Driverless: Waymo and the Robotaxi Race—Under the Hood
23:52 Share

With artificial intelligence, creating an ethical foundation isn't just the right thing to do, is crucial to success. Join IBM of the break to hear why from federal binet eris IBM consult into global leader for trustworthy ai.

Back in october, after years of pledging itself, driving technology was just around the corner, electric vehicle maker tesla finally unveiled its first prototype driver was tesla. C. E, O elon musk walked out to technical music alongside someone wearing a space sex space suit, waving to the crowd. He stepped into the passenger side of a sleek gowing, two seeder vehicle, which drove off to the main stage. No driver behind the steering wheel, and actually no steering wheel at all.

As you can see, I just arrived in the robot taxi, the cyber cab.

musk said these vehicles were designed to be entirely driverless. No steering wheels, no break or accelerator peddles.

no side view mirrors, even is here.

And IT feels great. IT was a flashy and tightly choreographed vent held on a set at Warner brother studios in burbank, california, but IT was light on details about the business and the cars, and that left some investors disappointed. The day after the event, tesla shares fell almost nine percent in one of its worst days in twenty twenty four.

Meanwhile, out in the real world, just a short drive over the hollywood hills, members of the public can already hail the driver was car to get around los Angeles, owned, Operated and designed by the biggest name in the business way. More robot taxes are officially in service in L. A.

You can now border ride across sixty three square miles from downtown L. A. To sanom ica without being forced to chit chat with the driver, if that's not your things.

Streets only, no highway driving at this point. This is what we got ta get used to. Again, the future is way more.

Robot taxi business is still wait ahead of tesla and other rivals like amazon's as of october, when o says it's conducting one hundred and fifty thousand paid rides a week across the four cities it's Operating in while its competitors are still testing their technology.

But even with the multibillion dollar bet google's parent company, alphabet, is making on way more and its expansion into new markets, how can IT maintain its lead in the robot taxi race and how might the industry change as the efforts to build and popularized driverless car technology continued? I'm danny Lewis, and this is driverless way o and the robot taxi race, a series from the wall street journal of future of everything. We're looking at what IT would take for way mo to stay at the top of the growing industry with its rivals snapping at its heals today. Episode two under the hood that's after the break.

How do you start to lay the foundation for responsible A I in your organization? Here's fator binder's IBM consultants, global leader for trust.

or the ai IT, starts with asking the question, what is the kind of relationship that we ultimately want to have with A I? The purpose of A I is not meant to display human beings. That is meant to all meant human intelligence. Soon, as you have a glimmer in your eye about how you're thinking, you might want to use A I then asking the questions like what would be required in order to earn people's trust in such a model.

In our last episode, we talked about how we o went from a slow and steady approach to developing driverless cars to first place and continuing to accelerate. But the company is facing some chAllenges on this next lap of the race, and it's also a crucial time for the industry as a whole. So let's start with some of way mos competition, namely tesla. As we just heard, elon musk's electric vehicle company is making a very high profile pivot .

to driverless cars. We saw elon's vision for how he wants to roll out this technology.

W S. J. Tech reporter miles cuppa.

we got a small sedan that he wants to be Carrying passengers as robot taxi, and we got a larger van that he thinks will be able to shuttle all upwards of a dozen people at a .

time from the outside, the robot taxi or cyber cab, looks similar to the testers already on the roads. The van, though, is a retrofuturism looking vehicle, all rounded edges and sharp lines, very architecture. But there are other ways that tesla is an outlier in the driverless car industry.

There's been this divide in the self driving car industry for a long time about whether you need light or or not to achieve fully autonomists driving.

Those are the laser based sensors that let a driverless car detect what's around IT in real time. If you've ever seen a way o and wondered what the spinning cyndi mounted on its sides are, those are light. Our sensors.

wao and zooks, taking the approach that you need light are.

while tesla uses light, are for data gathering. Moscow said he thinks cameras are enough for self driving cars. He talked about this, a tesla's autonomy day of that back in .

April twenty nineteen. Anyone to let rely on lighter is doomed. Expensive sensors that are unnecessary.

Tesla stopped installing light, our sensors in some of its cars in october twenty twenty two. Instead, its full self driving and autopilot systems rely on cameras that are cheaper than light are and artificial intelligence algorithms. Then what's called an end to end neural network takes in raw data and translates IT into the vehicles actions.

The deal with the end is to say, you know what, we're I can write code. We're just onna show a ton of examples.

fill cope man studies driverless cars and how to make them safe at carnegie in university.

camera goes in one end, wheel spinning comes out the other end. It's all machine learning through through us and end machine learning.

That means instead of tesla programmer manually writing instructions for how future driverless cars should handle any given situation, the company can take data in video from all the teslas out there on the road and feed IT into an A I program.

It's sort of like like training a pet and have the pet does something you don't like. You know, you say bad dog and that IT that all knows that like, well, we tried something else and eventually gets good dog.

Copen says, if you have enough data, this method could actually make IT a lot easier to get driverless cars to behave naturally in common situations.

driving down the street, taking a corner in the way a human driver would take a corner. Here's what the human driver did. You should learn how to do the same thing.

But copen says, end to end. Machine learning is not good at knowing how to respond to rare situations, and that's where questions about safety come into play.

How do you train IT when it's time to cross the w line to avoid a crash? How do you train IT to deal with fire trucks with the lights if you hadn't seen? How do you train IT not to drive into a cover, grow that might be flood water?

These are all issues that have cropped up with text less current technology. Earlier this year, a wall street journal investigation into data from over two hundred crashes involving tesla autopilot found that cars using the system sometimes struggle to recognize objects or stay on the road. We reached out to tesler for the series, and the company did not respond to requests for comment.

Another big difference between tesla and the other players in the driver was car world mosques company doesn't have the required california state permits to test its vehicles without a driver present, a spokesperson for the california department of motor vehicles said. Tesla has had a comment to tell self driving cars with a safety drivers since fifteen, which is set to expire at the end of twenty four. They also said tesla has not applied for a driverless testing or deployment permit.

There's different levels of autonomists driving, and you need different permits to do different levels of this driving.

Magan bruschi is a tech reporter for the journal. The levels generally used by the driverless car industry to measure autonomous capabilities are developed by the global standards organization S A E international. They ranged from zero purely support systems for a human driver, like automatic emergency breaking or blind spot warnings, to five, which would be a car that can drive itself anywhere under all conditions. Wao says its vehicles Operate at level because they only driving areas the company has mapped out.

so they have always permits and they've done what they're doing with light and cameras. Tesla, currently, its cars Operate at a like level.

Two S A E international defines that as the highest level of driver support features, which still requires constant supervision and is not automated driving.

A has said that will change. And I think the next year is gonna really interesting to follow and in the next year will see. Can you do self driving cars with just cameras? I think that's a big question mark.

especially in some of the other companies chasing emos success, like amazon's zoo s also use lidar and radar as well as cameras. But while the sensors are similar, they're conducting more limited Operations. Zooks is still testing passenger rides in the driverless vehicles, which, like text less cybercafe, lacks during wheels and breaks and accelerator peddles.

Zoe says. Those are on the roads in one neighbourhood in safran, cisco, the streets of a suburb just south of the city and in lost vegas, including on the strip. But the cost of the technology is one reason why musk is so keen on sticking with cameras and enter n ai. Cost is a chAllenge that everyone in the drivers car industry is facing because running a robot taxi company is starting with the cars themselves.

Let's just take when was an example.

Suta kjartan, as an analyst with market research firm wolf research, SHE follows wo as part of her .

work at a high level. The gust of a fully autonomous car, as IT stands today with wao, includes the cost of the .

card o currently uses jaguar ey Epaces e lectric S U V S.

approximately fifty thousand dollars. Then on top of that, there are different units of light, are about four units of those cost, about another seven to eight thousand dollars, plus their cameras that about two to three thousand dollars. And then on top of that RAID are another two .

three thousand k jerious says. Altogether, SHE estimates that each wave of vehicle costs about one hundred twenty five thousand dollars. A way more spokesperson declined to comment .

just for a context, we think maybe twenty five to thirty thousand four duser.

During the unveiling event we heard earlier, elon mosque said he expects the tesla to cost less than thirty thousand dollars, but in addition to the cost of the cars, caja ias says all the computing power and research and maintenance that goes into the robot taxi business is likely costing way o and billions of dollars per year, a way most spokesperson declined to comment. Alphabet did not respond to requests for comment.

The going consensus view is about three to five years lead that way more has over anyone else in the industry. What does way more need to do in order to keep that .

leadership and to become profitable will look at what's next for way most future and the business of driverless cars after the break.

Right now, members of the public can hale a amo in a handful of cities, some vertex o of course, but also los Angeles s and phoenix s arizona. But that will soon change. In september, we have announced that is partnering with uber to bring paid robot taxi rides to uber s platform in two new cities, Austin, texas and atlantic, georgia. This deal has some interesting details, first being that uber will take over all vehicle maintenance and right booking. W, S, J, tech reporter miles groupa.

uber is bearing the cost of cleaning, repairing these vehicles, storage Operations, a lot of the sort of native quality of running these vehicles in these local markets.

But there's a chance that uber rides and lana N N could get caught off guard when a hailed car shows up without a driver. Uber might automatically assign them a amo unless they intentionally opt out or decline the car within a certain time period.

You can see that easily driving more adoption of o as long as enough people opt in.

in exchange, the two companies will share the revenue from the way. More rights, if you haven't been keeping up with all the ends and outs of this technology, you may not realize that uber has a complicated relationship with driverless cars. At one point, IT had its own research division focused on the technology, but in twenty seventeen, we most suit the right haling company for allegedly stealing and using trade secrets.

Uber agreed to settle in february twenty eighteen, but then the next month, one of its test vehicles driving in tempe, arizona struck and killed a pedestrian crossing a street, while the safety driver behind the wheel was distracted by watching T. V on her phone and in two and twenty, uber got rid of its driverless cardiff sion entirely in uber spokesperson said the companies is current approach to autonomous driving is to partner with vehicle developers, fed Operators and cities, including with wao for rides and phoenix, though we o also Operates its own APP there. But mile says this new deal is different .

zuber has a huge base of people who have downloaded the APP and so by getting in front more people and increases the likelihood that people will be calling their cars um throughout the day.

the theory being that is more people ride in way mos, the more comfortable we'll become with riding in wees or other driverless cars. But also because the amount of time a robot taxi spends empty can have a big effect on how much money he makes the company.

Your not making money when when your driving test miles, you're only making money when you have paying passengers in the car.

W, S, J. Tech reporter Megan brusk looked at public data from Wayne's Operations in california.

I got some numbers in from september last year to make this year nearly forty percent of the miles that way o drove do not have passengers in them. A O spokesperson .

says the company is working to reduce the number of miles its cars drive without passengers. Some of that male's may be from testing, but unlike a right haling company, where drivers are only paid when they have writers in the car, way more cars have extra costs associated with their technology, which makes getting the vehicles onto services that already have a lot of users like uber really important.

That reduces the amount of miles that the cars are driving without paying passengers. And that kind of density of use is really important for way out gets to profitability ultimately. So you can kind of see them thinking a bit more about how they're going to turn this into a real business.

More isn't giving up its own APP yet, though in early december, the company announced plans to bring its way o one robo taxi service to miami, florida in twenty twenty six. But mile says this is still a big chAllenge that wyo. And its rivals are going to have to contend with as they try to make back their investments.

On the flip side, they don't have to pay drivers. And so the calculation becomes, for each car that you put on the road, how many miles can you get out of the car? And then of those miles, how many european passengers? And then for each ride with paying passengers, how does the revenue you bringing in compared to the costs? Those are kind of the economics that the company and its investors are thinking about right now.

A way more spokesperson said that only collects fares from paying passengers, but all miles driven provide meaningful experience to the automated driving system. Chauta kader ia, the wolf research journalist, estimates that could take a while for way o to break even.

Assuming that way, mo has invested about fifty to twenty billion in capital investment, even if you assume about forty five daily trips, which is three trips an hour, which is a budworth uber is um and twenty thousand vehicles IT would take them ten years to even break even, which is a very long time.

A way more spokesperson declined to comment. Cojero says a few things could shorten this timeline. First, rides have to be competitively Priced.

