As society looks to the future and the potential of AI and robotics, CIFAR Senior Fellow and MIT economist Daron Acemoglu reminds us that we must first look to the past.
“We have been here before,” Acemoglu told the crowd at the sold-out David Dodge CIFAR Lecture
in Ottawa on June 13.
Acemoglu cited a 1930 essay by British economist John Maynard Keynes that could have easily been written today.
“We are being afflicted with a new disease of which some readers may not have heard the name, but of which they will hear a great deal in the years to come … namely, technological unemployment,” Keynes wrote. He had hoped this would be a “temporary phase of maladjustment.”
There are two types of technology —replacing and enabling technologies—and this distinction is important because they create very different effects on the labour market.
The very first known computer was the Antikythera mechanism, which was used to predict the positions of the stars. It is considered an enabling technology because it allowed ancient Greek astronomers to do things that they were otherwise not able to do. Modern technologies like computer-assisted design still require people but make them more productive and in turn, increase wages.
On the other hand, replacing technologies displace workers and reduce wages. For example, the first power looms automated weaving and took weavers out of the production process.
Acemoglu said replacing technologies do not always create negative effects. When work is automated it can increase productivity and lower the cost of goods like clothing and cars. This boosts overall demand and makes everyone richer, Acemoglu said, so they can consume more.
There are two types of technology —replacing and enabling technologies—and this distinction is important…
“Actually what we should be scared of are not those brilliant technologies but the so-so technologies. They replace workers but they don’t reduce your costs all that much. If that’s the case you lose the workers in the displacement effect, but the productivity effect doesn’t change,” Acemoglu said.
It’s yet to be seen what affect many of the applications of AI will have on the labour market. Acemoglu said AI will never replace the human mind, but it will continuously take over tasks that humans have previously performed. Automation is already being introduced for things like financial planning and tax preparation as well as more creative applications like AI to perform bail decisions. Some soon-to-be-obsolete jobs Acemoglu listed include telemarketers, radio operators, drivers, assembly-line workers, and construction workers.
So where will new jobs come from? Acemoglu presented one optimistic and one pessimistic scenario. His hope is that new technologies will also create new tasks resulting in higher-skill jobs.
“When railroads came and replaced the horse-drawn carriages, we create a whole host of new occupations associated with the railroad; engineers, conductors, maintenance workers, financiers and managers. In the same way that when computers came we created a whole host of computerized radiology, computer engineers, software developers and very different versions of many of the tasks that were manual,” Acemoglu said.
Enabling technologies like robotics and AI should translate to increased productivity and wages, but we haven’t seen it yet
His fear is that job growth will be focused on low-paying service jobs. Over the last 35 years, there has been an “alarming” decline in employment rates and wages. With the tech boom in the 1980s, economists saw an increase in high-skill jobs. In the 1990s, there was growth at the top and bottom, but the middle suffered. Now, in the 2000s there is only job growth in the lowest skilled occupations.
Enabling technologies like robotics and AI should translate to increased productivity and wages, but we haven’t seen it yet, Acemoglu said.
The answer why may be found by looking at how society adapted to the Industrial Revolution.
There was essentially no wage growth from the beginning of the Industrial Revolution around 1760 to about 1850 despite very rapid technological change and technology adoption in Britain. For example, the first power looms could produce cloth several times faster than previous methods but it took time for overall productivity and wages to increase. This came in the second half of the 19th century with widespread adoption of power looms and complementary technology and training. Most importantly, Acemoglu pointed to major institutional changes. Britain took its first steps towards democracy and ultimately extended suffrage to all men and women. Universal education was introduced and labour unions started negotiating wages to protect workers. Acemoglu said similar changes may be necessary today.
Though there is much more to be understood about robots and AI, Acemoglu said the basic lesson is clear: The potential of technology can only be fully and equitably realized with a range of complementary investments.
Acemoglu said more research is needed to guide policies on what skills are important and how organizations and education systems must adapt.
“I think we are really very much in the middle of some very transformative changes and the way to deal with these changes is to understand them better, to make knowledge-based, quantitatively sound assessments about what they are doing and what they are likely to do and change our own investments, our own policy positions and our own institutions in a way that is coherent with this picture,” he said.