The Corner

Economy & Business

Automation, Unemployment and Moravec’s Paradox

Writing in the Guardian, here’s Larry Elliott on automation. The whole article is well worth a read, even if it’s too simplistic to argue (as he does) that the Luddites were wrong. Over the longish term they most certainly were. The industrial revolution paved the way for an immense improvement in living standards. But what that happy history omits is the fact that it took a while to do so, a phenomenon known as the ‘Engels pause’:

In the first half of the nineteenth century, the real wage [in Britain] stagnated while output per worker expanded. The profit rate doubled and the share of profits in national income expanded at the expense of labour and land. After the middle of the nineteenth century, real wages began to grow in line with productivity, and the profit rate and factor shares stabilized.

Put another way, the Luddites were (broadly) right about what the new technology could do to their prospects and those of their children, but hugely wrong about what it would mean for their grandchildren.

It’s worth noting that the Engels Pause was also a time of growing popular political discontent in Britain,

Convinced by the logic that the hit to demand from mass unemployment will (to oversimplify) constrain the extent to which tasks are handed over to the robots, Elliott argues that the robots “will create more jobs”. More jobs? I’m not convinced, but he’s on stronger ground when he asks this:

[W]hat if these jobs are less good and less well paid than the jobs that automation kills off? Perhaps the weak wage growth of recent years is telling us something, namely that technology is hollowing out the middle class….

Robots are likely to result in a further hollowing out of middle-class jobs, and the reason is something known as Moravec’s paradox. This was a discovery by AI experts in the 1980s that robots find the difficult things easy and the easy things difficult. Hans Moravec, one of the researchers, said: “It is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.” Put another way, if you wanted to beat Magnus Carlsen, the world chess champion, you would choose a computer. If you wanted to clean the chess pieces after the game, you would choose a human being.

In the modern economy, the jobs that are prized tend to be the ones that involve skills such as logic. Those that are less well-rewarded tend to involve mobility and perception. Robots find logic easy but mobility and perception difficult.

“It follows,” says Joshi [an economist at BCA Research], “that the jobs that AI can easily replicate and replace are those that require recently evolved skills like logic and algebra. They tend to be middle-income jobs. Conversely, the jobs that AI cannot easily replicate are those that rely on the deeply evolved skills like mobility and perception. They tend to be lower-income jobs. Hence, the current wave of technological progress… is hollowing out middle-income jobs and creating lots of lower-income jobs.”

Recent developments in the labour market suggest this process is already well under way. In both Britain and the US, economists have been trying to explain why it has been possible for jobs to be created without wage inflation picking up…. The relationship between unemployment and pay – the Phillips curve – appears to have broken down.

But things become a bit easier to understand if the former analysts and machine operators are now being employed as dog walkers and waiting staff. Employment in total might be going up, but with higher-paid jobs being replaced by lower-paid jobs. Is there any hard evidence for this?

Well, Joshi says it is worth looking at the employment data for the US, which tends to be more granular than in Europe. For many years in America, the fastest-growing employment subsector has been food services and drinking places: bar tenders and waiters, in other words.

AI is still in its infancy, so the assumption has to be that this process has a lot further to run. Wage inflation is going to remain weak by historic standards, leading to debt-fuelled consumption with all its attendant risks. Interest rates will remain low. Inequality, without a sustained attempt at the redistribution of income, wealth and opportunity, will increase. And so will social tension and political discontent.

The Guardian is what it is, thus the call for sustained redistribution, but the risk of “social tension and political discontent” cannot be wished away. And the risk of that will rise significantly as automation gnaws its way higher up the food chain. 

And gnawing away is what it’s doing.  Here (for example) is a recent story from CNBC on radiologists:

Arterys, a medical imaging startup, reads MRIs of the heart and measures blood flow through its ventricles. The process usually takes a human 45 minutes. Arterys can do it in 15 seconds. The remarkable power of today’s computers has raised the question of whether humans should even act as radiologists. Geoffrey Hinton, a legend in the field of artificial intelligence, went so far as to suggest that schools should stop training radiologists. Those on the front lines are less dramatic.

“There’s a misunderstanding that someone can program a bot that will take over everything the radiologist does,” said Carla Leibowitz, head of strategy and marketing at Arterys. “Radiologists still use the product and still make judgment calls. [We’re] trying to make products to make their lives easier.”

According to Dreyer, a radiologist spends about half the day examining images. The rest is spent communicating with patients and other physicians. There’s only so much that automated systems can take over.

“Our desire to have somebody in control, I don’t think that will go away anytime soon,” said General Leung, cofounder of MIMOSA Diagnostics, which is testing a smartphone device that uses AI to aid diabetics. “Someone’s always going to want a person to have made the decision.”

True, but what will they be paid to make that decision?

Meanwhile, at the lower end of the scale, the traditional retail sector is taking a battering from the impact of e-commerce,  but so far as retail workers are concerned, the hit  from the switch to online shopping  will be both direct (store closings) and, in a sense, indirect, as those stores that survive turn to automation to defend their profitability:

A recent analysis by Cornerstone Capital Group suggests that 7.5m retail jobs – the most common type of job in the country – are at “high risk of computerization”, with the 3.5m cashiers likely to be particularly hard hit. Another report, by McKinsey, suggests that a new generation of high tech grocery stores that automatically charge customers for the goods they take – no check-out required – and use robots for inventory and stocking could reduce the number of labor hours needed by nearly two-thirds. It all translates into millions of Americans’ jobs under threat.

None of this will happen overnight, and there will still be room for employees to work alongside them, but there will be fewer of them – and what will they be paid?


The Latest