It’s part of the human condition to assume that the future will be like the past. Such thinking has its evolutionary advantages, but is increasingly out of step with the modern world.

Living memory can be divided into two periods – the post-war recovery in the West, and globalisation, which began in the 1960s and 1970s, but really gathered pace over the last 25 years.

In the first part of this model, technological progress went hand-in-hand with Western prosperity.

  • Better machines meant increased productivity, earnings and profits.
  • It also meant higher wages for workers, and as these wages were spent, more demand for the products made by the workers.

The second half of the story is not so rosy.

  • Globalisation has taken work (and jobs) from the west and moved it over to developing countries.

Global inequality has reduced as emerging markets have emerged, but the working and middle classes in the West haven’t felt the benefits.

  • Despite continuing – though slowing – productivity gains, wages are stuck where they were in the 1970s and 1980s.
  • The cost of things like healthcare (outside the UK, where it is centrally provided) and education have soared, and living standards have been squeezed.

The money has gone to the western elites and the poor in the third world, and in Europe, to migrant workers from the post-communist east who have the courage to make a new life in the west.

  • Income and wealth inequality has soared in rich countries.

Increasingly, we’re beginning to see the frustration of the non-elites at the Western ballot box.

The optimists amongst us continue to assume that the future will be like the past.

  • As jobs disappear to the evolving machine workforce, new ones will appear that demand uniquely human skills.

I’m not so sure. For centuries, machines have been tools to help workers do more work. What happens when machines themselves are the workers?

  • When robots can do service jobs, and intellectual jobs, and can largely maintain themselves, where will the jobs for people be?

The traditional response to automation is more training and education, but for what?

  • When robots can do everything a human can, who will choose the human?

It’s possible that some new type of work not yet understood will emerge, but I see no reason to be confident.

  • Rich people will crave the human touch, and so jobs looking after rich people will still exist.
  • But that looks more like Downton Abbey than Star Trek to me.

Dissenters say that most work will never be done by robots. People are just too smart.

These are the same people who said that a computer would never beat the world chess champion. Or beat the quiz champion of Jeopardy! Or more recently, the world Go champion.

  • Sadly for those people, there are no more games left for computers to conquer.

Most office jobs are clerical and repetitive.

  • Journalists and lawyers rely “only” on natural language processing.
  • As this evolves, so will their world shrink.

Autonomous vehicles and drone deliveries will remove millions of driving jobs.

Education and healthcare require high skills, but are also ripe for disruption.

  • Why pay big bucks for a mediocre lecturer or surgeon on your doorstep, when you can remotely access the best in the world?
  • We’ll still need doctors and professors, but far fewer of them.

To put things into perspective, consider Moore’s Law.

  • Gordon Moore was the head of Intel, the microchip maker.

Back in the 1960s, he said that computer power (per dollar) would double every eighteen months to two years.

Remarkably, this has held true for almost 50 years.

  • Since 1958 (the invention of the integrated circuit), processing power has double 27 times.
  • That’s 134,217,728 times faster – 134 million times faster.

That’s the theory, what about the practice?

Well, it’s 30 years since I was first rich enough to buy a business computer to keep at home (I’m not counting my Sinclair Spectrum here).

  • The computer on my desk today (which is about 18 months old) is around 1,500 times faster than the first one that I bought.
  • And in real terms, it cost only 20% as much.

So my computer is around 7,500 times faster than it was 30 years ago.

  • Not quite 134 million times faster, but then we’re only looking at 30 years, not the 58 since 1958.
  • Over the past 30 years, computer power has doubled 14 times, which makes the computers of today more than 16,000 times as fast as those of 30 years ago.

Remarkably, around half of that has been passed on to me and my entry-level consumer machine.

Of course, some of my 7,500 times gains have been “wasted” on software to make the machine friendlier and easier to operate.

  • Back in the 1980s, only real computer geeks kept one at home.

But the improvement beats that in any other field of human endeavour over the same period, and underpins why this time things might be different.

It’s true that automation will initially – and perhaps for a long time, mostly – threaten low-paid and low-skilled jobs.

  • This remains significant, since in recent time these have been the majority of jobs that have been created.
  • Most people don’t want to work these jobs, and so tech firms and politicians can describe this change as “freeing up” people from drudgery.

That much is true, but if there are no other jobs to go to, will being “free” turn out to be a blessing or a curse?

The key determinant of whether a job will be taken by a machine is whether it is predictable.

  • Could someone – or a machine – learn to do your job by looking at how you’ve done it in the past?
  • Or by copying those actions?

The “big data” trend means that there is more and more information being collected about what people do.

  • Just around the corner is machine learning intelligent enough to process this data and work out what people did.

A good example from today is radiology.

  • The images produced by X-ray machines and other scanners don’t mean much to the untrained eye.
  • Radiologists generally have 13 years of post-school training before they are put in charge of diagnosis.

Yet there are already machines that can outperform them.

In truth, most jobs are routine and predictable.

  • Few of us are lucky enough to perform real creative work.
  • The new machines will be general purpose, able to learn almost anything.

And the new industries that do emerge will involve few jobs.

The classic example here is Instagram.

  • Kodak – the firm that dominated photography for more than a century – had 145,000 employees at its peak.
  • Instagram – the internet service that largely replaced Kodak – had 15 people when it was sold to Facebook for $1bn.
  • Facebook itself had fewer than 5,000 employees, but was much more valuable than Kodak had ever been.

So there’s real potential for a lot of jobs to disappear, and for serious disruption to the economy.

Already in the US, 40% of consumer spending comes from the richest 5% of households.

  • As the value of labour falls, and the share of labour held by humans shrinks, we need to think seriously about how we get purchasing power out to the masses.

That of course, is assuming that we want to stick with consumer capitalism.

Until next time.


Don’t forget to share this via , Google+, Pinterest, LinkedIn, Buffer, , Tumblr, Reddit, StumbleUpon and Delicious.

Published by Mike Rawson

Mike Rawson has recently re-awoken a long-standing interest in robots and our automated future. He lives in London with a single android - a temperamental vacuum cleaner - but is looking forward to getting more cyborgs soon.

Leave a comment

Your email address will not be published. Required fields are marked *

Where did all the jobs go?

by Mike Rawson time to read: 5 min
Hi there - can I help you with anything?
[Subscribe here]
More in Economics, Editorial
AI and jobs
The big debate: Artificial Intelligence – Digital Catapult

We recently held a heated debate - would AI increase the number of jobs? Peter Karney and Darren Murphy went head...