Robots: upside, downside

With the help of technology we might all live the high life with our robot slaves. But is this what we want? And who will own the robots?

They’re coming.

Let’s not even think about what could happen if the “singularity” – the point at which machines acquire a human-like intelligence better and smarter than ours – is reached. We have a problem right now.

Artificial intelligence, automated systems, robots – whatever we choose to call them – are almost certainly going to be able to do a large proportion of the work humans do now, more effectively and more cheaply.

Any sort of job in which people are asked to act predictably in response to specific information is likely to be automated. We won’t only have driverless cars; it will be possible to operate all forms of transport without direct human control, including buses, trains, taxis, ships and yes, even aircraft.

Jobs in manufacturing, already on a slide, will be decimated. Mining, which already uses robots for some jobs, could use them for all jobs. Automated machines could also take over construction, working more precisely and without any need for sleep or taking work breaks.

Retail may soon not need the vast majority of its workers. As online shopping gains even more popularity, goods will be dispatched from automated warehouses and delivered by driverless couriers. Even burger flippers may be a thing of the past, as smart machines produce fast food without human intervention.

Some industries that have been job creators will soon be job destroyers. Finance and banking, big employers now, are prime candidates for artificial intelligence to do a better job than humans, particularly given the fact that their infrastructure is mainly virtual.

Computer programs, such as those used in finance and banking, are easily and cheaply replicable. Even more alarmingly, like all software it is increasingly created with the aid of other software, which reduces the need for programmers.

‘Why are we not seeing the productivity growth that common sense suggests would come from information technology?’

But is there really a reason to worry? Labour-saving machines have been revolutionising industries for the past two centuries. Agriculture, in which most people once laboured with back-breaking toil, now employs only a tiny fraction of the population. Initially, during the industrial revolution, displaced agricultural workers moved to towns and cities and found jobs in manufacturing, which then also became more reliant on machines. Indeed the Luddites, in early 19th century England, destroyed weaving machines because they feared mechanisation would kill their craft-based jobs.

Their jobs certainly did disappear, but more appeared in their place, and as manufacturing became more efficient workers earned more and thus had more to spend, which, in a virtuous circle, spurred more production. Productivity, the economic measure of how much output is produced by each unit of a person’s labour, rose rapidly. In the golden era of productivity growth – from 1920 to 1970 – output per hour in the US rose by 2.8 per cent a year.

The constant creation of new jobs in new industries meant that higher productivity did not lead to mass unemployment. People acquired a taste for services as well as goods. Entertainment, health and personal care, travel and recreation became huge and created a host of new jobs.

Then came what would appear at first sight to be the productivity boost to end all boosts – the computer revolution. Now, not only could machines automate physical work, they could also automate work that required thinking and intellectual skills. Wouldn’t this send the economists’ measure of productivity through the roof?

Oddly enough, it hasn’t. As Nobel prize-winning economist Robert Solow famously said, “computers are everywhere but in the productivity statistics”. This has become known at the Solow paradox. In the US, growth in output per hour from 1970 to 2014 – the era of the computer – grew at a miserly 1.6 per cent.

Why are we not seeing the productivity growth that common sense suggests would come from information technology? Part of the reason may be that people now place high value on things that don’t cost much. For example, software, once created, can be replicated extremely cheaply. As a result many IT products such as apps, or services such as Google and Facebook, cost the consumer next to nothing. Even though they are highly valued by people, this value is not expressed in what they pay for them and their use has little direct impact on GDP.

This is a controversial topic because Google and Facebook do successfully monetise their software, so some of the value is picked up elsewhere. But there is a widely held view among economists that rapid innovation, in which products become much better even as they become much cheaper, is likely to cause GDP to be underestimated.

‘Personal care is a growth industry but since machines can’t do it it’s not going to have much productivity growth.’

But what will happen as the IT revolution gathers pace and intelligent software takes over more and more human tasks? Is this just a repeat of the industrial revolution? Can we expect that, as the machines advance relentlessly, humans will migrate to new jobs in new fields that are beyond the capacity of lifeless intelligence? If so, perhaps there is no reason for concern.

