The day you catch a robot taxi is still a long way off


The day you catch a robot taxi is still a long way off

Google subsidiary Waymo has launched an automated taxi service but it's not ready to ditch the safety man yet

James Titcomb

In 2012, Sergey Brin made a bold prediction. Within five years, Google’s co-founder claimed, the public would be able to ride in cars that could drive themselves. Not everybody took Brin seriously at the time, and not only because he was going through a phase of wearing the peculiar Google Glass headset at every public appearance.
Google had been testing driverless cars for just three years, and few knew how much progress the company was making. However, it turns out that Brin’s prediction was only off by a year.
Last week, Waymo, a subsidiary of the search company’s Alphabet parent that was spun off from Google two years ago, launched a commercial driverless taxi service to the public. In an area of Phoenix, Arizona, people can now open an app on their smartphone, select a destination, and within a few minutes have a driverless car arrive to take them there. The future, it seems, has arrived.
Well, not quite. Several important caveats apply.
Only a select group of riders – 400 individuals – are able to take part. The service applies to just a small part of the city at first. And the driverless cars will still have a safety driver, to take control should the software driving the car put its passengers in danger. The last point here is the most glaring, because Waymo had until recently claimed that by this point, human supervisors – a safety measure that has been in place since it first began tests almost a decade ago – would no longer be needed.
But like an anxious parent who resists removing the stabilisers from their child’s bicycle, not yet confident enough that they can ride safely without them, Google is reluctant to do without its safety drivers. It is not the most encouraging sign.
This, by the way, is all in an area in which Waymo has been conducting tests for 18 months, and has mapped in painstaking detail, which is currently a requirement for driverless cars. It will take many years to map enough of the world in such detail to remove human driving from the equation.
The strings attached to Waymo’s big moment show that, however much progress driverless car technology has made in recent years, it is still a long way, probably further than most people believe, from the point at which an unsupervised vehicle can take a passenger to any point of their choosing. In the past year or so most experts in the field have pushed back their expectations for the day on which “level 5” automation, the term for a fully independent driverless vehicle, arrives. Suddenly, Brin’s prediction is not looking so good.
Driverless car optimists point out that the technology has come a long way in the past few years. The number of “disengagements” – the times an emergency safety driver must take control of a car – has fallen sharply in recent years. In California, which makes these statistics public, Waymo reported 63 disengagements over more than 567,000km of testing – a rate of about 0.18 every 1,600km. Two years earlier, the company recorded 0.8 disengagements per 1,600km. That is undoubted technical progress. But it does not tell us the point at which driverless cars are “ready”.
The consensus is that even as autonomous technology gets better by the month, the point it must reach seems to be getting further away at an equal rate.
Waymo, by the way, is seen as the leader in driverless technology. Others are faring worse. Uber suspended its driverless car programme in March when one of its vehicles hit and killed a pedestrian. The company is planning to resume tests in the coming weeks, but is doing so only on a 1.6km loop between two of its offices in Pittsburgh. The ride-hailing app once sold investors on the idea that it would become wildly profitable by removing its biggest expense – drivers – from its cars, but no longer makes such claims, even as it prepares to go public in 2019.
Tesla, too, has cut back its bold predictions. The electric carmaker recently stopped offering buyers the option for their car to become “fully self-driving” through a future software update, apparently because such a feature is not as close as it had once hoped. Earlier this year Tesla settled a class-action lawsuit brought by a group of customers who had been promised this upgrade years ago.
This is not to say that driverless cars are never going to happen. At some point, perhaps 10 years away but probably more, they will. The promised benefits – lower congestion, more free time, fewer deaths – are too great to not pursue. In the meantime, semi-autonomous features, such as cars that park themselves and drive unassisted on motorways, will become increasingly ubiquitous. But the day when most of us can order a robot taxi that can take us wherever we please is undeniably further away than was hoped for until recently.
For years, proponents of autonomous cars have insisted that the technology is nearly ready, and that the real barriers to a driverless future are other factors, such as regulation, public perception and cost. The reality may be the opposite: We may be waiting for driverless cars to be good enough long after we are ready for them.
– © The Daily Telegraph

This article is reserved for Times Select subscribers.
A subscription gives you full digital access to all Times Select content.

Times Select

Already subscribed? Simply sign in below.

Questions or problems?
Email or call 0860 52 52 00.