Daron Acemoglu on Robots and Jobs

By Jacques Crémer and Priyanka Talim

Posted on June 27, 2017


New technologies have often prompted fear and unrest, perhaps most famously when angry English textile workers smashed weaving machinery during the 19th-century luddite protests. Today’s concerns about the future of jobs are centered largely on the threat that they will be replaced by automated technologies such as artificial intelligence and robotics.


Does the rise of AI and robotics spell disaster for human jobs and economic growth? Can humans keep up with the pace of technological change, creating and adapting to new roles? Two groundbreaking working papers published last year by TNIT (Toulouse Network for Information Technology) member and MIT professor, Daron Acemoglu, and his co-author Pascual Restrepo of Boston University, have begun to change the way economists address these questions.


In the first of these articles, ‘The Race Between Machine and Man: Implications of Technology for Growth, Factor Shares and Employment’, the researchers proposed a new conceptual framework, raising the theoretical possibility that rapid automation need not signal the demise of labor, due to powerful self-correcting forces in the economy.


Focusing on real-world data, the researchers then drew a much gloomier picture in a follow-up paper, ‘Robots and Jobs: Evidence from US Labor Markets’. This is the first study to quantify large, direct and negative effects of robots on employment and wages. Their estimates imply that for every new robot per thousand US workers, about seven workers lost their jobs and wages fell by 1.5 percent.


We talk to Daron about the prospects for humanity in an increasingly robotized future, as seen by the press, the public, the Trump administration and economists. We also feature two fascinating essays on the subject by Daron and Pascual, which discuss their research and its implications.

Image: Daron Acemoglu

A TNIT member and MIT professor, Daron Acemoglu is among the 10 most cited economists in the world according to IDEAS/RePEc. For ‘originality, thoroughness, and prolificacy’ in economic research, Daron was awarded the John Bates Clark Medal in 2005. TNIT News caught up with him recently to discuss his research on automation.


Your paper ‘Robots and Jobs: Evidence from US Labor Markets’ has received extensive press coverage. Do you have any concerns that your findings might be misinterpreted? Is the public reliably informed on these issues by journalists and academics?


Daron Acemoglu: I don’t think the public is reliably informed at all. And yes, I am worried about how academic findings, including my own, will be and are interpreted. The problem is that there is a lot of uncertainty about the effects of new technologies on employment, and this creates room for hype. The media, in this field as in many other areas of technical expertise, is attracted to extreme statements rather than focusing on a balanced discussion of what is known in the academic area. In the context of the future of work, this takes the form of statements claiming that new technologies will bring the end of work for humans. Nothing in the serious research in this area suggests that something like this is in the cards.


‘The Economist’ and others have used your research to suggest that robots might displace humans in the same way that cars ousted horses. To what extent is this a useful parallel?


Daron Acemoglu: That’s not the conclusion I would have drawn. There are different ways of reading our results. On the one hand, the results are large, because the stock of robots in the United States is still small, and our numbers suggest that this might have led to 0.34 percentage points lower employment to population ratio in the United States between 1990 and 2007 than it would have been the case without the additional buildup of this robots stock. So one could go from this and project into the future that with many more robots in the next several decades, we will have a lot more jobs displaced by machines. But the numbers are not that large. Even if we have a huge acceleration in the adoption of robots, we are still talking about a few percentage points lower employment in the next several decades.


So my bottom line is that we have to take the displacement created by new technology seriously, but the research does not support a picture of the near future where robots and other machines will do all the work, and we will all stay at home and play video games (or sip the Burgundy wine all day).


Your colleague at MIT, David Autor, has struck a more optimistic tone about preserving jobs in increasingly automated workplaces, arguing that machines cannot replicate human traits like common sense and empathy. Do you think society is veering toward complacency or paranoia about the dangers of automation?


Daron Acemoglu: David and I do not fundamentally disagree. The way I would put it is that some tasks are harder for machines than others. But we have also seen that many predictions about what machines cannot perform have not come true. People used to claim that robots could not perform tasks that required hand-eye coordination, but current robots can do this quite well. (And for full balance, I should say that probably more predictions about what machines will do in the very near future, some from leaders of artificial intelligence and economics, have fallen flat on their face than those on what robots cannot do.) In terms of the reactions of the public, I think we have the worst combination. We have a huge amount of complacency punctured by bouts of paranoia.


To what extent might the market ease the transition to a robot era?


Daron Acemoglu: I am a big believer in the market and very cautious about regulation in general. But in this instance, I think we cannot just leave this to the laissez-faire dynamics of the market. First of all, given everything else that’s in place (for example, subsidies to capital, tax credits, accelerated advertisement), there is a strong bias for using machines rather than people. In other words, on first principles, I would expect excessive automation in both Europe and the United States (but there is no systematic work to date to back this claim up). Second, the biggest adjustment we need to make is in preparing our workforce to work productively with robots. For this, workers need to have skills that are complementary to robots, artificial intelligence and other new technologies, rather than skills that are going to be substitutable and thus easily replaced by machines. But we do not at the moment know exactly what bundles of skills are complements rather than substitutes to technology. How can we expect people in their teens to know this and make investments accordingly? Finally, the education system, especially high schools, is broken in the United States and many other advanced economies. There is little that an individual student can do as long as the high-school system is dysfunctional.


What is your reaction to the position of US Treasury Secretary, Steve Mnuchin, who said earlier this year that AI’s displacement of human jobs was ‘not even on our radar screen,’ and ‘50 to 100 more years’ away? How should policymakers be preparing for the future?


Daron Acemoglu: This is exactly what I would expect from a member of the Trump administration.


David Autor said about your paper: ‘I don’t think it is the last word on its subject, but it’s an exceedingly carefully constructed and thought-provoking first word.’ What do you think might be the next word?


Daron Acemoglu: It’s very nice of David to say so. I think the next word will have to come from a much more detailed look at how firms deal with these new technologies. But we also need to start looking at how artificial intelligence is going to be used in service occupations such as accounting, finance, law and so on. We are very much at the beginning of this process of radical automation.



The preceding is republished on TAP with permission by the Toulouse Network for Information Technology (TNIT). “Daron Acemoglu on Robots and Jobs” was originally published in TNIT’s May 2017 newsletter.


Read more from Professor Daron Acemoglu on Artificial Intelligence and the Labor Market: