We simply don’t know what sort of jobs will be available in the future. After all, imagine yourself in the year 1900 peering into the future. How could you know then that the proportion of people employed in agriculture in the USA would fall to a twentieth of what it was then?
Or that there would now be more people employed as mental health nurses in the NHS than there are sailors serving in the Royal Navy? Or that large numbers of people would pay good money to personal trainers to put them through their paces and ensure that they suffered the requisite amount of agony?
History is full of people who have made long-term predictions and who have been proved utterly wrong. Among economists one of my favourites is the great William Stanley Jevons, one of the most distinguished economists of the nineteenth century. In 1865 he predicted that industrial expansion would soon come to a halt due to a shortage of coal. Poor old Jevons.
So we must tread warily. Having said that, and having dosed ourselves with lashings of humility, and drunk deep from the well of scepticism, there is a lot that we can say about the future of employment in the new robot- and AI-dominated future.
One of the most widely talked about categories of jobs supposedly at risk is drivers: bus drivers, truck drivers, taxi drivers, chauffeurs, delivery drivers, and many more. A 2017 trucking industry report predicted that by 2030, out of 6.4m trucking jobs in America and Europe, about 4.4m of them could have disappeared as “robots” do the driving.
But things aren’t quite as straightforward as they seem. On driverless vehicles there is a yawning gap between the hype and the reality. I will start by discussing the positive case and reviewing the potential impact, before moving on to the criticisms and problems.
The potential and promise
Driverless cars are not a fantasy: they are already working. Philip Hammond, when he was chancellor, told the BBC that he aimed to have “fully driverless cars” in use by 2021.
You can readily understand the reasons for the enthusiasm. This is only partly about cost saving. Human drivers kill 1.2m people a year, and additionally injure between 20 and 50m people. Some estimates put the cost to middle income countries at about 2pc of annual GDP.
And these accidents are generally caused by the common failings of human beings – drunkenness, tiredness, sickness, and distraction.
Moreover, imagine all those old and infirm members of society who can no longer drive, as well as all those who never could drive, who, in a world with driverless cars, would not need to drive. They would have as much mobility as the rest of the population, freed from the inadequacies of public transport and the expense of taxis.
Meanwhile, parents would be freed from the regular chore of dropping off and picking children up from parties, ballet lessons, football, and so forth. And going to the pub would no longer bring on the agonised choice over whether to drive and not drink.
Over and above all this, there is the saving of the time that we all spend driving ourselves to work, to meet friends and family, doing the shopping, going on holiday, or performing some errand or other.
It is fine if we actually enjoy the driving but most of us don’t – particularly not congested urban driving or being stuck in a motorway traffic jam.
Think how much better things could be if someone – or rather something – else did the driving. We could watch films, learn a language, work, drink to our heart’s content, or go to sleep. What bliss! Although the effects would not show up in the GDP figures, the result would surely be an increase in human wellbeing.
Wider still and wider
Furthermore, if driverless vehicles really take off, the ultra-enthusiasts talk of a transformation of urban land use as people en masse forgo their individual cars and are transported in driverless, shared-use, electric vehicles. It is quite possible that car ownership would fall sharply as people predominantly chose to take rides in driverless vehicles from a floating pool.
The results would include fewer cars needing to be built (as well as sold, repaired, insured, etc.). Additionally, there would be less demand for space to park cars that remain idle most of the time.
Would a world of driverless vehicles also be a world without traffic jams? CREDIT: GARETH FULLER
While they are waiting for users, driverless cars can be parked end to end and stacked. This could potentially transform urban landscapes and free up much scarce space for other uses.
In 2016, Google’s Chris Urmson told a US Congressional committee that in the US parking takes up an area the size of Connecticut. By implication, if everything went according to plan with driverless cars, this space could be freed up for other uses.
And there are potentially major effects on the insurance industry.
In the US, vehicle insurance accounts for about 30pc of all insurance premiums. There are particular issues with regard to who bears the liability when a driverless vehicle is involved in an accident. Doubtless this would provide a fruitful area of business for insurance companies – as well as, inevitably, for an army of lawyers.
But a sharp reduction in the number of vehicles to be insured would deal a heavy blow to insurance companies’ revenue streams.
So the implications of driverless vehicles are potentially huge. But it is now time to take account of the more sceptical view about their prospects, before making an overall assessment.