We did an experiment with twenty rides within sentences go and twenty different ride requests within photo x, two of the biggest markets for o and across those forty different rides, wao was consistently approximate twenty five to thirty percent more expensive than an override.

SHE says rides that are cheaper, or at least similar to what writers will pay for a person to drive them in a taxi, will encourage more people to give way more a shot, uber says. Writers who are matched with emos in Austin, in atlanta will pay the same rates as if the APP had assigned them a human driven car.

The more the usage, the more the Price gap is gonna narrow, and we are just not there just yet.

But caja IaaS says one reason the rides are so expensive is probably because the vehicles are so expensive. Like we said earlier, he estimates that each wave of vehicle currently costs about one hundred twenty five thousand dollars before getting into regular madness.

That is a huge amount of investment towards what IT got to ensure the car, the hardware, the cost to serve that is managing these units and then the research that goes behind making the autonomists technology Better every day.

But kjersti says there's another option we moo can take to become profitable, getting out of the day to day business of Operating taxes entirely.

We are more likely, in my view, to see a scenario in the next three to five years whether way mo technology is then integrated into fords and Christ lers and one days a uh toyotas etra. But we will get a high margin licensing fee for that technology.

We most partnership with uber presents a potential blue print for establishing markets in new cities as IT tries to accelerate its growth, though new cities also mean new chAllenges for its technology.

When these companies scale up, a lot of lessons will become apparent cause you just more opportunities to have a problem.

Carnegie melon university professor.

all these companies have pressure to make their investors happy. And the question is always, are the companies holding back the pressure well enough that they can expand their Operations at a reasonable printed pace? Or they coming to the pressure in ramping up faster than they should?

Some of way most founders have expressed support for its efforts to get the execution right. In october, we more closed its latest investment round, raising five point six billion dollars from firms including Andrews and horror z, silver lake and tiger global. In a statement at the time, tiger global founder chase coleman said wyo.

Had baLanced ambitious goals with responsible execution in silver like coffee o egan urban said waymore driving technology leads in earning trust a way more. Spokesperson said the company is not facing investor pressure to expand quickly and its expansion plans are guided by a safety framework over the last two years, wo has gone from just ten thousand paid rides a week to one hundred fifty thousand across all its markets. And IT announced plans to bring its driver lost cars to three more major U.

S. Cities as the company tries to make the most out of its lead in the robot taxi race, a lead which just got wider. A few days ago, one of the most biggest competitors announced IT is abandoning its robot taxi efforts. General motors will continue to work on self driving, but its scrapping cruizers work on robot taxes because of the time and costs needed to scale the business, G M, invested upwards of ten billion dollars into cruise. As for what might happen to way mo and the driverless car industry at large, W, S, A reporter Megan bruck says, this is way most moment to find out.

Now is the time for them to do this. There is no other competitors in the market. And so if you can get the name recognition with people, that is the self driving cars with you and you feel safe in a way. Mo, and so you're going to take a way to feel brand loyalty right now might really pay off a way out once other competitors do that for the market.

The future of everything is a production of the wall street journal. Stephanie egan, fritz is the editorial director of the future of everything. This episode was produced by me, Daniel, special thanks to miles groupa mega brasi antipa ki Peterson, our fact checker is a partners, Nathan Michael, level and just contenting our our sound designers and rode our theme music.

Catherine house office are supervising producer we had helped from section editor digna island, I shout, all muslim is our development producer Scott salary and presidents are the deputy editors, and philonous patterson is the head of news audio for the wall sty. Jero, like the show, tell your friends and leave us a five star review on your favorite platform. Thanks for listening.

Earlier, we discuss what responsible AI looks like in practice. Here's fator born aeris from consulting again on why that begins with data.

My favorite definition, the word data, is an other fact of the country experience. A, I is like a mirror that reflects our biases back towards us. But we have to be brave enough and introduction cof enough to into the miron decide, does this reflection actually alive to my organization? Values if IT aligns, be trees cared about. Why did you pick the data that you did? If IT doesn't like that, when you know you needed to change .

your entire approach, learn more about IBM artificial intelligence consulting services and IBM 点 com flash consulting。