The American futurist Ray Kurzweil predicted a “singularity” by 2030: an exponential increase in technologies, robotics and nanotechnology that would see AI overtake human intelligence in a sort of biotechnical convergence of man and machine.
While the singularity may still be far off, the impact of technology on the world of work can hardly be disputed. But is it a good thing?
The jury’s out, says Daron Acemoglu, Professor of Economics at MIT. On one side are the technology optimists who see a bright future immeasurably enhanced by technology. On the other side are those who see AI as a threat to jobs – with the potential to replace human beings altogether.
Professor Acemoglu, who was recently awarded an honorary degree by London Business School (LBS) for his groundbreaking research on institutions and long-term development, is well known for thinking big. At a conference hosted by the LBS Wheeler Institute for Business and Development, he stresses the complexity of how automation impacts labour.
The truth is, it’s just not that simple, says Professor Acemoglu. Because while robots can in many instances do the job of humans, resultant savings in cost can drive productivity. And where there’s a hike in productivity, there’s usually an increase in demand for labour.
Better questions, he says, might be: Are we developing the right skills to exploit the opportunities that innovation is creating? Do we have what it takes to spot those opportunities? And do we have the talent to meet any future increase in demand?
In conversation with the Wheeler Institute for Business and Development, Daron Acemoglu highlights how we can build a more inclusive system.
The disappearing middle
One thing that is clear is that employment growth has fallen in the US in the last 40 years. There is an appreciable drop in the employment-population ratio among prime-age males, says Professor Acemoglu, from 95% in the 1970s to less than 85% in 2015.
The worst affected are the middle classes. “We’ve observed a growth pattern in the US job market that is best described as perverse. If you divide employment groups into the very high skilled, the moderately skilled and the very low skilled, you see that the top echelon – the senior consultants, executives, engineers – this group continues to grow right up to the early 2000s. But thereafter we see something strange. Growth here lessens, the low-skilled group continues to grow, but the group that is squeezed the most is the group in the middle.”
This group, he says, are the moderately skilled staff who have historically worked the factory floor or in clerical occupations. Professor Acemoglu calls them “the basis of middle-class existence in the so-called American Dream.”
He has found similar patterns in European economies like Germany and Sweden: “We’ve seen an increase in the development and use of technology especially from the 2000s that is making us richer, but, in the process, seems to be creating readjustment pains.”
Enabling versus replacing
To understand why this is, says Professor Acemoglu, you have to first distinguish between two distinct types of technology. Computer-aided design software, 5-G cellular networks – these are examples of what he calls “enabling technology;” technology that makes staff more productive in their work. Designers can sketch with greater precision. Surgeons will be able to operate on patients remotely, treating special cases in a timely and efficient manner wherever they are in the world.
Then there is “replacement technology”. These are the automation technologies that can do the entire job of human beings, effectively replacing them and leading to wage reductions and job losses.
Professor Acemoglu cites the first industrial revolution in Britain. Jobs and livelihoods were sacrificed as machines took the job of weaving and spinning materials. He finds a corollary in the present economic climate of the 21st century. In today’s manufacturing, services and healthcare industries, the most “iconic” technologies being used are replacement technologies.
“Look at Hyundai factories and you will see cars being put together by industrial robots. You won’t see workers on the factory floor – these workers have been replaced by automation. And it’s a similar story in services and manufacturing. There is a displacement effect at play across a significant fraction of the US labour force. If you previously painted cars or affixed tyres, you’re simply no longer needed.”
But it’s not all bad news for employees.
The productivity hike
While it’s estimated that around 400,000 jobs in the US have been lost to replacement technology, that figure still only represents about 0.25% of the nation’s 6.6 million unemployed.
And let’s not forget that trade-off in productivity. “Companies that invest in replacement technology benefit from improved productivity and go on to spend less on wages. But replacing salaries with machines also reduces costs.”
When costs are reduced, says Professor Acemoglu, businesses are in a position to sell more of their products and services.
“Car manufacturers can sell more cars. And consumers who have access to cheaper cars might want to buy better stereo systems – and in other sectors, they might want better services, better healthcare, better education.”
Many of the sectors using replacement technology today are labour-intensive, says Professor Acemoglu, and so an increase in productivity is likely to translate into an increased demand for labour.
“There’s an interesting dynamic at play here. The productivity effect works against the displacement effect in such a way that replacement technologies can ultimately become a positive force for job creation and wage increase.”
So why is it not happening more in developed economies? It’s down to the quality of the technologies, says Professor Acemoglu. That and our capacity to exploit them.
The wrong kind of innovation
“It’s not the brilliant stuff, the nanotechnologies or robots that do amazing things and accelerate productivity that cause the problem,” he says.
“The real threat to employment and to the economy are the so-so technologies – those that are just productive enough and just cheap enough to be adopted, but not so significantly better than the human labour they replace to have any substantive impact on real productivity. In other words, if I use machines that yield very small reductions in cost, I’m just left with the displacement effect and none of the good stuff in terms of increased productivity.”
In this scenario, profits might increase but workers suffer. In addition, short-term gains like this ignore the significant longer-term benefits of exploiting technological and knowledge opportunities to innovate better.
Simply put, it’s the wrong kind of innovation: a use of technology that squanders opportunity and at the same time scores the “twin own-goals” of destroying existing jobs and failing to create new ones , and running into diminishing returns through replacement without generating the productivity gains that create new roles, tasks and jobs.
“We’re devoting our resources to the wrong kind of AI. AI is a broad platform and it can be used to create lots of new ways of organising jobs, but at present we just use it to perform very simple predictive tasks faster than humans.”
It’s imperative to re-think how we invest in and exploit technology to generate a win-win scenario in which productivity hikes create substantive gains in cost saving and open up new business opportunities.
“It’s time to get innovative about how we work with automation so that as we automate tasks, we also create new ones that generate a labour advantage for humans.”
This means getting innovative about education.
Educate to innovate
In the developed world, education systems have remained largely unchanged since the 1950s. This means that school leavers and graduates are not acquiring the skills they need to do today’s technology-driven roles. It also means they are underprepared to spot the exploitation opportunities within emerging technologies.
“We’ve seen this with the decline in employment and wage growth in the west over the last 60 years,” says Professor Acemoglu.
Again, he suggests a precedent in the first industrial revolution.
“Then, as now, there was a period of almost 100 years where you had this bewildering array of new technologies – more dazzling in a sense than what we’re seeing today –
but there was no resulting improvement in living standards or wages. And it’s a big puzzle. All we know is that things start to improve around the time that significant reforms came in that affected democratic structures – and more importantly, education.”
The lesson for modern economies, he believes, is that similar “big changes” might be necessary if we are to take fuller advantage of innovation.
It’s time, in other words, to get smart about technology.
“AI is still in its infancy and it will continue to revolutionise the labour market. I believe that we are not yet taking real advantage of the suite of technologies based on the silicon chip. How we make the shift from the wrong kind of innovation to the right kind and how we can get smarter about skills building are crucial questions.
“Because unless we really understand what we are grappling with we will not only miss the great opportunities out there, but we could end up making a social mess of our changing world.”
Daron Acemoglu was speaking at the event ‘Automation in Emerging Markets: Killing Jobs Before They’re Born?’, organised by London Business School’s Wheeler Institute for Business and Development. Established through the support of Lonely Planet Founders Tony and Maureen Wheeler, the Wheeler Institute aims to illuminate and help solve the world’s most pressing global issues through sharing and applying business expertise. It acts as a platform to identify innovative and sustainable solutions in partnership with local, national and global organisations.