Uber’s self-driving car fatality should not end autonomous driving

Self-driving technology could still save many lives.

Portrait of Tammy Strobel

Self-driving technology could still save many lives.


On March 18, a self-driving Uber SUV struck Elaine Herzberg. Herzberg, 49 and a mother of two, was pushing her bicycle across a dark street at night. As she emerged from the shadows and in front of the Volvo XC90, it did not attempt to break and ran into her at 61 km/h. Herzberg died of her injuries at the hospital, making her the first pedestrian killed by an autonomous vehicle.

Everyone working in the autonomous vehicle industry saw this coming in some capacity. It was not a matter of if, but when. After all, cars kill people, and 1.3 million people die in road crashes every year.

Herzberg’s death threatens to derail efforts by automakers to push the technology to market. An egregious failure like this casts everyone in a bad light and makes it tempting to paint the entire industry with the same brush.

But that doesn’t mean it should. The incident needs to be investigated and processes need to change. Yet this shouldn’t be cause to call for the complete abandonment of autonomous driving.

There’s a disconnect between how we view accidents caused by people and those caused by machines. Staggering numbers of people die in road accidents daily, and yet you don’t hear people talking about kicking driving to the curb.

We accept the risk every time we get behind the wheel. But when AI is involved, we freak out. And that’s understandable. Human error is unpredictable, and it stands to reason that we can do more to ensure that AI performs than we can to stop people from making mistakes.

This also means accepting that AI is also fallible. Ideally, autonomous driving would have a zero failure rate, but that’s a huge ask of any technology. Imagine if we went from a hundred deaths a day caused by humans to 10 caused by AI. Those 90 lives saved by self-driving cars could still be considered a success.

Progress is continually being made on self-driving cars. At GTC 2018, NVIDIA announced its Drive Constellation system. Drive Constellation tests driving software in simulated real-world driving environments. This means you can test autonomous vehicles in multiple scenarios without endangering anyone.

Looking forward, autonomous vehicles could save many lives. To get cold feet now would seem almost foolish.

Imagine if we went from a hundred deaths a day caused by humans to 10 caused by AI.

By Koh Wanzi