Riding in a self-driving vehicle requires you to suspend your belief that man is better than machine.
You have to trust that the steering wheel moving below your hovering hands is turning in the right direction and the right amount.
You have to hope that the acceleration you feel will stop when you reach the right speed and not careen out of control.
You have to believe that this truck, a mashup of metal and gears and circuit boards, is a better driver than you, for you have given control of your life to a machine and are blindly trusting that it actually sees the road better than you.
It’s a leap of faith, and one that I only realised how miraculous it was after I saw how the self-driving truck I was riding in knew the difference between an open lane and the open sky.
Flipping the switch
I’d never been in a self-driving vehicle until I rode in the back of an Otto truck on Friday.
Uber had just revealed that it bought the company for an estimated $680 million for its technology that takes existing, newer trucks and retrofits them with a combination of sensors and radar.
Otto’s goal is to make a truck drive itself in a way that’s so safe and reliable that a long-haul trucker could take a nap in the back. Its promo videos show a truck with an empty seat in the front, the steering wheel eerily turning by itself.
Otto’s product lead Eric Berdinis says the goal is for a trucker to be able to go to sleep in El Paso and be able to wake up in Dallas — about nine hours later.
But it’s still the early days for self-driving technology. An ardent Tesla fan died after a truck turned in front of his car, and neither he nor the Tesla saw it. So Otto, which hasn’t had any accidents, knows something will go wrong at some point.
So far, Otto’s testing still requires that a trained and licensed manager, like senior program manager Matt Grigsby, to sit behind the wheel, hands hovering just in case something happens. Systems testing engineer Brian Gagliardi sat next to him as a co-pilot, reading the screen on a laptop to make sure the truck sees what he sees and doesn’t make a mistake.
The dream as Berdinis describes it is that one day the dual-trucker system most companies use is replaced by one trucker and a computer. Instead of splitting shifts for sleep, the partner would be the computer and the trucker would handle things like exits and driving in cities.
Leaving Otto’s headquarters in the SOMA neighbourhood of San Francisco, Grigsby, the human driver, took the wheel until we got on to Interstate 280. Then he pushed the “engage button”, a bell chimed a few times, and his hands floated away from the wheel.
A sky vs a lane
The computer taking over felt just like when you engage cruise control. There’s a bit of whoosh as the speed changes and the truck corrects itself to the center of the lane, and then everything evens out.
It’s not a dramatic switch until you notice the small things, like Grigsby’s hands in his lap for a few seconds or his feet flat on the floor even when the truck is accelerating. For the most part, it felt normal — relaxing, even.
While looking at the road, I started to notice just how bad the human drivers around us are compared to the even keel the truck is keeping.
You see cars drifting in-and-out of the center of their lanes, even if they’re staying inside the lines. Some cars fly by, while other cars are going slower than we are. Otto sets its software at the match the speed for trucks on that stretch of highway. In this case, it’s going 55 m.p.h.
As the interstate became a bridge and my view shifts to rooftops and blue sky on the left of the cab, I started to realise just how much of a miracle it is for the truck to not be flying off the edge.
Its combination of sensors and cameras have to not only detect that it’s on a bridge, but also know not to direct the truck towards that empty expanse of sky. The empty area to the left of the truck doesn’t mean an empty lane, but actually something much more fatal than that.
In Pittsburgh, where Uber (now Otto’s parent company) is testing its self-driving cars, a Bloomberg reporter documented how the self-driving system chimed to get the safety driver to take the wheel for a few seconds as they crossed a bridge in a test ride.
“Bridges, unlike normal streets, offer few environmental cues — there are no buildings, for instance — making it hard for the car to figure out exactly where it is,” Max Chafkin wrote. Uber’s car flipped back to autonomous a few seconds later.
With that in mind, I thought I’d be terrified when I looked at the rooftops rushing by as we wound our way into San Francisco, but really I was in awe.
This is a technology where an error could cost lives, yet I thought my odds of going off the bridge were probably about the same as if I was behind the wheel myself. Or even less. From our steady perch in the self-driving truck, it was the human drivers that looked out of place and chaotic — not the truck that was miraculously driving itself.