Self-driving zealots should not put cars above people

It was a request that raised eyebrows, even in the Tesla fan community: “Is there anyone in the Bay Area with a child who can run in front of my car on Full Self-Driving Beta to make a point? I promise I won't run them over.”

The request on Twitter came from a Tesla owner angered by a video published online. A group called the Dawn Project had put a Tesla onto a test track, and then switched it into what the company calls “full self-driving” mode (FSD). It was then set on a collision course with a static mannequin of a small child. The car hit the mannequin every time. 

Even with seven seconds notice, “an attentive human driver cannot fail to notice a stationary child in the middle of a clear roadway,” the study noted. “Tesla’s FSD software does fail this simple task, repeatedly and with deadly results.” 

The angry Tesla fan was convinced the Dawn Project was wrong and made the request for a sacrificial child to recreate the experiment at home, only with a live guinea pig.

Dawn is part of a campaign by a wealthy software entrepreneur, Dan O’Dowd, who thinks Tesla’s self-driving feature is lethal. O’Dowd isn’t exactly impartial. 

Customers for his software include drone manufacturers and Tesla rival BMW. But its videos serve as a vivid demonstration of how far away we are from truly safe autonomous driving

After years in which the auto industry has marked its own homework, regulators are beginning to agree. 

The US traffic safety agency, the NHTSA, is not only investigating why Teslas keep hitting things, but also why the vehicles brake suddenly when there is no danger at all. 

Since 2016 the agency has investigated 19 fatal accidents involving vehicles fitted with Tesla’s automated driving software

Also taking note is Ralph Nader, the renowned consumer campaigner and serial Presidential candidate, who at 88 is still revered by American baby boomers for his efforts to make cars safer.

None of this was in the script. A decade ago, the auto utopians argued that since human drivers are involved in almost all traffic accidents, once you remove them, we should see far fewer accidents — or perhaps none at all. 

Cities could be redesigned around this technological miracle; people may cease to want to own a vehicle, since they could just summon one, and it would drive itself over to you. And they assured us that this miracle was just around the corner. 

Back in 2011, GM reckoned human drivers would be obsolete by 2021. Musk declared that it was “basically a solved problem” back in 2016. Years roll by and the prophecies remain unfulfilled.

What happened? Surely, since autonomous trains are known to be safe and reliable, then autonomous cars can’t be that much different? Actually, they are. 

On rails, the risk environment can be reduced to a handful of known factors: pedestrians don’t stray into the path of the train, for example, and that path is very rarely shared with other vehicles. 

But a busy road such as the North Circular is a symphony of chaos. In their arrogance and naivety, the auto engineers underestimated the challenge. Despite the addition of more sensors and more powerful AI, it isn’t getting better.

The Holy Grail of the engineers remains full autonomous driving, which is defined in the industry jargon as “Level 4” and “Level 5”. These are both “hands-off” and “mind-off” modes. 

But with success far away and receding, what the industry has done instead is move the goalposts. An example is Tesla marketing, which is really automated lane assistance labelled as “full self driving”. 

Musk’s bluster was contradicted by his engineering managers, and was reportedly a factor in prompting the head of Autopilot at Tesla to quit.

Tesla’s own website acknowledges that “the currently enabled features do not make the vehicle autonomous.” Tell that to the Tesla owners who post videos of themselves on YouTube taking a nap as their car speeds down the freeway.

This practice of exaggeration, of linguistic inflation, has become known as “autonowashing”, and it matters, because it deludes owners into thinking the car is much cleverer than it really is, and so they pay even less attention to the road. 

In January, the Law Commission of England and Wales recommended two new offences to restrict the use of misleading terms, and prohibit practices that induce complacency.

The real puzzle for me is why anyone thought this was a desirable goal at all. Other than urban planners, nobody really wants self-driving cars. We’d rather have safer or more comfortable cars. 

It’s very telling that the biggest boosters have been public officials, willing to sacrifice public safety for kudos, so they can boast about hosting a piece of the future in their municipality. Pure civic chauvinism, in other words. 

I have another theory, too: if we think of technology as a religion, then true believers require regular miracles. These vast, fruitless sinkholes of capital are now sustained entirely by faith. 

I think a significant correction is overdue, and given that layoffs have begun sweeping through the sector — at Cruise (GM), Argo (Ford), and even Tesla itself — it’s a correction that may already have started. And I wonder who will remain solvent once it’s over.