are going to be delivering everything, will I even need to get my driver's license when I'm 16, or will robots just do everything?

Well, you may not have to get your driver's license, but go ahead.

Ilia, how old are you now?

13.

You're 13. 3 years away.

Ilia, I think 3 years from now you will still want to get a driver's license because even though these AVs are coming, mass production of these AVs is going to take some time. They're very expensive today, much more expensive than a regular car.

But 10 years from now, you won't have to drive, and you'll have plenty of choices. It's kind of fun to drive, though, so I'd encourage you to get your license.

Nice.

10 years ago, I made 2 predictions. Our sons were 4 years old at the time, and prediction 1 was that my son would never go to university, and prediction 2 was that he would never get a driver's license. I predict today that in 2 years he won't need to get a driver's license. He'll just get it so he can get away from his parents. But the need won't be as great, and certainly it's going to collapse over the next few years.

Yeah, my son is 21 years old. He doesn't have a driver's license, and he doesn't seem to have any intention of getting one. He's got a big Uber bill, but that's okay.

He's got contacts.

Great question.

George.

7. Automation Changes Human Work

Hey, Dara. I'm George with Untapped Ventures. We invest in the autonomous economy, and for our investors and me, it's very exciting to see this world where there are billions of agents and billions of all kinds of autonomous robots like the ones you're talking about.

But the human concern comes up, and I love what you said earlier about how you're already thinking about it and extending the kind of work that Uber does. I wonder if you could expand on that. What are your thoughts about what role Uber might play in that, as well as the rest of the ecosystem? I think in this community we talk a lot about how there are going to be tens or hundreds of millions of people who are going to be out of work because of AI and robotics, and it sounds like this is something you guys have obviously already thought about.

And certainly, it would be great to get Peter and Salim's thoughts on this as well, because I think it's a big problem, but I also think it's a great big opportunity for some new startups to come in and solve that. What do we do with all of this new human labor in the next 10 or 20 years when this transition happens?

Yeah, totally. I think the first thing I'd say is the press loves the drama of machines replacing humans because it's dramatic, right? It's interesting. It may not be pleasant, but it certainly gets people's attention. You've got to realize that the press is like any business as well. They will tend to dramatize things to be able to earn revenue, just like any other business in the world. It's not a bad thing, it's not a good thing, but it's a reality.

The reality that we've seen historically is that automation typically doesn't replace work, but it augments work. If you go to probably the Chinese OEMs, which are the most sophisticated, when you go to their factories, yes, there's a ton of work that's automated, and there are robots all over the place. There are humans, and instead of doing the work, they are overseeing the robots doing the work, performing quality checks, et cetera. Yes, there are fewer humans working in that plant than there would have been 10 years ago, but there are plenty of them around.

We see the same thing in terms of AI at the office as well. Just like any other company, we're looking to automate tasks, et cetera. Usually, we don't get to 100% automation. We'll automate 20%, we'll automate 30%, and there's plenty of work to do, so to speak. I do think that over the next 10 years, you will see a lot of augmentation of work and some replacement of work.

Historically, our society has been able to adjust, and as some work has been augmented or replaced, new work shows up. That is my hope as it relates to our society as well. There is a real question because, with the singularity and everything else, the pace of change is moving so quickly. The pace of change that we're seeing as it relates to AI and automation over the past 5 years is happening faster than even I expected, and I'm an optimist when it comes to technology.

I think there's an open question as to how quickly society is going to adjust. Just as it has historically, can it adjust fast enough? From my standpoint, if that's a question, what can we do about it as a big, flexible labor platform? We are extending the kinds of jobs available so that by 2035, we've got 20 million people on our platform versus the 10 million people on our platform now. They may be doing different kinds of tasks, but there will be plenty of work to go around on our platform.

Ultimately, whether it's retraining or new jobs, it will be up to both private enterprise and public enterprise to take a lead here, to lean forward. I don't think it helps to worry about this stuff, because what the hell is worrying going to do for you? My push is, let's do something about it. We, as a platform, are doing something about it. I think Peter's push on labor being able to own assets is another great push as well.

There's no question that in our capitalist society, capital is the thing that's being optimized, sometimes at the expense of labor. To bring labor along so that they can be owners, that is a very, very positive direction for us to go in. Hopefully, that's something else that we can—

I love this. Uber is the societal capacitor, being able to absorb and then discourage electrons as required.

I mean, the kind of work at Uber is obviously—when I was growing up, it used to be that you'd have a job for life.

Yes.

Now, a lot of times, you have a job for a week. Then you move on to something else, right?

