The human touch: Machines can already perform many even complex human tasks better than humans can, so we should build capacity at the moment of interaction between the two. Photo: Oli Scarff/Getty Images
Over the past century, society has increasingly emphasised the work a person performs as a defining attribute of their identity. This trend became particularly salient from the 1990s, with a growing tendency to relentlessly pursue career advancement, a phenomenon often referred to as careerism.
Machines are, however, fast becoming better than human workers in a wide variety of professions. There is broad recognition of this fact, based on a range of rapidly advancing technologies. For example, MIT researchers Erik Brynjolfsson and Andrew McAfee argue that human workers are increasingly finding themselves in a “race against the machine”.
As we observe International Workers’ Day on 1 May, we should perhaps also reflect on what this will mean for us in the coming decades.
Central to Brynjolfsson and McAfee’s thesis is one particular statistic about the American labour force. They observe that since 1975 the gap between GDP per capita and real median household income has widened. Prior to the 1970s, the two indicators stayed fairly close together, suggesting that productivity gains were achieved through human capital and that these increases were reflected in the rise of median household income.
However, since then, GDP per capita has continued to rise while (real) median household income has stagnated.
Brynjolfsson and McAfee argue that the growing gap between these indicators reflects a fundamental change in how income and wealth is apportioned in the economy. Through the adoption of increasingly advanced machine technology businesses are able to maintain or increase their productivity without recruiting more human workers — a trend referred to as jobless growth.
The median worker, Brynjolfsson and McAfee conclude, is losing the race against the machine.
The impact of machines on human productivity is visible across all sectors and industries. It is easy to recognise in industrial settings such as modern factories or construction projects.
However, it is important to also recognise it in the context of what is sometimes referred to as knowledge work. Prior to the broad adoption of computers and internet-based platforms in academic research, which is my own area of work, research work was often hampered by tedious tasks. Finding and organising literature was central among these, as was the calculation of statistics.
In a 1937 article introducing a new statistical method (bifactor analysis), Karl Holzinger and Frances Swineford praised the method for its “relatively easy” calculation. When done by a single person, they state, the calculation can generally be done in fewer than 10 hours. Today that same calculation can be done computationally in less than a second.
Moreover, researchers today use internet research databases to retrieve, in a couple of seconds, neatly structured literature collections from across the world without leaving their desks. I personally haven’t set foot in a library in years.
Not surprisingly, on the back of these technological advancements, researchers are substantially more productive today.
In his analysis of trends in research productivity, Rob Warren from the University of Minnesota found that newly appointed sociology professors in the US published twice as much research in 2017 than those who preceded them by 20 years.
Brynjolfsson and McAfee accordingly propose that rather than framing this debate as one where human workers and machines race against each other, we should seek ways for workers to race with machines. Productivity increases substantially when human workers are able to develop and harness the capacity of machines in areas where our own abilities are limited. There are and still may be for some time various tasks that are not well suited to machines. These tasks typically involve fine motor skills, vision and locomotion, which are harder to automate than tasks that involve processing information.
The prospect of racing with rather than against machines certainly paints a more optimistic view of the future of work. However, it is critical to recognise that this prospect rests heavily upon our ability as society to develop the appropriate human capital.
As our machinery advances, we will require knowledge and skills that complement it. This is the real race we should be putting our effort into. On the one hand this implies that our economy will require a growing supply of computational expertise such as programmers, roboticists and sociotechnical system designers. On the other hand we should avoid building significant capacity in domains where work is likely to be automated in the near future. Some careful, future-oriented navigation is required here.
The composition of our labour force suggests that we’ve made some progress in terms of the general level of education of workers, but it’s been slow going. According to data from Statistics South Africa, in 2010, only 3% of the workforce held degrees or higher diplomas, fewer than 1% held honours degrees or postgraduate diplomas and only 0.6% held higher degrees (master’s or PhD).
Data from 2019 suggests that things have improved slightly. Just more than 6% of the workforce held degrees or higher diplomas, almost 4% obtained honours degrees and 1% had higher degrees.
These numbers are still a far cry from those of developed nations at the forefront of the digital revolution. Much more will have to be done to widen access to high-quality education at all levels if we want to develop a workforce capable of racing with the machines and competing in the global economy.