Automation: Friend or a Foe to Humankind?
Kailash Raj Pandey
Technological change has become a prime phenomenon in the 21st century. With the rapid development in processing power, and chip development doubling almost every 18 months (Moore’s Law), the world has globalized with the help of technology, and that technology has changed the way we live. However, technology is also structurally disrupting many labor markets and that is a concern for governments. Over the past few decades, technology has replaced workers, particularly in the manufacturing industry, and continues to do so for workers in different industries that perform routine tasks that do not require creativity. As much as technology has been the main driver for job loss, some economic pundits and recent research recommend that a solution to adapt to the technological future lies in fundamentally revolutionizing our education and training programs.
Jobs that currently require 40-hours per week and require medium-level skills appear to be at the highest risk of being automated. Figure 1 shows that occupations offering medium wage have the potential to be automated compared to the extreme tails of the wage distribution. Also, jobs with forty hours of work in an average week are mostly prone to automation across all wage levels. The US Job Market has been polarized after the 1980s, especially due to the increased accessibility to computing technology.
Computerization has substituted jobs with routine tasks. Alternatively, it has increased productivity in jobs that require more abstract and creative thinking skills. Figure 2 shows a high correlation between repetition and degree of automation, especially for jobs that offer medium wages (in blue). Additionally, the jobs that require the highest level of accuracy are labeled, which show that jobs that are more technical and require accuracy are more prone to be automated.
Workers with college degrees have a comparative advantage in non-routine tasks.In a paper published in 2003, Autor and co-authors assert that technology changes the task content of different jobs. This effect of technologically replacing low-skill repetitive jobs, referred to as routine-biased technological change, has motivated low-skill workers to switch to service occupations where they are less likely to be replaced by machines. In another paper co-authored with Dorn, Autor also observed that employment prospects among non-college workers are higher in service occupations than in other low-skill occupations. This switch to service sector is due to technological change and its effect on employment prospects in other industries. With recent development in smart intelligent systems through machine learning and computer vision, even these tasks that were initially sheltered from technological change are becoming vulnerable.
Some economists like Robert Gordon claim that the IT revolution has already taken place and productivity growth is the lowest it has been because technology hasn’t fundamentally changed in the 21st century. Gordon claims that these new technological developments in the last decade aren’t changing the way we work like automobile, electricity, and telephone did in the past century. He believes that the big transition from paper to digital platforms was mostly complete by 2005 and there hasn’t been a big invention of a “general purpose technology” since.This may be partially true, but productivity could change direction and start speeding up in near future. There is a lot of development happening right now that has the potential to completely turnaround industries like medicine, retail, and communication, and in general how we live. For example, Internet of Things (IoT) is an area of technological progress that has been popular in recent years with the development of auditory and visual sensors in objects that allow them to gather, store, and share data with other objects over the internet. This could truly revolutionize agriculture, retail, transportation, personal fitness, and many other industries. Moreover, nanorobotics and 3D tissue printing are close to revolutionizing medical care. Gordon doesn’t foresee multiple technological frontiers like these. As much as he has a great explanation for historical events and shows how different inventions have affected productivity, it is important to acknowledge that a short decline in productivity growth recently does not necessarily mean that there would not be a groundbreaking technology soon.
Mckinsey’s 2013 report suggests development in artificial intelligence and Human Computer Interaction (HCI) is opening more avenues for automation including tasks that are knowledge-based and were taken in the past as irreplaceable by machines. These technologies could potentially even change the structure of labor share in the economy. This is expected to impact more than 9% of the global workforce and 27% of global employment costs. IBM’s Deep Blue Chess champion software developed in 1997 was a huge feat for mankind in terms of technological progress. However, that has advanced vastly to technologies like reinforcement learning using deep neural networks. This method allows the program to teach itself to be advanced using reinforcement learning, without human data, guidance or knowledge beyond the basic rules.
A product of this methodology, Google Deepmind’s AlphaGo Zero is featured in an article published in October 2017 in nature, titled “Mastering the game of Go without human knowledge”. This suggests how we might be going into another era of technological disruption where we are not only automating routine tasks, but we are building self-sufficient smart technologies that are able to make a subtle judgement and complex decisions like humans do. Moreover, advances in robotics with sensor technology and computer vision are making robots more portable and accessible on a mass scale. This is expected to affect about 12% of global workforce in manufacturing. All of this suggests that the comparative advantage of humans over robots in service jobs is likely to diminish over time, creating further problems for all the low-skill workers that have switched to the service sector.
Much research and discussion has been focused on automation threatening occupations instead of tasks. However, Arntz, Terry, and Ulrich, in their working paper, look at tasks instead of occupations to distinguish between workers performing different tasks even if they are in the same kind of occupation. Ulrich and co-authors also emphasize the importance of more training for low-skill workers to adapt to new tasks that are complementary with those replaced by the machines, within their own occupations. This is possible through specialized vocational training programs. Coding bootcamps are an example of this new emerging model.
Coding bootcamps like Horizons have tapped into the model of “getting-paid to learn” where they make up for the costs for training from salaries trainees receive in their first job out of the program. There are several of these bootcamps that are promoting this accelerated-learning model and teach very specific skills in the technological sector. They also are more career specific than a typical college degree. They support students in obtaining a job, with just 12 weeks of learning. These bootcamps proivide rapid training, focus on experiential learning and practical skills necessary on the job, and are catered towards specific jobs that are demanded in the labor market. This model is the future for middle-wage jobs where people can go and get new skills in a short amount of time to switch tasks in the same occupation or switch to a different industry. Even though these bootcamps currently favor college-level students, this model can be expanded into vocational training programs for low-skilled workers as well.
The World Bank Group, with an initiative called “Decoding Bootcamps,” focused on evaluating the effect of these coding bootcamps on youth employment in emerging markets. In a recent official report published in 2017, they identified these bootcamps as a tool to build “future-proof skills through rapid skills training.” They have compiled five case studies from the US (Hack Reactor), Peru (Laboratoria), Kenya (Moringa School), Lebanon (SE Factory), and Colombia (World Tech Makers + Coderise). They find that these programs had high employment rates in low-entry tech positions and this is dependent upon the network of the programs in the corporate sector and how they increase industry exposure for their participants. As a result, they conclude that these results how understanding this model can allow for public-private partnerships with an intent to target the low-skill and marginalized workers in different industries that have become vulnerable to automation.
Figure 3 clearly shows that there is going to be a huge increase in the employment prospects for jobs that require some degree, and in particular Bachelor’s degree. There will be a lot more value in holding a degree in the near future, which isn’t very surprising. Goldin and Katz, in their paper “The Race between Education and Technology,” also highlight that stagnation of educational development has historically been a significant factor for increase in US wage inequality. They suggest a quality preschool education for disadvantaged children, preparing more students for college with better primary and secondary education, and making higher education more affordable for all.
Now, there is an increasing need for dynamic reform in the education system. Introducing short-term accelerated vocational programs like those mentioned above is a step forward to address new demands in the job market due to technological change and automation. Governments should not shy away from embracing technology as it has tremendous abilities to transform economies. Increased automation in different industries every year is demanding adoption of rapid skills training programs, like different coding bootcamps mentioned above, for various industries being threatened by automation. Only then, automation could be a friend, and not a foe to humankind.