Or how I learned to stop worrying and love self-driving cars
by Pearl Tesler • March 18, 2015
If you live anywhere near Silicon Valley, then you may already have glimpsed the eerie sight of a driverless car flying down the freeway.
My initial reaction to these rolling Google ghost ships, shared by many, was to be afraid—very afraid. But one short symposium at the 2015 AAAS Annual Meeting disabused me of my human hubris: Not only are autonomous vehicles coming inexorably down the pike—they are going to kick asphalt. Which is to say, they are going to perform much, much better than we can.
For starters, there’s reaction time. At best, it takes even a closely attentive driver between 1 and 2 seconds to notice that the car ahead is braking and to start braking in turn. A self-driving car can react at least a hundred times faster, shrinking the response time to a mere 10 milliseconds.
Such superhuman reaction times confer obvious safety benefits, yet there are other practical upsides. Trucking companies are already beginning to implement robotic platooning systems that wirelessly link the throttle and brake of one truck to another close behind, allowing them to link up into snug aerodynamic pairs that save fuel.
Robocars will also best us with their sheer computational finesse, a fact they’ve already proven on the racetrack. Chris Gerdes of Stanford University and his team are fine-tuning an autonomous racecar, an Audi TTS called Shelley. Like a modern-day KITT, Shelley scans the road far ahead and instantly calculates the optimal line through a series of curves, closely matching and sometimes exceeding the performance of professional racecar drivers.
Finally, your autonomous car will be privy to information that you, mere mortal, are not. Coming soon by federal mandate, vehicle-to-vehicle (V2V) beacons on cars will broadcast their current stats—their location, dimensions, speed, direction, and whether they’re braking or not—to all other vehicles in the vicinity. Much like the navigation systems found on ships, V2V beacons will help autonomous cars identify potential collisions long before you can, and spot hazards you can’t even see.
Imagine: The moment that guy in your blind spot starts to veer into your lane, your autonomous car springs into action like Batmobile, shunting you safely aside. Holy handbrake, Batman—that was close!
Sure, there are some engineering details to work out. But ultimately, autonomous cars will do for driving what IBM’s Deep Blue did for chess: outmaneuver us at every turn. And maybe that’s just as well. There were 5.7 million crashes in the U.S. in 2013, and an estimated 93% of them were the result of driver error. If machines can drive better than we can, wouldn’t it be reckless to stand in their way?