I’m sitting in the back of a Waymo self-driving Chrysler Pacifica parked on the side of a road inside Castle, a former air force base in central California which was leased in 2012 by the company then known simply as Google.
On the beige headliner, behind the traditional driver’s seat, is a black rectangular panel and a series of four uniform buttons. My eye is drawn to the only bright blue button on the far right labeled “start ride.” I push it and a few seconds later the the minivan pulls out onto the road.
The Pacifica is programmed by Waymo to operate at low speeds, but it’s moving at pace that is more than a slow crawl. We maneuver around cyclists, a couple of pedestrians, and a crestfallen man standing next to a broken-down Hyundai, all of whom are among the cast of actors contracted by Waymo to simulate the real world at its 9-acre test facility.
It’s a smooth, yet totally unremarkable ride. Only there’s no one behind the wheel. There’s not even anyone in the passenger seat.
And I feel completely safe.
In the past two years, I’ve ridden in four self-driving cars. I’ve even been behind the wheel of one. In addition to being mundane rides, the one thing these cars have in common is they always had a safety driver at the wheel. After riding in the Waymo Pacifica, I’ve finally experienced what big car and tech companies are chasing: full autonomy.
Waymo, the self-driving car division of Alphabet, has been testing its autonomous vehicles at the Castle facility outside Atwater, Calif. since 2012. But yesterday, the company took the unprecedented step of inviting a group of tech and auto journalists for a behind-the-scenes look at how the Google spinoff would achieve the tall task of deploying fully driverless cars. Not just highly automated cars, mind you. Cars with no human required.
To date, Waymo has been close-lipped about the nature of its autonomous testing. It has periodically provided updates about the 3.5 million miles accrued since it launched its self-driving efforts, much of which were logged at Castle. The press has had limited access to its ubiquitous Google car prototypes, the modified Pacificas and Lexus SUVs, and the engineers and designers charged with building its program.
Monday’s event marked a shift in Waymo’s strategy, as it offered unprecedented access to it test vehicles, facilities, and engineers to a group of media. (The Atlantic got an exclusive look earlier this year, but no fully autonomous rides.) Waymo is intent on showing that its test cars are able to handle the complicated, complex, and increasingly dangerous task of navigating a two-ton machine on public roads. Not only that, they can do it more efficiently than humans. Waymo appears to be moving toward making the flesh-and-blood driver obsolete.
“In level four mode, you can imagine a completely empty car coming to where you are, you open the door, hop in the back seat, and it can take you — relaxed and happy, perhaps it has WiFi — wherever it is you want to go,” says John Krafcik, CEO of Waymo. “That’s what we’re striving to achieve every day.”
But Waymo is still a long way from being able to bring its fully driverless cars from the protected, highly scripted scenarios of its test facility to the wild, unpredictable open roads. Regulators and local governments are still grappling with how to regulate cars without steering wheels and pedals, and surveys indicate that most Americans don’t know enough about the technology to fully understand it, much less trust it.
Plus, Waymo is finding itself in a space that is becoming increasingly crowded with other big, well-heeled players. The Alphabet subsidiary boasts that it has the most experience, the best technology, and the smartest minds working on its self-driving cars. But what it has not set forth is a clear strategy for making money off its advanced systems. It’s not a car company, an automotive supplier, or a ride-hail company like Uber or Lyft. It’s part of Alphabet, one of the world’s biggest and best-capitalized tech giants, so profit isn’t an immediate concern — but it will be soon.
Ironically, its lawsuit against Uber may provide it with its first revenue, if a judge agrees that the embattled ride-hail company conspired to steal its self-driving trade secrets. That case is set to go to trial at the end of the year.
During the tour, Krafcik wouldn’t comment on reports that the company is getting ready to launch its first commercial product. But he did describe the types of ventures Waymo was considering: ride-hailing and ride-sharing; trucking and logistics; working with cities to help better connect residents to public transportation; and selling and licensing its technology to OEMs (original equipment manufacturers). “Whether they are Waymo-branded or not, it’s probably too early to say,” Krafcik says. Waymo has also set up shop in suburban Detroit, in close proximity to several of the predominant automotive industry suppliers.
The company has partnerships with Fiat-Chrysler, Lyft, and Avis, and its been manufacturing its own sensors and hardware since late last year to reduce dependence on suppliers. And since last April, the company has been operating a limited-run ride-hail service in a suburb outside Phoenix, Arizona so it can monitor how real people would use and react to its vehicles. Those rides always feature a safety driver behind the wheel, but Krafcik said that the driver is increasingly “fading into the background.”
