According to several surveys conducted over time in various countries on the cause of traffic accidents, the driver is to blame for over 90% of occurrences. Speeding, disobeying traffic signs, using mobile phones, drunk-driving, sleep-deprived driving are the main offences wrecking our roads. The large-scale deployment of a fleet of computer-driven ground vehicles will alleviate these factors. Clearly, accidents will still happen, however in significantly lower rates, and that will directly impact the insurance sector.
By nature, we are less tolerant of mistakes made by machines than of those made by people. After all, machines exist with the sole purpose of tending to our every need, working efficiently and silently. Accidents involving autonomous cars, especially those taking place during the transitional period facing us over the coming decades, will be thoroughly scrutinized and discussed. Driving is an extremely complex activity for a machine, and the revolution of self-driving cars is only made possible by advances in techniques associated with Artificial Intelligence, Robotics, sensors, communications, and processing capacity.
In February 2015, a fully autonomous car in development by Google collided (at an extremely low speed) with a bus — the first accident after almost 2 million miles driven which was not caused by human error. Three months later, a Tesla in autopilot mode collided with a vehicle while crossing a Florida highway, killing its driver. According to Tesla, the accident was caused when the car’s sensor system failed to detect another vehicle as it blended with the background: the side of the vehicle was white, and the day was extremely bright.
Imagine an autonomous car perceiving a pedestrian illegally crossing the street. This car will follow instructions programmed in its memory with the purpose of saving lives. What if it is impossible to protect all lives involved? What if the system must choose between the life of the pedestrian and the lives of the passengers, considering a route detour might cause an even worse accident? The complexity of the algorithms controlling autonomous vehicles is tremendous as they reflect real-world scenarios.
And how will insurance policies be adapted to provide coverage for these cars? In the event of an accident, is the car manufacturer, the loaded software, the set of sensors or the car owner to blame? In October 2015, Volvo declared that, for insurance purposes, any incident occurring with one of its vehicles in autopilot mode will be their responsibility. What if the problem is caused by connectivity issues regarding the communication amongst vehicles? What if a hacker breaks into the system? Note that everything online is subject to intrusions and alterations.
As previously stated, first indications suggest machines are much better than humans per mile driven. However, we have been driving for more than a century, whilst self-driving cars are at an early stage.
Next week, we will discuss some of the prospects for the future of the insurance sector in a world in which accidents will be less regular, but much more complex. See you then.