The idea of driverless vehicles has been doing the rounds for almost as long as vehicles have existed: General Motors introduced the idea at the 1939 World’s Fair in New York.
Of course, the technology has become much more capable than what could even be imagined then. (The 1939 concept was of a radio-guided car.) But this not-withstanding, the whole issue has always been characterised by over optimism.
In 2012 Sergey Brin, the founder of Google, said that driverless cars would be available to Google’s employees within a year and would be available on the commercial market in “no more than six years.” That means 2018. At the time of writing, we are still waiting.
A driverless world is still a long way off from becoming a reality, confounding the prediction of Google founder Sergey Brin CREDIT: KIM KULISH/CORBIS NEWSQ
In practice, there is no reason for the three elements of the ultra-radical vision of the future of car travel – driverless, shared-use, and electric – to occur together. We need to unpick this attractive and, supposedly inevitable, triad.
We could see the widespread use of shared-use vehicles or electric vehicles, or both, without seeing a large-scale move to driverless vehicles.
Feasibility is not the issue. Safety is. Demis Hassabis, one of the founders of DeepMind, said in May 2018: “How do you ensure, mathematically, that systems are safe and will only do what we think they are going to do when they are out in the wild.”
His misgivings are fully justified. Despite the claims of the manufacturers and developers of driverless vehicles that they are ultra-safe, a 2015 study from the University of Michigan discovered that the crash rate is higher for driverless vehicles.
The study suggested that, when they occur, crashes are almost always not the fault of the driverless cars. The problem seems to be that human drivers find it difficult to interact with other vehicles when the latter are driverless.
This is such a problem that some tech companies are trying to make driverless cars less robotic, even inducing them to cut corners, be aggressive, and inch forward at junctions.
In fact, things aren’t quite so simple as even this might seem to imply. For all the boasts about what their autonomous vehicles can do and the reports that their vehicles have passed so many tests with flying colours, the claims of the manufacturers and developers of autonomous vehicles cannot be taken seriously.
For these tests are usually conducted in secret and without independent verification. We do not know – and we are not allowed to know – the road and weather conditions that the vehicles were subjected to, nor how far they were dependent on any human intervention.
Even one of DeepMind’s founders has voiced doubts about autonomous vehicles
It is significant that so much of the experience with driverless vehicles so far has been in locations like Phoenix, Arizona, a place blessed with a predictable and attractive climate and good driving conditions. No snow, no fog, and no convoluted road systems or random congestion.
A more serious test would be to put these cars through their paces in London, Moscow, or Istanbul – in February.
Legislators, courts, and insurance companies are having to deal with some very tricky issues created by driverless vehicles. Under new UK legislation, drivers of self-driving cars must not take their hands off the wheel for more than a minute. And in April 2018 a motorist was banned from driving after being caught on the M1 motorway in the passenger seat, with the driving seat vacant as the AI “drove” the car. The British government is planning to scrap the requirement for a “safety driver” to enable advanced trials on public roads of fully automated vehicles by the end of 2019. It will be interesting to see how far this goes.
The American experience invites scepticism. In June 2018 a report from the Tempe, Arizona police department into the crash of a “self-driving” Uber vehicle that killed a 49-year-old woman who was crossing the street said that the crash was “entirely avoidable”. The report said that evidence showed that the safety driver had been distracted as she had been streaming a show on her phone.
In fact, among the developers of driverless vehicles and AI enthusiasts, such distinctions have gone much further. Discussion is dominated by the so-called six levels of autonomy, ranging from level 0 to level 5. But these levels are problematic.
There is a well-known case of an accident in Florida in 2016. Joshua Brown, a keen advocate of Tesla cars, had put his car on autopilot but the car’s sensors failed to register that a large truck was crossing the car’s path. The car steered itself under the truck, killing Brown.
Recognising these problems, within the industry the great ambition now is to develop vehicles at level 4.
Actually, a level 4 autonomous truck is awaiting regulatory approval in Sweden. There is no driver cabin or controls but, if need be, the vehicle can be operated remotely by a supervisor sitting hundreds of miles away, who can supervise up to 10 vehicles at a time.
This vehicle, known as a T-Pod, would initially travel only six miles a day and would be allowed to operate on only 100 miles of public roads where it might encounter vehicles with human drivers.
Yet substantially the same issues arise, unless the car really is capable of dealing with all circumstances. If not, how can you be ready to intervene in an emergency if you are not constantly alert? What if you are drunk – or asleep?
