The screen tells you when the car’s driving itself, your speed, and your route. The lower two-thirds of the screen depict the car and its surroundings as understood by the spinning LIDAR unit on the roof. The idea is to give riders an idea of what the car sees, so they don’t wonder if the robot has noticed that truck up ahead.
My ride starts with a tour of Pittsburgh’s warehouse-filled Polish Hill, before we use one of the city’s many bridges to cross the Allegheny River. As the car navigates four-way stops and traffic light-controlled intersections, the ride is mostly smooth, with scattered moments of aggressive acceleration or braking. The engineer at the wheel takes over control every few minutes. Once, he’s not happy with how long the car is waiting before slowing for a pedestrian. Another time, he manually steers around a double parked truck, knowing the system will just stop and wait for it to move.
All along the route, passers-by gawk and take photos—Uber has been working in Pittsburgh for 18 months, but autonomous cars still aren’t as common here as they are in Silicon Valley. The San Francisco-born company chose Pittsburgh for a few reasons. The City of Bridges is home to Carnegie Mellon University, with which Uber partnered on autonomous tech, and from which Uber poached dozens of researchers. And the city offers challenging test conditions, from rain and snow to a tangled grid of narrow streets. Maybe best of all, it’s enthusiastically pro-tech.
“Startups will choose places based upon the ability to be innovative and there’s never a time where regulation comes before innovation,” says Mayor Bill Peduto. “If you stop the clock all you’re doing is assuring that that tech and those jobs will be in another city.”
While states like California, Michigan, and Nevada encourage and regulate autonomous tech, Pennsylvania’s been all cheerleader. The state has no real rules about who can operate these things, or where. Which is extra cool, because it means that after half an hour in the back seat, an Uber rep asks if I’m interested in taking the wheel—or, you know, sitting in front of it.
Still, after hitting the nickel-sized silver button that makes the car drive itself, I can’t demand a trip back to San Francisco. Like just about everyone working on full autonomy, Uber runs its car only in areas mapped in extreme detail. Those maps provide key context for the cars and let them focus on temporary obstacles like other cars and people.
Building those maps requires sending a human-driven, sensor-laden car down a stretch of road seven or eight times; using software to extract the important stuff like traffic light locations, crosswalks, and speed limits; having a human vet all that; and getting it onto the car. Uber will soon expand the cars’ territory to include the main road connecting downtown to the airport, and Pittsburgh’s bar- and student-filled South Side. But figuring out how to streamline that mapping process is key to the company’s dreams of replacing a planet’s worth of Uber drivers with software.
Uber won’t say when, exactly, it expects to remove the bags of meat from the driver’s seat. Its competitors are promising fully autonomous, market-ready cars in the next three to five years, and the ridesharing behemoth is likely on a similar timeline. But Uber has already managed to translate its experience to the autonomous age, complete with a “driver” rating. And no small talk required.