91av

DRIVING in San Francisco is like watching a robot uprising in the making. Last week, I was crossing an intersection, turning from one major downtown thoroughfare onto another, and realised that my vehicle was surrounded by four experimental self-driving cars.

They weren’t exactly difficult to spot, their bumpers encrusted with radar and other sensors, their roofs topped by enormous lidar rigs, including bulbous, whirling cameras shaped like turrets.

These kinds of cars aren’t uncommon in San Francisco and Silicon Valley to the south of the city. But their numbers are growing fast, bringing many questions with them. The California Department of Motor Vehicles reported that there were 900 self-driving vehicles registered in the state in November 2021. As of December 2021, the number had jumped by a huge 55 per cent to 1400.

Most of them are Waymo vehicles, fitted out by Alphabet, Google’s parent company. And most cruise the streets of the San Francisco Bay Area using sensors to gather data on road conditions. The idea is that the data will be used to improve the algorithms that guide these cars through every possible hazard and traffic snarl.

Here in California, safety rules require that self-driving vehicles carrying people on public roads must have someone behind the wheel, even if they don’t do much. But in Arizona, self-driving taxis are ferrying people around with nobody in the driver’s seat.

This has led to some hilarity, as in the case last year when a passenger recorded a video of his Waymo taxi becoming confused by orange road cones. First the poor car simply stopped, blocking the road. Then, when roadside assistance from Waymo arrived to intervene, it tried to escape twice, pulling away from the person sent to help.

In that case, nobody was hurt and there was no crash. But when self-driving cars go wrong the results can be devastating. In 2018, a woman named Elaine Herzberg was pushing a bicycle across a road in Arizona when she was struck by an Uber car in self-driving mode, albeit with a safety driver in it. Herzberg died of her injuries.

Because of incidents like this, truly self-driving cars are unlikely to advance beyond limited taxi-like services for decades. Even Waymo, which agreed a deal in December with automaker Geely to produce a fleet of self-driving vans designed for ride-share customers, has been vague about the launch date. Promotional materials say the vans will arrive “in the years to come”.

“First the poor car simply stopped. Then, when roadside assistance arrived, it tried to escape twice”

Other companies have simply given up. Uber announced in 2020 that it was scrapping its long-touted plan to develop its own driverless vehicles.

So if all those sensor-clad cars on the streets of San Francisco aren’t about to start whisking me off to the grocery store, what are they good for?

Obviously Waymo is hoping that there is still a market for future ride-share firms, one with fewer pesky human drivers that sometimes demand pay rises. If they can just get enough data, maybe their cars will never get caught out by traffic cones again.

In addition, a number of companies are developing self-driving delivery trucks that can use simple, predetermined freeway routes to deliver goods hundreds of miles away. Because these trucks stick to major roads, they may be easier to program.

As economists like , Berkeley, are fond of pointing out, automation giveth and taketh away. If these companies are right, there will be fewer human drivers, but there will also be new jobs for people to develop and maintain self-driving cars and for workers to keep the roads optimised for the AI fleet.

There is an even weirder possibility, however: self-driving cars really could lead the robot uprising. For decades, roboticists have argued that human-equivalent consciousness can only evolve if it is “embodied”. A brain in a box will never think like a person. But a brain in a body? That could work.

After all, nothing is more human than spending your first few years of life learning not to run into things and avoiding danger. This is one reason why, for example, researchers at Stanford University are putting AIs inside simulated “bodies” to help them get smarter.

It could turn out that a giant metal body on wheels is the missing link that we need to develop artificial general intelligence. Maybe we saw the first inkling of that future in the unfortunate adventures of that Waymo taxi that rolled away from the person who attempted to get it back to work.

It makes you wonder. What are the self-driving car firms of the world going to do when their vehicles go on strike?