Also during the tour, Waymo provided a closer look at the interiors of its self-driving minivans, as well as the backseat user experience. In addition to the blue “start ride” button, there is another button that reads “pull over,” in case a rider need to unexpectedly suspend the trip. There is also a “help” button, a la GM’s OnStar, that someone could push in case the car breaks down. (During my ride, I was able to chat briefly with a Waymo technician based in Austin, Texas.)
Two screens on the back of each of the driver and passenger seat headrests show a top-down visual map of what the van’s sensors allow it to “see.” But rather than the raw “X view” version used by Waymo engineers, the version passengers see looks like a design studio product — it’s simple, congruent, and the script is uniform. Other vehicles and cyclists are represented by blue rectangles. Emergency vehicles are distinguished by a red light that circulates around the rectangle. The symbol for pedestrians appears in a ghostly white hue. And orange traffic cones look like orange traffic cones, only tiny.
“It’s all about curation,” said Ryan Powell, head of Waymo’s UX design team. “Knowing what to show, how to show it.”
Other features, such as navigational directions and time-to-arrival, are meant to give passengers basic information they need about their journey. They also help assure riders that the car can see construction zones and all the other variables that make daily driving unpredictable. (Waymo says it has an app that test subjects in Arizona use to hail its minivans, but the company wouldn’t let us see it.)
But the interior of a self-driving car is worthless if no one agrees to get in one or trust their belongings in the cars hands. Many Americans say they have serious reservations about self-driving. This is where Waymo’s real work at Castle comes into play: testing and retesting all the different scenarios one encounters on the road. Aggression. Unpredictable behavior. Bad infrastructure. Debris in the road. Unrestricted intersections.
Waymo first began conducting its structured tests in a partially secluded parking lot near the Shoreline Amphitheatre near Mountain View, but has since moved all this testing to its Castle facility. That’s where Stephanie Villegas, Waymo’s lead on structured testing, ran a couple demonstrations for us. Villegas (wearing aviator sunglasses, a witch’s hat, and a black cape embroidered with gold stars, in the spirit of a press event held on the day before Halloween) explained how Waymo installed and constructed a variety of driveways, lane-changes, and hairpin turns to let the cars run dozens of tests. Waymo’s engineers then collect the data from those tests to recreate them in simulation, allowing to rack up even more miles and experiences for its vehicles.
“We really put our cars through the ringer of real world conditions,” Villegas says. “Things like deep, severe potholes, over harsh speed bumps, steep incline scenarios, and shooting concentrated jets of water at the car to see if water gets in any of the crevices. And that will model conditions we see in the real world and want to be ready for.”
We watched Waymo’s self-driving minivan dodge an aggressive Honda convertible, an obliviously reversing black Fiat, and a trio of Waymo employees pretending to fumble a bunch of moving boxes in the middle of the road. (Waymo calls these actors “fauxes,” pronounced “foxes,” as an homage to the wild foxes that sometimes infiltrate their site.) But unlike our test ride, Waymo always puts a safety driver behind the wheel during structured tests, just in case. These scenarios are then run through a simulation, where engineers will “fuzz the variables” to account for a variety of different outcomes, Villegas says. Waymo has even orchestrated a number of collisions to see how its vehicles operate in a crash scenario, but Villegas said there has never been an unintended accident.
But what happens if the movers with the boxes don’t get out of the way? And Waymo’s car refuses to cross the double line to go around them? Villegas says the car would never intentionally break the law, even minor infractions that a human driver would have no qualms violating in order to get cars moving. That could create bottlenecks, with self-driving cars holding up traffic because its law-abiding brain won’t allow it to proceed, even if a stoplight is malfunctioning.
It’s these types of complications and unanswered questions that Waymo is furiously working to address as it speeds toward making full autonomy available. “I think for a lot of the world right now, this can feel really theoretical,” Krafcik says. “This new world of fully self-driving — what is it? I think one of the reasons for that is because all of the folks playing in the [Level 4] space right now have drivers in the driver seat. It doesn’t feel like fully self driving. When you drive on public roads, you have someone in the driver’s seat. Sometimes it’s part of the law, or they’re doing it because it’s just the safe thing to do.”
The leap to fully driverless cars will be a hard one to make. But Krafcik — and Waymo’s parent company Alphabet — intend to be the ones leading the charge. “Our goal, our intention, is to bring this technology to the public,” he says. “To the world.” The question is, are the rest of us up to go along for the ride?