That social contract is dead.

Amazing, Per.

Thank you. My name is Per. I'm from Chile. My family has a shipping company, and part of what we transport is oil, gasoline, and jet fuel. My question is: Do you think EVs will completely displace combustion engines? And if you think so, how quickly will it happen? How can I calculate that?

I think the onset of these technologies in the physical world always takes longer than in the digital world. You need an ecosystem to move these cars over to EVs. Based on what we see, EVs are a better product, but the ecosystem, especially in the US in terms of charging infrastructure, is just not there yet. That's why EVs have not gained the kind of traction in the US that they have in Europe.

One of the benefits of AVs is that AVs are EVs. As autonomous vehicles become a bigger reality in our network, our network is going to move to EVs as well. I think 20 years from now, I don't think you're going to see too many combustion engines. But 20 years is a long way away, and the world is going to be a very, very different place by then.

Mm.

We'll go to Sagar on mic 1.

Thank you, Peter. Thank you, Dara. My name is Sagar Chopra. I'm a co-founder of a company called Empower Sleep. Before Empower Sleep, I was actually at Uber for about 4.5 years, focused on international expansion.

I always tell people it's one of the coolest jobs that I ever had, so I've always really admired your leadership as well. One thing I really admired about the culture was that everyone ran toward the fire: super entrepreneurial and very driven. Everyone had a lot of agency over their own decisions, and it just created this magic within the company. How do you preserve that as the company grows and scales and doubles or triples in size? I would love to get your perspective on it.

Great question, Sagar. It's something that companies, as they scale, all have to face. We don't get it perfect by any means, but the 2 things that I would point out are, first, one of the core values at Uber is “Do the right thing. Period.” That was a value that we introduced during a difficult time, when you could argue we weren't doing the right thing.

At the time, most companies, when they have values, have the headline and then a description of the value. With “Do the right thing,” the description is “Period.” I remember when we introduced the value to our employees, people were like, “What does that mean? What does ‘Do the right thing’ mean?” What I told the team is, “It means you are responsible for doing the right thing.”

Every single person in our shop has to take that responsibility personally as far as what they're doing. Sometimes doing the right thing may be to play it safe, or maybe to take a risk, or maybe to run fast, or maybe to fold your cards. You never know exactly what the right thing is. But for us, we have a culture where individuals at Uber, 3 or 4 levels down from me, still can impact the direction of the entire company.

That's something that I was able to take from Travis and the culture that he built. There's a lot of criticism about him, et cetera, but he built a culture of go-getters. I got to inherit that culture and keep pushing it forward. “Do the right thing” is also a message of responsibility at all levels of the company.

For me, as companies get larger, they tend to get more conservative. The exact opposite should be true. Peter, you were talking about when we joined the company, we were losing $4 billion. If we made a decision and lost $1 billion on that decision, it would be a freaking disaster. We're going to generate $10 billion in cash flow this year, so we should be taking big bets.

While I'm not rooting for these bets to fail, some of them are going to fail. The fight that I have all the time—it's a fight with myself, sometimes with my board, and with my team—is that companies tend to get more conservative as they get bigger. The exact opposite should be true, because you can take big bets at large scale and be just fine if they don't work out.

It ain't perfect. I fight it every single day and talk to the team about it, but they're the ones who are the doers. There's only so much I can do.

Love it.

Awesome.

Love it. We can take one last question from my deputy son Bear over here. Bear.

Hey there. Following up on that last question, I was wondering: When you run one of these huge companies, how do you decide when to pursue a new field rather than stay conservative and continue just collecting cash flow?

Yeah.

So we want to make sure that we are continually experimenting. Usually, when we go after a new field, the way I put it is, it’s got to rhyme with what we do. If it’s something new and crazy out here, a startup is probably better equipped to do it.

But if you look at the path of Uber, we started with rides. If we’re going to move people around, why can’t we move things around? So then we went to Eats. If we’re going to move food around, why can’t we move bulk around? So we actually went and started Uber Freight.

If we’re going to move people on the ground, why can’t we move them in the air? That was Elevate, et cetera. So what we try to do—the pattern that we find—is, hey, what are areas where we have some core skill set that gives us a right to win in this adjacency?

Uber AI Solutions has nothing to do with movement, but it actually is a kind of flexible work opportunity that we’re bringing either to our drivers or entirely new people coming onto the platform to go out and design new models, do labeling for new models, or test out models, et cetera. Again, it rhymes with us because we are that platform for flexible work.

Adjacent.

So if it rhymes, I’ll listen. If it doesn’t rhyme, it’s better for someone else to do it.