Slate is Cherry Picking the Data on Self-Driving Cars

On July 8, 2020, Elon Musk announced that Tesla was “very close” to creating completely autonomous cars and predicted that they would have the “basic functionality” ready by the end of the year. Fully autonomous vehicles aren’t yet available for consumer purchase, but they are being tested on public roads. Waymo (formerly Google’s brand of self-driving cars) is even offering fully driverless taxi service in the Phoenix metro area. But some commentators, such as Slate’s David Zipper, think driverless technology is moving onto public roadways too fast, too soon. They cite YouTube videos of Tesla drivers leaving the driver’s seat while using autopilot and a few high profile crashes to argue that federal regulators should step in and create stricter safety and testing protocols before more lives are lost to the imperfect technology. Others are arguing for an outright ban on the driverless tech. But these critics are wrong; though driverless tech is still in its infancy, it can and will save millions of lives, and the sooner it hits our roadways, the better.

It’s hard to overstate how bad humans are at driving. The National Highway Traffic Safety Administration (NHTSA) reports that 36,560 people were killed in car accidents in the U.S. in 2018, including over 10,000 deaths involving alcohol-impaired driving. On average, human error causes over 90% of car crashes every year. Motor vehicles also accounted for 43% of law enforcement officer deaths in the line of duty from 2006 to 2019, including both crashes and officers struck by vehicles.

Driverless cars don’t get drunk, drowsy, or distracted; they have infinitely faster reaction times and they can see every side of the vehicle at once. And though many argue against using public roadways as testing labs, the more time that automated vehicles spend on the road, the faster the technology will improve. If these vehicles can be shown to be even 1% safer than human drivers right now, we should get them on the road as quickly as possible.

So how does today’s driverless tech compare to human drivers? There is no single source of data (the NHTSA doesn’t track accidents that specifically involve autonomous vehicles), so direct comparisons are tricky. To estimate crash rates, we must rely on what little state-imposed tracking there is.

California is currently the only state tracking crash data for autonomous vehicles, and Wired reports that in 2019, autonomous vehicles in California logged 2.9 million miles in driverless mode on public roads. The California DMV reports 105 collisions involving these cars during that period, and the vehicle was operating in autonomous mode in 47 of those. Only three of these 47 reports state that police were called; 17 don’t specify. If police were called in all 17 of the unspecified cases, that’s 20 police-reported collisions involving an autonomous vehicle operating in driverless mode.

The reports don’t assign fault, so can we know how many of those 20 collisions were caused by the driverless car? A 2019 analysis of 113 crashes involving autonomous vehicles determined that only 13% were the fault of the autonomous vehicle. Assuming a similar percentage of the 20 autonomous, police-reported collisions in California in 2019 were the fault of the driverless car, that’s about 2.6 collisions blamed on the car, or one per 1.1 million miles driven, making them over twice as safe as human-driven vehicles when compared to NHTSA data.

Rather than stricter regulation as many commentators are demanding, every state should implement California’s requirement of separate, detailed reports for autonomous vehicle-involved collisions. We don’t need stricter laws; we need data to improve these cars’ abilities and monitor their impact on roads and other drivers. Currently data scientists have limited sample sizes to work with, but every driverless car on the road has the potential to be a data point. More data and better tracking of those data are the most effective way to improve safety and speed up innovation—especially since available evidence shows that even the primitive driverless cars on the road today are already safer than human drivers.

Imagine streets where traffic is a thing of the past, where every vehicle can talk directly to every other and cars can move in harmony like a flock of birds. Stoplights would become unnecessary; congestion would disappear; emissions produced by transportation (whether from the cars themselves or from electricity produced to power them) would drop exponentially.

Imagine roadways where every car moves aside immediately and automatically for police and EMS vehicles, then efficiently closes the gap behind them like zipping up a jacket, cutting response times dramatically. Self-driving cars that are incapable of breaking traffic laws would also remove the need for traffic stops, reducing contact between police and Black and brown drivers as well as the risk to cops of being struck by a moving vehicle.

Will driverless cars someday put truck drivers, taxi drivers, Uber and Lyft drivers, and even delivery drivers out of work? Yes. But this concern isn’t unique to self-driving cars; blue collar workers all over the world are already being replaced by machines. A 2019 report estimated that 20 million manufacturing jobs would be lost to automation by 2030. The robot revolution is already happening in our factories and will eventually happen on our roads, and it’s the job of governments not to try to stop it, but to help our societies adapt to it.

Self-driving cars have the potential to save tens of thousands of lives annually in the U.S. alone. Lawmakers and the people who vote for them will help decide how quickly that happens. Let’s not allow a few high profile cases to obscure the reality that they are already far safer than human drivers, and let’s not hold back an industry that can create so much fundamentally positive change.