I am enthralled with the book I’m reading right now. Boy, is it good. This is a frequent problem for me because I like to read in the morning when I’m in my car on my way to the office. Today, as I read behind the wheel of my Chevy, I risk a ticket for a moving violation at best—and vehicular homicide at worst. The inevitable arrival of driverless cars will not only keep me out of jail during my morning commute but will also solve a host of competing problems. The inability of the elderly to drive due to infirmity, highway paralysis from commuter traffic, and deaths due to driver inattention, inebriation, or an ill-timed sneeze—will all be eliminated by cars which have no need of a driver’s input beyond a desired destination.
People are the weak link in the automobile ecosystem. If we weren’t part of the control loop, things would be safer. A lot safer. It’s estimated that 94 percent of the world’s yearly 1.2 million automotive deaths involve human error. We’re uncorrectable bugs in the system, you and me. We get into accidents once every 500,000 miles, and cause fatalities once every 1.3 million miles. But the software that’s driving autonomous cars? It was performing at levels equal to or better than that error rate three years ago. We can assume the algorithms that model human driving behavior are even better now.
Several companies want to bring about this wheel-borne revolution. Among the most eager are Internet search giant Google and electric car manufacturer Tesla. Google worked on driverless technology for several years and broadened its efforts to include an all-new car platform to carry it. Looking like a cartoon vehicle designed for Paddington Bear, Google’s test vehicle aims to convert the driver into completely passive cargo. This Car of the (near) Future carries a suite of scanning and sensing hardware to know where it is in space, what’s around it, and how to plot a safe course to deliver the passenger. Not only will this car deliver you safely home after your colonoscopy, it will electronically partner with other autonomous vehicles to eliminate the need for intersectional stop lights, giving you the shortest possible ride.
While Google has offered no public dates for a debut of its adorable gnome mobile, Tesla took a phased approach in which its currently available vehicles are outfitted with a driver assistive technology called Autopilot. The name Autopilot certainly leads me to expect the similar system on a Boeing 747, where I can take my hands off the controls and have a sandwich in the back of the cabin. Tesla says Autopilot automatically steers down the highway, changes lanes, and adjusts the car’s speed in response to traffic. Many customers have taken Tesla at their word and delighted in posting videos of themselves flying down highways in Tesla vehicles, hands at their sides, the steering wheel gently oscillating.
Some of the shine of this technological marvel was dulled in June 2016 when a Pennsylvania man died while Autopilot was in control of his Tesla Model S. It was reported that neither Autopilot nor the driver noticed a tractor-trailer crossing the vehicle’s path. Autopilot failed to detect the trailer because it was colored white and was framed against a bright white sky. Unable to differentiate the obstacle from its background, the car’s software acted as though the car’s path was clear. To compound the problem, the “Mobileye” system that controls braking is designed for rear-end collision avoidance, not side-impact avoidance for objects which are traveling perpendicular to the car. For its part, Tesla says that Autopilot should be regarded as a test feature and that drivers should keep their hands on the wheel.
A specific combination of circumstances revealed a key deficiency that will be corrected. The software will ultimately improve. A human actively driving might have made a mistake that led to the same outcome. Yet, somehow it feels different hearing that a piece of software drove an occupied vehicle into a semi. As we’re told to relax and let the machine relieve us of the tedium of daily life, we come to feel that the machine is magical. It operates on principles and technologies that are beyond the comprehension of anyone who’s not an engineer. When a failure of this sort happens, the truth of the matter is revealed. It’s not magic. It never was. It’s just very, very clever.
Now that the dawn of driverless cars is here, there are important questions that we need to collectively answer. For starters, how much mental activity are we willing to cede to machinery? Anyone who’s driven on an Interstate highway has imagined how boring professional trucking is, but it employs millions of people. Driving a taxi in Manhattan is extremely stressful but it provides employment for thousands. Where will the soon to be displaced drivers go? Will professional driving become yet another industry automated out of existence?
How many complex technologies have we seen operate flawlessly? My first thought is of weekends lost to troubleshooting PCs past or home wireless networks. Every consumer knows what engineers and statisticians know: the greater the complexity of the system, the greater the capacity for system failure. Let’s not forget that a driverless car is a networked computer on wheels.
There’s a maxim in computer security circles that what can be hacked will be hacked. There’s no reason to think that future cars will be any different. For evidence of that, we can look back to 2015 in which “researchers found hackers were able to access the ignition on Audi, BMW, Ford, Honda, Hyundai, Kia, Lexus, Mazda, Mitsubishi, Nissan, Range Rover, Subaru, Toyota, and Volkswagen cars.” Nightmarish stories in which major cities like Manhattan are brought to their knees by cyber attacks that deliberately crash exploited cars are already written. These aren’t wild fantasies. They could be front page articles on the next decade’s news websites.
So far, the sell for autonomous cars is all upside: transport for the disabled, reduced traffic, fewer automotive fatalities. History shows us that there are unanticipated consequences to all technological advances. All of them. What will they be for cars without an organic brain behind the wheel?