On the other hand, there could be a lot to be worried about. Adair Turner, former chair of the UK Financial Service Authority and now chair of the Institute for New Economic Thinking, argues in a recent paper that rapid growth in a highly productive new technology can lead to a productivity boom in one sector but poor productivity growth, and a slowing economy, overall.

In his paper, “Capitalism in the age of robots: work, income and wealth in the 21st century”, Turner points out that the overall impact on productivity depends on what jobs the displaced workers move to.

Let’s say that workers whose jobs are destroyed by artificial intelligence become low-paid personal care workers. They look after the sick, the elderly and the disabled, doing jobs the machines can’t do – simply because people prefer to be looked after by people.

We already see evidence of this happening. Personal care is a growth industry but it’s not well-paid, and since machines can’t do it, it’s not going to have much future productivity growth. So as more and more workers migrate into it, overall productivity growth will continue to slow.

This, says Turner, is possibly part of the explanation for Solow’s productivity paradox. We can actually have major productivity advances from applying artificial intelligence in some parts of the economy while productivity in the economy overall is slowing, which in turn drags down growth.

But is that so bad in a world where many of the material needs people have can be met very cheaply by the robots, who will provide them with food, create their clothing, build their houses and entertain them in a world of non-reality that will make today’s computer games look childish? Maybe it won’t be such a bad life. Provided you are not ambitious and have no aspirations.

However, because humans are human, they will want more. What will success look like in this new world?

‘Wealth is likely to reside in things the robots can’t create and make available to everyone cheaply.’

Wealth is likely to reside in things the robots can’t create and make available to everyone cheaply – things such as waterfront property, or houses in other attractive but scarce locations.

People who are talented and creative and design fashionable items that are not robot created, whether it be clothing, jewellery, computer games or music, could also do well.

However the owners of the robots – those who have the intellectual property rights to the artificial intelligence or the expertise to create new and better forms of it – will be at the top of the tree. If you consider the power already amassed by Google, Facebook and Amazon, this picture looks convincing.

It gets worse. There will still be cyber criminals trying to steal or subvert the AI. In response, there will be ever more costly efforts around cyber security. There will also be legal disputes over immensely valuable intellectual property rights to robot technology, and highly paid lawyers needed to defend these rights. And the rich and powerful will employ ever larger armies of lobbyists to bend government regulations in their direction.

These are all jobs that keep intelligent, highly paid people engaged but offer no overall benefit to human welfare. Economists call them “zero sum” activities because they produce no added value. We don’t really want more of them.

Overall, Turner’s vision of the future looks like one in which the trends we see in the present have deepened and solidified – high levels of inequality in the economy with sectors of low productivity growth growing ever larger.

As he points out, it’s an inefficient rentier’s paradise. The returns will go, not to those who are productive, but to those who own things such as highly desirable land or valuable intellectual property from which they extract rents – the economist’s term for excess income earned on an asset that is a drag on the economy.

How will the political or social systems develop in such an economy? Is it a society that is socially mobile? Is it compatible with democracy? Will the power of the minority over the majority lead to a new form of feudalism?

How should we deal with this? In the short term, what is important is education. Jobs requiring the least education generally will be automated soonest. Perhaps it will become very important to educate people in creative pursuits and craft skills, because those with the means will want to pay for human-created products as an antidote to the monotony of living with smart machines.

Perhaps also we should tax property more highly and ease protections for intellectual property. Maybe we should look closely at why AI-driven businesses such as Google and Facebook tend to so quickly crowd out competitors and dominate their markets.

Of course none of this is certain; we can’t know the future. But we can intelligently extrapolate the present, based on the advances in automation we know are coming. We can’t stop it; it has its own powerful momentum. But we can choose how we react to it. It’s something we need to be thinking about very deeply indeed.

Tim Dodd is The Australian’s higher education editor