We need more driverless car accidents

A driverless car was involved in a traffic accident on a California city street earlier this year. No one was hurt in the small fender bender, but the accident does signal we are making incredible leaps forward on the road toward driverless cars. It may sound counterintuitive, but this crash shows just how far autonomous technology has come in such a short time.

This wasn’t really an accident in the traditional sense — intentional software changes implemented just weeks earlier were likely a contributing factor. With any luck, we will continue to accelerate real-world experimentation and the possibility of more accidents to come.

The incident was partially caused by a subtle software update, implemented a few weeks prior in all of Google’s autonomous cars, that enabled them to “hug the rightmost side of the lane,” a common social norm that allows other drivers to pass on the left. According to the accident report, the Google autonomous vehicle was shifting within its lane to bypass an obstacle in its path when it made contact with a bus approaching from behind. The car was traveling slower than two miles per hour at the time of impact — the bus, about 15 mph.

The fact that Google was testing this new behavior to hug the right side of the lane shows that the technology has developed beyond simply following the rules of the road, but actually driving more “human like” — in line with the social elements of driving.

Accidents like this are vital learning exercises. Google’s driverless vehicles cover more than 10,000 miles a day, in addition to the three million miles of computer-simulated driving taking place daily. But these real-life tests are crucial.

The only way to fully understand our world is to explore it by taking risks.

Driverless cars won’t just change who or what is behind the wheel — they hold the potential to change where we drive, which in turn can change our commutes, vacation plans and how we connect with family and friends. Driverless cars are positioned to forever change the world in which we live… but before they can do that, they must fundamentally and fully understand that world. And the only way to fully understand our world is to explore it by taking risks.

The history of innovation is built upon pushing the frontier of what has been done before. From Lindbergh’s nonstop solo flight across the Atlantic in 1927 to Chuck Yeager breaking the sound barrier in 1947, we have long been demystifying the unknown in the name of progress. While shifting within the same lane may seem like a minute detail, autonomous car developers, like Lindbergh and Yeager before them, are fundamentally changing transportation.

Subtle changes in technology are what will propel us forward. From air bags to automatic windshield wipers, on-board diagnostics to collision avoidance systems, our cars look vastly different than they did 30 years ago, thanks to incremental innovation over a long period of time.

Innovation builds exponentially, but requires risk. The first seven U.S. astronauts were military test pilots. When they were selected in 1959, no one had come close to leaving the earth’s atmosphere. And at the time, no one knew whether any of these seven men would be successful.

These innovators pushed the frontier of what we knew to be possible. Back then we were pushing for discovery and exploration, but we were also in a space race with our Cold War rival, the Soviet Union. Countless Americans risked their lives, and some of the very first Americans in space gave the ultimate sacrifice, to win that race.

Make no mistake: Today, we find ourselves in an even more important race. More than 1 million people are killed annually in traffic deaths globally, and almost all of these deaths are caused by human error. Autonomous vehicles hold the greatest promise to eradicating one of the most deadly forces on earth, but first we must push the limits.

Autonomous vehicles need to learn to be aggressive to the right degree and in the right ways. Much of this can be determined within computer simulations, but some of it must be determined on open roads, where obstacles are dynamic and complex.

There will be accidents along the way. And from these incidents, we will gain treasure troves of intelligence that will push us further along the innovation curve.

Hopefully, we haven’t seen our last autonomous-vehicle accident; hopefully everyone will see them for what they are if/when they occur: invaluable steps pushing us across uncrossed frontiers.