Flaws in the driverless logic
And what is the point of going driverless if you, the human driver, the “safety driver,” or whatever they call you, have to pay attention the whole time? Isn’t the point of going driverless that you, the erstwhile driver, can read the newspaper, fall asleep, or get drunk?
A further problem derives from the de-skilling of drivers as a result of relying on technology.
This is ironic because it is precisely when, for whatever reason, the technology fails, or cannot cope with a particular set of circumstances, that intervention by humans is required, humans who are supposed, at that moment, to be more capable than the machines/automatic systems that have failed.
But how can they be more capable if they have been used to sitting passively while an automatic system did all the work and made all the decisions?
This isn’t an issue with level 5 automated driving because in this case human intervention isn’t even possible. But, to reach level 5, driverless cars will need to be able to cope with all weather conditions, including fog, blizzards, and snow, be able to distinguish between a football being kicked into a road and the child chasing it, distinguish between a dog and a child, and be able to negotiate their way along streets crowded with people, often doing unpredictable, and sometimes apparently nonsensical, things.
None of these things can driverless cars readily do now.
What’s the point of a driverless car if the ‘driver’ has to remain alert? CREDIT: ANDREY RUDAKOV/BLOOMBERG
Robert Dingess of Mercer Strategic Alliance, which is a lobbying firm that specialises in automobile technology, has put the current state of play most pithily. He notes that the manufacturers have become good at “developing self-driving systems that operate safely 90pc of the time, but consumers are not happy with a car that only crashes 10pc of the time”.
Accidents are one thing but suppose that a nefarious person or organisation was able to hack into the system controlling a vehicle or set of vehicles?
This risk is scary enough with regard to disaffected individuals or criminals, but think about terrorist groups. If there were no human drivers in vehicles, including planes and cars, and a terrorist group were able to hack into the computer systems controlling these vehicles then they could deliberately cause mass slaughter on an industrial scale by turning the whole transport system into a weapon.
It falls appropriately to an economist to raise another key issue that seldom seems to be discussed by the techies, namely cost.
The kit necessary to enable a car to operate autonomously is phenomenally expensive. Just because something is technically feasible does not make it necessarily economically desirable, as the operators of the Concorde aeroplane painfully discovered.
This is not to say, though, that there is no scope for driverless vehicles. Already cars can drive themselves, unassisted, on motorways and can park themselves.
These features can and do bring some benefits to their users. And doubtless over time more people will want to avail themselves of these services. But, of course, this is a long way from the human user, the erstwhile “driver,” being able to switch off altogether.
And while human drivers are still needed to be capable of taking control then most of the much-hyped economic consequences of driverless vehicles won’t happen.
Mind you, journeys with restricted routes, where the scope for things to go wrong is limited, will surely be wide open to the full replacement of humans by AI drivers.
In fact, we already experience such AI drivers quite frequently.
Driverless rail shuttles at airports and underground trains have been common for some time. Four-person driverless pods to shuttle passengers between Heathrow’s Terminal 5 and the car park have been in operation since 2011. And there is clearly scope for an increase in the use of driverless tractors and other farm vehicles operating on agricultural land.
Norway has introduced a driverless ferry. Mind you, so far it has been confined to a 320-foot stretch of water in Trondheim. The journey takes only 60 seconds. I am sure that there will be plenty of instances of such “auto-ferries” moving people and goods short distances across restricted waterways. But this is a far cry from a fully automated oceangoing cargo ship or liner sailing without a captain.
Whether the vehicle is a car, a plane, or a boat, what has been achieved so far, and what is likely to be achieved in the foreseeable future, is a far cry from the wholesale replacement of humans by AI drivers that the AI enthusiasts think is imminent. In that case, the benefits of driverless vehicles have been massively oversold.
The whole thing has an aura of the emperor’s new clothes about it.
It reminds me of the excessive enthusiasm about all things digital in the run-up to the bursting of the dot-com bubble. Of course, there was much to admire about some of the ideas and the companies that proliferated then. And some have more than endured: they have grown like Topsy and transformed the business landscape.
But there was also an awful lot of dross that was blown away as soon as the bubble burst and people recovered their senses.
If it is right to think of the time and treasure spent on driverless cars as reflecting a bubble, then there is going to be one hell of a reckoning when it bursts.