Article Single


Autonomous Car

Are self-driving cars really safer than human drivers?

With over 90% of car accidents resulting from human error, the self-driving car seems to be the perfect solution to making roads safer. But autonomous vehicle accidents in recent months raise serious doubts about their reliability.

An autonomous car is a vehicle capable of sensing and functioning without human intervention in its environment. At no time shall a human passenger be required to take control of the vehicle, nor shall a human passenger be required to be present in the vehicle at all. Everywhere a conventional car goes, an autonomous car will go and do everything that an experienced human driver does.

Endowed with inexhaustible forces and a lifespan allowing it to cover billions of kilometres, the autonomous car has undeniable qualities. Advances in artificial intelligence even suggest that it will one day accumulate more experience and knowledge than humans. No worry of fatigue, no alcoholism or lack of attention ...

Yet accidents do happen. In March, an Uber car killed a pedestrian due to a software "error". Indeed, the sensors did detect the pedestrian, but the car did not respond properly - that is, as common sense required. There are multiple sources of errors and possible accidents: poorly programmed software, a road with invisible signage on the ground, weather conditions hampering proper use of the sensors, etc. Human errors are avoided, but others appear to the place. To be effective, artificial intelligence - whose mechanisms are modelled on the human brain - also needs to learn. However, each time an accident occurs, the future of the autonomous car is called into question.

No room for error

Yet, technically, an autonomous car is more secure than a traditional vehicle. Quite simply, machines have no right to make mistakes. Should we join RAND's in-depth research under the speaking title "The Enemy of Good. Estimating the cost of waiting for nearly perfect automated vehicles", according to which the ideal is the enemy of the good in the case of the autonomous car? In debates about the right to run semi and 100% autonomous cars, should we not focus on their ability to improve and learn rather than whether they are reliable enough when marketed? According to the study, introducing autonomous vehicles that are only 10% more reliable than traditional cars could save more lives in the long run than expecting a near-perfect autonomous car.

The answer may lie elsewhere. In fact, the autonomy of the car is gradually being established. No manufacturer is talking about marketing their level 5 autonomy vehicles in the coming weeks. On the other hand, all of them already offer models equipped, more or less, with partial autonomy solutions. Volkswagen’s latest Touareg is a showcase of what is working perfectly well today. Let us quote in particular its "traffic jam assist", thanks to which the car drives… all alone. Or the Nissan Leaf's ProPilot Park, which parks… all by itself. No one would ever think of rejecting this equipment on the pretext that it was not developed a few months ago. Let’s be patient! The 100% autonomous is for tomorrow.