[:en]Car companies and internet giants are all betting big that self-driving cars will be the wave of the future and they are racing to own as big a share of that future as possible. The relative success of self-driving cars, and the degree to which the motoring public embraces them, will be an interesting case to see how willing people are to trust algorithms.
The public’s faith in machine learning techniques has never been put to the test quite like it will be by the emergence of self-driving cars, which are reliant on the technology.
Of all the companies trying to get in on this modern-day gold rush, each is reliant on machine learning algorithms. Google claims to have unleashed its cars for more than 1.5 million miles of driving. That is not just testing. The point of all those miles is to refine the machine learning algorithms that underlie the decisions the cars make. The more miles cars drive themselves, the more experience they earn. The algorithms collect data on everything including weather conditions, human actions and objects surrounding the cars. They use this data to predict what objects will do and decide how to react. Just like any other deep-learning exercise, the more data the better.
Now we will see if the general public believes these algorithms will make the right decisions. Ride-hailing service Uber has said that it soon will start offering trips in self-driving cars to passengers in Pittsburgh. These trips will still have humans in the driver’s seats to take the wheel if need be, but that will not be an option in the future that is coming.
This basically amounts to a big bet on the public’s trust of machine learning algorithms to make decisions. After all, there is no greater sign of trust than putting your life in someone’s (or something’s) hands. But there are some reasons to be dubious on this point. More than 37,000 people die in car crashes in the U.S. every year. But if you ask the average driver, they’d probably say driving is more or less safe. It is generally believed that self-driving cars will drastically reduce the number of traffic fatalities, but to the family who loses a loved one in a freak crash, self-driving cars will be unsafe. People think in anecdotes, not statistics: it will be worth keeping an eye on whether news about the benefits of autonomous vehicles outweighs the stories about mishaps. This could play an important role in the public’s willingness to accept self-driving cars and the decisions they make underwritten by machine learning.