Back in 1978, when British television only had three channels, the impact and audience of any individual TV show was much larger than today. One episode of Horizon, a high-profile BBC science show, was large enough that, almost fifty years later, it has its own Wikipedia page.
The episode was on a then still new(ish) technology called microprocessors and, as Wikipedia notes:
Britain’s lagging place in the worldwide technology race was widely acknowledged after the documentary was screened. The UK government launched the Microelectronics Education Programme in 1981, with a budget of more than £10 million. This included nationwide discounts on computers to schools and colleges, and was followed by government backing of the BBC’s Computer Literacy Project.
Your columnist could note, at this point, that the very first computer he ever used was a BBC Micro, which was available in his school as a direct result of that broadcast. But more interesting than the documentary’s effects was the tone in which the economic potential of microchips was discussed.
As the programme’s original listing explained:
A machine that can read aloud; a driverless tractor; a production line without humans; a warehouse that needs no staff. Science-fiction fantasies that have already arrived! The reason is an inch square chip of silicon called a microprocessor. For less than £5 it can do the same job as the giant computer of a few years ago. Programmed and linked together, they promise a future that is both sensational and frightening. Offices, shops and factories are already being made more productive in a way that will cost millions of jobs. Now the chips are down, what are we going to do about it? Must we accept the widespread unemployment to come? Can we survive if we don’t? Above all, why is nobody talking about it?
The fear, at least in the BBC studio discussion in 1978, was that this new technology would lead to such a surge in productivity that unemployment would rise. Machines would take the jobs of humans. This program description could, of course, have been written, with very minor changes, about the potential of artificial intelligence (AI) in 2026.
Such worries are surprisingly common in the long sweep of economic history. Rising agricultural productivity meant that fewer people had to work as farmers to feed even a strongly growing population, and rising manufacturing productivity meant the share of the workforce working in manufacturing fell even as output rose. Nowadays, there is a frequently heard concern that rapid advances in AI will throw coders, lawyers, accountants, and a host of others out of work.
How real, though, are these concerns?
Asked this week whether “use of artificial intelligence over the next ten years will lead to a substantial increase in the unemployment rates in advanced countries”? The Clark Center’s US Economic Experts Panel was rather unsure.
Weighted by confidence, almost two-thirds of respondents expressed uncertainty, whilst 9% agreed, 23% disagreed, and 2% strongly disagreed.
The high levels of uncertainty are understandable. This is, after all, a question to which the answer depends on the development of a technology, how it is rolled out, how firms and workers respond, and also, to an extent, on the policy response of governments.
As Aaron Edlin of Berkeley argued, “It is too early to know. Past technological innovations have not. But this time could be different”. As Eric Maskin of Harvard put it, “Many jobs will surely be lost, but we don’t know yet how many new ones will be created”.
Interestingly, though, while the panel was deeply unsure about the impacts of AI on unemployment, it had more confidence in a couple of related questions. The panel was next asked whether the “use of artificial intelligence over the next ten years will have a negative impact on the earnings potential of substantial numbers of high-skilled workers in advanced countries”? Here, 4% of respondents strongly agreed, and 47% agreed, while just 13% disagreed and 36% expressed uncertainty.
While not exactly clear-cut, this does represent substantial agreement. Daron Acemoglu, of MIT, was one of those expressing uncertainty. He argued that “There are many claims that AI will impact high-skill workers. The evidence outside of coding is scant. AI will displace customer service and many other service workers. If displaced, high-skill workers may be able to adapt better to AI and take jobs away from low-skill workers”. Even though he agreed with the proposition, Robert Shimmer of the University of Chicago noted that there was a lot of nuance here, “Use of artificial intelligence over the next ten years will have a negative impact on some high-skilled workers and a positive impact on others”.
The panel was finally asked whether “Use of artificial intelligence over the next ten years will lead to substantially greater uncertainty about the likely returns to investment in education”? This time around, the result was more decisive. 9% of respondents strongly agreed, and 60% agreed. In the words of Larry Samuelson of Yale, “At this point, AI creates a great deal of uncertainty, in many areas, and is likely to continue to do so”.
Taken together, then, the only thing the panel was reasonably sure of was that AI would increase uncertainty about the returns from investment in education. On the really big question of whether this time really is different and if mankind is about to experience some form of technology-driven unemployment, the jury is still out.
