Autonomous Vehicles Might Be 100 Times Worse Than Human Drivers

The progress hype is ongoing; every year autonomous car developers hint, maybe even proclaim, that they are closer to deployment. Whether introducing a cool new concept in Vegas or a boastful speech at an industry event, the promises continue to pile on.

But they aren’t really that close to deployment – far from it.

While advancements are being made, it’s not as if you can look outside your apartment window and eyeball dozens of driverless cars speeding by. It might feel like the roads are being overrun by autonomous vehicles (AVs) if you live in one of the few heavily tested corridors of the world. For most consumers, however, driverless cars are still a fantasy.

And for better or worse, they might remain a fantasy far longer than anticipated. Shane Elwart, deputy-chief engineer of Autonomous Driving Systems at American HAVAL Motor Technologies (a wholly owned US subsidiary of Great Wall Motors), explained why. He compared crash data from the National Highway Traffic Safety Administration to the data from the presumed leader of AV development (which he wouldn’t mention by name to avoid calling out any particular company). The results, while not definitive, were not terribly encouraging.

“This is a rough assessment, but it looks like autonomous vehicles might be 100 times worse than a human driver just by looking at some really simple numbers,” said Elwart. “Maybe AVs are actually better than that, but I think we have to seriously look at the numbers and say, where are we? We can take a risk and go out there. Like Uber, they took a risk and had an incident. I’m not sure if they could have reacted to it in time anyway. Nonetheless, what would be acceptable? That’s one of the key questions.”

When taking that assessment into consideration, Elwart started to question the claims that autonomous vehicles must drive millions or billions of miles to reach deployment. He is not sure that any manufacturer can apply an accurate mileage estimate at this stage in development.

“When I think of driving, we already have something that exists – the human,” said Elwart. “If we’re going to make something that replaces humans, I would think the public would only accept something that’s better.”

Despite all of the players involved in driverless car development, Elwart said he has not seen anyone attempt to answer two key questions: when will development be done, and what does being “done” actually mean?

“I know that’s a really simple statement, but I have to answer that question,” Elwart added. “Maybe AVs are 100 times worse than humans. But if we are able to achieve this, the really great story behind this is, we would be on a journey toward reducing the number of injuries just in the US by over 2 million a year and potentially 45 million globally. And save potentially more than 27,000 lives and over a million globally. There’s a really awesome responsibility here and something that’s really motivating.”