an article recently published on the MIT Technology Review, author Andrew Rosenblum sat down with the National Transportation Safety Board (NTSB) Acting Chairman Christopher Hart to discuss self-driving cars.
Hart's biography on the NTSB website, he's had a long career in transportation safety, dating back to 1990, including a brief stint at the FAA. Hart holds a law degree from Harvard University and Master's and Bachelor's degrees in Aerospace Engineering from Princeton University. Hart is clearly experienced and knowledgeable.
CHRISTOPHER A HART
Hart believes "his agency's experience investigating accidents involving autopilot systems used in trains and planes suggests that humans can’t be fully removed from control." He believes humans will have to act as "co-pilots" in some capacity for many years to come. This opinion, from the master of transportation regulation, is in complete contrast to the opinions of autonomous industry heavyweights Tesla, Uber, Google and most major automakers. Autonomous vehicle pioneers believe we'll be regularly riding in fully-autonomous vehicles within years, without having to co-pilot the vehicle.
Hart is optimistic that self-driving cars will cut into the auto accident death toll, and that tens of thousands of lives will be saved; however, he is "not confident that we will ever reach [fully-autonomous cars]. I don’t see the ideal of complete automation coming anytime soon."
He worries about complacency and a lack of skills if a human co-pilot will be required. He also says that "Some people just like to drive." Others just won't trust autonomation. He also believes that software developers, even as a collective, won’t ever be smart enough to anticipate every possible scenario.
Hart cites an investigation that found a maintenance problem was to blame for an accident involving a people-mover at an airport. "Even if you eliminate the operator, you've still got human error from the people who designed it, people who built it, people who maintain it."
We think that Hart's response discounts two very important elements of autonomous vehicle development being artificial intelligence, and the standard to which he holds autonomous vehicles. Artificial intelligence, or machine learning, means that software developers won’t ever have to be smart enough to anticipate every possible scenario. A network of autonomous vehicles can learn and adapt over time. Autonomous vehicles shouldn't have to be ready for every possible scenario. Hubris won't be a factor. Autonomous vehicles will know how to stop safely and ask for human input if they encounter the unknown.
What do you think? We encourage you to leave your thoughts and comments below.