What We Can and Cannot Say About the Tesla Crash
Last year, the National Highway Safety Transportation Association (NHSTA) conducted an investigation of Tesla’s Autopilot Software following a deadly, May 7, Model S. accident in Williston, Florida. The accident – which killed the car’s driver – was the first in a series of high-profile AV accidents, involving a variety of AV manufacturers, which skyrocketed public concern about the safety of autonomous vehicles.
Although the accidents are tragic, it is still premature to draw any conclusions about AV safety.
Statistically speaking, it is not yet possible to make a clear comparison between the safety of autonomous vehicles and human-driven vehicles. As of July 2016, the most autonomous miles driven by any developer — which turns out to be Tesla — was about 1.3 million. According to Susan M. Paddock, a senior statistician at RAND and a professor at the Pardee RAND Graduate School, “this does not come close to the level of driving that is needed to calculate safety rates. Even if autonomous vehicle fleets are driven 10 million miles, one still would not be able to draw statistical conclusions about safety and reliability.”
With so few AV miles driven, experts can’t estimate the impact AVs would have on, for instance, the 2.3 million car-related injuries reported in 2013, of which 32,719 resulted in deaths. That’s one fatality per 100 million miles driven.
According to the NHSTA, human errors such as driving too fast, alcohol impairment, distraction, and fatigue are the cause of more than 90 percent of automobile crashes. While it is safe to say that an autonomous vehicle is not likely to get drunk or sleepy, we are still in the data collection phase of evaluating the overall safety of these autonomous systems.
That said, it would also be premature to draw conclusions that discredit the potential of this autonomous technology.
After all, we’ve been here before. Take a look at the airline industry: in the short history of airline safety, the first great turning point occurred in the 1950s with the introduction of the jet engine, which was far more reliable than the behemoth piston-engine that preceded it. The second turning point followed with advances in sensor technology, computing, and artificial intelligence: like the introduction of GPS, aircraft avoidance systems and ground-proximity alarms in the 1970s and ‘80s.
Today, an aircraft is generally flown by a computer autopilot that tracks its position using motion sensors and dead reckoning, corrected as necessary with GPS. Software systems may even be used to land commercial aircraft. In a recent survey of airline pilots, those operating Boeing 777s on a typical flight reported spending just seven minutes manually piloting their planes; pilots operating Airbus planes spend half that time.
The safety outcomes of commercial airline automation have been tremendous. The year 2015 marks the safest year of airplane travel to date. The chances of dying on any given flight with one of the world’s major airlines are just one in 4.7 million; in any given year, you have a higher chance of getting struck by lightning, at one in 1.9 million. In the 1970s, an average of 68 commercial planes crashed each year. Last year, of the total 33.4 million flights, only 21 crashed.
As aviation safety expert Carl Rochelle puts it, “the most dangerous part of your airline flight is the trip to the airport.”
So how did flying become so safe?
The unfortunate reality is that the industry learned a great deal from its failures, most notably from its crashes. By examining downed plane wreckages and black-box recorders, they engineered solutions to the problems.
In the wake of the deadly Tesla accident, what we know is that the vehicle was on a divided highway with Autopilot engaged when a tractor-trailer took a left-hand turn across the road, perpendicular to the Model S. Neither Autopilot nor the driver seemed to notice the white side of the tractor-trailer against a brightly lit sky. The brake was never applied.
Sadly, in this case, the technology appears to have failed. But let’s not allow unsubstantiated fear about the overall safety of autonomous vehicles dictate how we move forward. Instead, let’s do what we’ve always done: embrace innovation and look for technology-based solutions that captivate our imaginations while ensuring that safety is the number one priority.