AI, industrial renewal, and the common good
Carl Benedikt is Oxford Martin Citi Fellow and co-directs the Oxford Martin Programme on Technology and Employment, the world’s leading programme on the future of work which researches the transition of industrial nations to digital economies.
Convoco: What do you think of the following statement by the inventor of HTML and founder of the World Wide Web, Tim Berners Lee: “It is difficult to argue against the thought that artificial intelligence will in the future be smarter than human intelligence and could thus take over the world.”
Carl Benedikt Frey: Artificial intelligence is already outperforming humans in a variety of tasks, such as chess, Jeopardy!, and more recently even Go. But general artificial intelligence is a different matter. No algorithm outperforms humans in every single task. We still hold the comparative advantage in complex social interactions and creative tasks, for example. It is of course possible that the comparative advantage of humans will vanish in those domains as well, but for the time being that future seems distant.
C: What made you start your research on the link between AI, computerisation and the future of employment?
CBF: Our understanding of the potential social and economic consequences of AI has not been keeping pace with how fast the technology has been developing. Our models no longer reflect technological and economic reality as algorithms are getting better at a variety of tasks, such as driving a car, performing medical diagnostics, and translation work. What got me and my engineering sciences colleagues in Oxford started on this topic was trying to understand what recent technological advances in artificial intelligence and mobile robotics are likely to mean for the workforce in the future.
C: Do you think that AI and computerization are beneficial to the idea of the common good?
CBF: On average automation has always benefited society, but sometimes at the expense of large parts of the workforce whose jobs have disappeared. It is up to society to make computerization a common good by making sure that its benefits are widely shared. It is not a common good per se.
C: Is there a big divergence in how and when computerization hits the different continents and differently developed worlds? Let us say that computerisation hits the jobs of low and high skilled workers in Europe and Asia first, is there a chance people might wish to emigrate to Africa and South America, reversing current migration trends in order to find jobs?
CBF: Despite the promise of the digital economy to make geography redundant, it has done exactly the opposite. Since the computer revolution of the 1980s, new jobs have increasingly clustered in places with a highly skilled workforce, such as London, Stockholm and Munich, while places that have specialized in low-skilled jobs are likely to see a large share of their labor force being automated away. It used to be an advantage to have an abundance of cheap labor to be able to produce at lower cost. With advances in automation that advantage is gradually disappearing. This is likely to exacerbate current trends in migration from low- to high-income countries. The problem is likely to be particularly great in Africa where the population is growing fast, but manufacturing industry is creating very few jobs. The current migration crisis is likely to be minor compared to the forthcoming one, as a large share of the African population will search for jobs elsewhere.
C: Is creativity and lateral thinking a domain that will remain firmly in the hands of human intelligence or will it also be taken over by artificial intelligence?
CBF: At least for now, creativity seems to remain firmly in the hands of humans. Over the much longer term this is obviously an open question.
C: Can you give us a positive outlook on the impact of computerization?
CBF: Over the very long term automation has always been an engine of comfort and prosperity. During the British Industrial Revolution, the benefits of automation initially went to the owners of capital: it took six decades for workers to see its benefits, but then finally, between 1840 and 1900, real wages grew by 123 percent. It is possible that we are seeing history repeating itself. Indeed, so far, the economic trajectories of the computer revolution closely resemble those of the Industrial Revolution. But if many workers lose out to automation in the short term, they are likely to take a stand against it. To avoid further populist rebellion and a looming backlash against technology itself, governments must find ways of making the benefits of automation more widely shared.
C: Working is an important element of identity. If we lose this, what is going to replace it? Are we at a turning point in history?
CBF: To be sure, our jobs are an important part of our identity. But clearly, social norms have changed in the past and could do so also in the future. If our main concern becomes how to occupy our leisure time, people will have to form their identities around things other than their jobs. And some already do. But it would without doubt be a turning point in history and not an easy transition.
C: If you had to make a prediction, what would you say working life will look like in ten years?
CBF: Over the past century, automation has relieved us from the most physically demanding, dangerous, and boringly repetitive jobs. As automation progresses in the near future, humans will still hold the comparative advantage in creativity and social intelligence. In other words, we will have to reallocate our labor to the things we enjoy the most. I suspect that for the workers that have the relevant skills, working life will be much more interesting and enjoyable in ten years time. My worry is that many workers who do not have such skills will be left behind in the transition.
C: What is your biggest worry for the future of mankind? Do you fear that humans will make themselves redundant?
CBF: The only thing we know about the future is that we are pretty bad at predicting it. It has always been a known unknown. Since the early days of the Industrial Revolution, people have worried that humans will automate themselves into obsolescence. Today, a frequent belief is still that the human worker will go down the route of the horse, which was gradually displaced in artillery, transportation, and agriculture, and which has now been reduced to ceremonial purposes and sporting events. The reason that the human worker did not experience the same fate as the horse is our insatiable demand for services, which we lack the technological capabilities to automate. Of course, it is entirely possible that technology at some point might be able to fulfill all of our demands. In such an event, our main concern will be how to distribute the gains from technology and how to occupy our leisure. For now, such a future seems remote.