There was no screech of tyres, no last minute change of direction. As the grim video footage now visible on You Tube makes clear, the speeding Uber taxi didn’t even brake before mowing down 49-year-old Elaine Herzberg last weekend as she pushed her bike across a dual carriageway in Tempe, near Phoenix, Arizona.
It was exactly the sort of accident that was supposed never to happen. For this was a ‘safe’ self-driving car, a vision of the future armed with a battery of sensors and automatic brakes for just such an emergency.
For Ms Herzberg, however, the new technology proved fatal. Rushed to hospital, she later died of her injuries – and so became the first pedestrian casualty of the driver-free era.
If the crash was a catastrophe for the victim and her family, it is disastrous, too, for the billion-dollar proponents of the fledgling technology.
Uber has suspended its experimental autonomous taxi fleets in Phoenix, Pittsburgh, San Francisco and Toronto pending multiple enquiries. And the questions will go well beyond the events of Ms Herzberg’s tragic death last Sunday.
Because, despite the huge investment and many promises, autonomous vehicles remain adrift in an ocean of practical problems, moral doubts and, as we now know, fatal consequences. As a transport writer and analyst for more than 25 years, I believe the promise of a world dominated by self-driving cars is a mirage, and a deadly one at that.
Here, I examine what’s really going on behind the glossy spin churned out by the industry – and ask some very worrying questions.
The future beckons
It is a deeply seductive vision: cars so intelligent you can put your feet up, sit back and relax on even the most stressful journey. Traffic jams and crashes will become a distant memory, we are told, along with expensive driving lessons and irksome parking tickets. Moreover, this future is just around the corner, with some authoritative voices claiming 95 per cent of city car journeys will made in self-driving vehicles by 2030.
No wonder technology giants and car manufacturers are licking their lips at the prospect of a fully robotic transport system – and profits on a vast scale, naturally.
Google alone has thrown billions at the technology, testing out a sci-fi world of driverless electric pods at its Googleplex headquarters in Mountainview, California.
It is a crowded field, with car manufacturers and tech giants including Intel, Uber and Amazon investing, too. More prosaically, Jaguar Land Rover is running experiments with self-driving cars on public roads in the Midlands, while the Government is testing small, self-driving vehicles in South London. Chancellor Philip Hammond has gone so far as to suggest that fully autonomous cars will be on our roads by 2021.
The nightmare failings
Last week’s crash was supposed to be impossible, yet many – myself included – believe it was all-too predictable. Indeed the accident has exposed something the tech giants would rather keep to themselves: that the self-driving technology isn’t yet ready, and might never be.
What, for example, do two driverless cars do when they meet on a typical country lane? How do they cope with unexpected roadworks, or a sudden flood – persistent features of British motoring? How can you get the car to understand the difference between a policeman waving a car on or trying to get it to stop?
These are the sorts of calculations that are intuitively and instantly worked out by human drivers on a daily basis, but which remain baffling to a computerised control system. And then there is the question still haunting the citizens of Tempe, Arizona: what is a robotic car to do about pedestrians who venture into the road? Autonomous cars must surely be programmed to stop – something that signally failed to happen in Arizona – but that in turn poses enormous difficulties in busy cities where pedestrians routinely step out from the pavement.
Other worries follow on from that. If any driverless car can be stopped by something as simple as a person standing in front of it, what will prevent criminals from blocking a car and then attacking the people inside? This is an especially pertinent question in the South African context.
Bear in mind, too, that with computers comes hacking. Criminals already have the technology to open and drive away expensive cars, despite sophisticated electronic security systems. Driverless cars would be even more vulnerable.
The legal minefield
Even if the technology could be conquered, there are multiple legal, regulatory and economic conundrums facing autonomous cars.
The question of who is liable for an accident could soon be tested in the case of Ms Herzberg, for example, whose family could be in line for millions in compensation. Is it the manufacturer, the owner or the operator who must pay? The Herzberg case alone could provide lawyers with years of lucrative work. Imagine the outcome if driverless cars and, presumably, driverless accidents, become more widespread.
Nor is there any clear business model. None of the firms has said how much these cars will cost, and none is yet on the market. Nor is there any coherent plan as to how conventional vehicles will be replaced by autonomous ones, presumably after a long period when the two will run side by side, with all the dangers that will create.
The dozy reality
For all the billions spent on research, the manufacturers have come across one major and very human failing: we fall asleep, sometimes at the wheel. Companies such as Google and Jaguar had hoped they could adapt existing cars step by step, gradually removing the driver as the technology advanced and individual problems were solved. But that incremental approach has proved impossible because scientists have found that when drivers are told to supervise cars instead of driving them, they soon get bored and doze off.
This is why Ford abandoned its attempts to develop partially autonomous cars: the ‘driver’ nodded off. Then, when it tried putting a second ‘driver’ alongside the first, they both fell asleep.
As Chris Urmson, the former head of Google’s driverless car project, has admitted: "Human drivers can’t always be trusted to dip in and out of the task of driving when the car is encouraging them to sit back and relax."
Last weekend brought telling proof. The human ‘operator’ of the Volvo that killed Ms Herzberg wasn’t even looking at the road ahead. Only as the crunching impact happens can she be seen gaping open-mouthed in alarm.
As a result of our all-too-human failings, then, the manufacturers have been obliged to leapfrog straight into what is known as Level 4 in the industry: a fully robotic car that can deal with all the challenges normally faced by a human driver.
The cars must drive by themselves in all conditions, coping with heavy rain, snow, icy roads, flooding, dirt roads, roadworks and potholes, while avoiding other cars, pedestrians, cyclists, motorcyclists, children, dogs, wild animals, debris and so on.
Except that they cannot, of course, which is why industry promises have morphed from near guarantees into vague predictions.
Volvo, which in 2013 suggested we would have fully autonomous cars by 2017, has now said it will take at least another four years to develop them. A report by the London Assembly suggested the 2030s were more likely. I am not convinced they will happen at all.
The growing doubts
The Arizona crash has led to the first serious outbreak of bad publicity for the fast-expanding industry, but it comes as no surprise to those who’ve looked closely at the new wonder machines. Take, for example, the technology reporters who went for a ride in Pittsburgh, Pennsylvania. Taylor Soper and John Cook, who write for GeekWire website, tried out a driverless Uber taxi, a Volvo.
It will be some time before Soper and Cook get into another one. In fact, their experience of ‘fly-by-wire’ driving ended when the car hit a pothole, causing it to revert to manual mode.
Not that it was much of a ride, in any case. Soper said the journey felt as if he were being driven by a 16-year-old novice as the car stopped and started at random, jolting its occupants when it sensed non-existent danger or failed to negotiate mundane hazards such as a car ahead being reversed carefully into a parking space.
"Most of the excitement I had for these futuristic vehicles has been replaced by wariness," he wrote. "There’s no way I’m letting a robot drive me around without human backup, at least not until the technology improves."
What about the rest of us?
Naturally, the promoters of driverless cars have a clear set of priorities - and other road users come second. It has been suggested, for example, that cyclists should wear beacons so they can be detected by autonomous vehicles. How long will it be before they say pedestrians must be microchipped like dogs?
Governments, meanwhile, seem dazzled. As things stand, they are hell-bent on creating a brave new world without assessing the downsides.
So, for the moment, we should keep a tight hold of our car keys. There is no such thing as a driverless car capable of, say, taking someone to work, then returning to whisk the children to school. The driverless vision of Google, Uber et al remains as futuristic as it has ever been.
If the death of Ms Herzberg serves any purpose, it is to show that driverless technology is nowhere near ready. And might never be.
Christian Wolmar’s book, Driverless Cars: On A Road To Nowhere, is published by London Partnership Publishing.