Driverless cars pose unique safety risks that must be mitigated before they hit the roads.
As automobile and technology companies increasingly partner to develop autonomous automobiles, it is important to consider whether such vehicles promise improved safety on the roads. The proponents of self-driving cars correctly note that the technology would eliminate driving under the influence, distracted driving, road rage, and driver errors such as pushing on the wrong pedal or backing into something. These potential benefits, however, are offset by a number of safety risks that are unique to self-driving cars.
To begin with, the technology underpinning driverless cars is mapping – very advanced 3D mapping. The map guiding a self-driving car must be capable of precisely noting landmarks such as lamp posts, specific building characteristics, and street markings. As Christopher Mims, a technology reporter for the Wall Street Journal, and Mary Cummings, a Professor at Duke University, have noted: “If the map is wrong, then the car is going to do something wrong.” Given the amount of construction work done on our roadways and their surroundings every day, and the physical changes such work produces, there are legitimate concerns that driverless cars may be unable to navigate changing roadways effectively.
In addition, insofar as autonomous automobiles rely on lasers and other sensors to navigate and maintain safe distances, precipitation and other forms of inclement weather can substantially undermine sensor performance. It may be that whether a car can be driverless will, like so many things, depend on the weather.
As powerful as the analytics underlying autonomous automobiles may become, they are unlikely to be able to match a human’s ability to observe her surroundings and draw the correct inference from the relevant observations. For example, a human driver may look at an adjacent car and notice the other driver is constantly looking down, which suggests he is reading something on his phone or trying to send a message. The human driver would know that this creates risk and then honk and maneuver her car to a safe position. Would a driverless car observe the distracted behavior? Would it draw the right inference from that behavior? These are critical and, as of now, unanswered questions.
Lastly, as Ralph Nader has noted, driverless cars are inherently vulnerable to hacking and related threats. In a world where the largest data companies are constantly victimized by successful cyber attacks, do we honestly believe that autonomous automobiles will have substantially better defenses? If the answer is “no,” we need to think more than twice before putting these cars on the roads, especially when one of today’s leading terror techniques is to use cars and trucks as weapons.
Billions of dollars have been bet on the successful development of driverless cars, and this bet may pay off. But before we declare this bet to be a winner, we must satisfy ourselves that autonomous automobiles would be as safe as humans driving cars with the latest safety features. We are not anywhere near meeting this safety standard today, and it will be a very long road to get us there. So while we marvel at the pace of technological development, let’s not forget that there are many things we can do today to Drive2Save.