Self-driving cars – or Autos, as I hope we’ll call them – passed several important milestones in 2018, and they will pass several more in 2019. The big one came at the end of the year, on 5th December: Google’s Autos spin-out Waymo launched the world’s first commercial self-driving taxi service, open to citizens in Phoenix, Arizona, who are not employees of the company, and not bound by confidentiality agreements.
This service, branded Waymo One, was an extension of the company’s EasyRider programme, which was launched back in April. In that programme, selected members of the public who were willing to sign non-disclosure agreements (NDAs) got free rides in cars where sometimes no-one sat up front: no driver, no supervising engineer. There is much debate about how often the cars in both these programmes run with the front seats empty. Google and Waymo won’t say, but the answer seems to be sometimes, but not often. Some people argue this means that self-driving cars won’t be ready for prime time for years to come. Others see it as commendable caution.
Waymo is the clear front-runner in this business. In October it announced that its test cars had driven 10 million miles, and they have not been the unambiguous cause of a single accident. In simulations, they drive that many miles every single day.
General Motors, America’s biggest car maker by volume, is determined not to lag far behind, and has said for some time that it will launch a fleet of self-driving taxis during 2019. In October it announced a $2.75bn JV with Honda in Cruise, its self-driving car unit, which added to the earlier $2.25bn investment by Softbank to bring the valuation of Cruise to $14bn,i which is almost half the parent company’s equity value.
The rest of America’s car industry is also in hot pursuit, especially its newest and most valuable participant, Tesla Motors, which is pursuing the contrarian strategy of offering more and more driver assistance rather than jumping straight to full automation.
Autos are still expensive, not least because production volumes of their LIDAR sensors are still low. So for some years to come, these vehicles will probably only be sold to commercial fleets, especially taxis and trucks. Unless, of course, Tesla’s Elon Musk is proved right, and Autos can operate solely with cameras, and don’t need LIDAR. So far he’s in a small minority, but his contrarian views have been vindicated before. Even if Musk is wrong, city dwellers in particular may well stop buying cars and start using Auto taxis. In which case, how long would the switch take? A famous pair of photographs taken on the same New York street on the same day in 1900 and 1913 shows that it took just 13 years to effect a complete swap in that city from horse-drawn carriages to automobiles. The switch took longer in rural areas of the US, and much longer again in less developed countries.
In short, anyone who thinks that self-driving vehicles will not be in widespread use by the mid-2020s is probably in for a shock.
The US is in the vanguard of the Autos revolution, but other countries are keen to catch up. Both the UK government and London’s leading private hire company (Addison Lee) have stated their intention to have Autos operating in London by 2021. Driving in London is a whole different proposition to driving in Phoenix, so this two-year delay does not denote a lack of ambition.
But as usual in AI, it is China which is most likely to catch the US if there is a race to deploy self-driving technology. Baidu, often described as China’s Google, is the leader so far, with more than 100 partners involved in its Apollo project, including car manufacturers like Ford and Hyundai, and technology providers. The Chinese government is keeping close tabs on these developments, not least in obliging foreign companies to source their maps from Chinese companies.
Are we ready for the arrival of Autos? Can our infrastructures cope? The belief that Autos require modifications to our road infrastructure is a misapprehension. Waymo’s cars don’t need smart lane dividers, special traffic light telematics, or dedicated local area networks. They drive on ordinary roads, just like you and me. No doubt Autos will lead to our cities and towns becoming smarter and more intelligible, but they don’t require it to get started.
What about resistance? Will there be road rage against the machines? The most tragic thing to happen in the self-driving car industry this year was also perhaps the most revealing. In April, an Uber Auto ran over and killed a woman walking a bicycle across a busy road. There is still disagreement about what caused the accident, and Uber stopped its self-driving test programme immediately. But the most interesting thing is that no other company followed suit – and there are over 40 companies trialling self-driving cars in the US alone. Despite this, and despite blanket press coverage, there was no popular protest against Autos. It seems that people have already “discounted” the arrival of Autos: it’s a done deal.
Even if the arrival of Autos is a done deal for society as a whole, there may well be pockets of resistance. On a low level, this will come from petrol heads who find themselves banned from more and more roads because they are much more dangerous drivers than machines. Eventually they will only be allowed to drive on designated racetracks, after signing detailed indemnifications. We should welcome this, not resist it: right now, we kill 1.2 million people around the world each year by running them over, and we maim another 50 million. We are sending humans to do a machine’s job, and there is a holocaust taking place on our roads. We should hurry to embrace Autos. And anyone tempted to vandalise Autos will quickly find that they are bristling with cameras: if people start spray-painting their LIDARS to disable them, they will find themselves on the wrong end of a criminal prosecution.
But there is another form of resistance which may not be so easy to assuage. In June, I gave a talk about AI to a room full of senior US police officers – just outside Phoenix, Arizona, appropriately enough. When I argued that a million Americans who currently earn a reasonable living driving trucks are going to be out of a job fairly soon because the economics of truck driving is going to flip, there was an audible gulp in the hall. They didn’t need me to point out that many of these people have guns.
One of the most significant impacts of Autos may well be to play the role of the canary in the coal mine: they could alert people to the likelihood that technological unemployment is coming – not now, and not in five years, but in a generation. If it is coming, we had better have a plan for how to cope. Otherwise there could be a panic which makes the current wave of populism look mild. At the moment we have no plan, and we’re not even thinking about developing a plan because so many influential people are saying that it cannot happen. They might be right to say that it will not happen. But to say that it cannot happen is dangerous complacency.
So what of 2019? Assuming success in Phoenix, Google is likely to roll out its pilot to other US cities – we could maybe see a dozen of them start during 2019. GM will be anxious not be seen as lagging, and no doubt Tesla will make startling announcements followed by almost-as-startling achievements. I’ll be surprised if there aren’t some significant pilots in China by the end of 2019 as well. And who knows, maybe all this will spur Europe into getting more serious about AI in general. Here’s hoping.
This article was first published by Forbes magazine