ARTIFICIAL INTELLIGENCE, AUTOMATION AND THE LABOUR MARKET
Ryan Khurana // 14 February 2017
Part 2, for see part 1 here.
The great 20th century economist Joseph Schumpeter noted that it is not our technological capabilities that hold back our innovative powers, but the economic and social conditions of a time. This aversion to “creative destruction” has been present throughout history, with even Queen Elizabeth I rejecting the patenting of a machine knitter in 1589 for fear it would cost hard working men their jobs. Britain became the first nation whose lawmakers realised the emancipatory powers of technology despite its challenge to the status quo, developing rapidly over the following few hundred years. The form of these technological gains was one in which the lowest skilled gained disproportionately. Easy to operate machines replaced the skilled craftsman who produced fine textiles, and by the late-19th century almost eradicated the artisan guild. At the same time as destroying certain professions, they created far more, employing former agricultural workers in much more productive fields at higher wages.
That the creation of new technology creates more jobs than it destroys is a historically observable phenomenon, but it is the character of these new jobs that draws concerns. While the machine age disproportionately improved the opportunities for the lowest skilled, the computer age did quite the opposite. Computers, as opposed to factory machines, are not simple to operate, and require a great deal of training to reap their productive rewards. Since the 1960’s the new jobs created in the wake of more efficient automated processes where in the more skilled, not less skilled sectors. As companies became more profitable, and by extension larger, the number of managerial positions grew rapidly, with most of these careers requiring some form of higher education. Simultaneously, non-cognitive careers, especially ones that required a high degree of manual labour began to disappear. Many of the workers who had lost well-paying manual jobs entered the labour pool for service jobs such as retail, increasing the number of applicants, which concurrently held wages down. The growth in service jobs since the 1980s exceeded their growth in the preceding decades of the 20th century. As a result of the increasing demand for high skilled labour, and rising numbers of low skilled labourers, wage bifurcation began to emerge. This meant that new entrants into the labour market no longer had the same opportunities to work their way up in the pay scale, but rather competed for either low paying or high paying jobs.
With Artificial Intelligence, even the safety of many of the managerial jobs that commanded a high wage premium are now at stake. The new technological revolution does not discriminate between cognitive and non-cognitive tasks, meaning that simply being more data literate or technologically capable is not a guarantee of job security. The careers that have low susceptibility to automation are of a few varieties: either those that require certain non-automatable skills, those to which the cost of automation is too high, or those to which social barriers to automation exist.
The first group of careers are those that involve skills that current technology cannot replicate, such as empathy or critical thinking. Nurses, Lawyers, and skilled salespeople all fall into this category. It is important to remember, however, that these skills are only non-automatable given current technologies, and the skills required for these jobs will have to adapt rapidly to changing technologies. As technologies get better, those within non-automatable professions will either need to improve their own skills at a faster rate, or become better at harnessing the technology available to improve their productivity. As the combination of a person and a computer still outpaces a computer on its own at many tasks, the evolution of these professions may be one in which they can accomplish a lot more, faster.
The second group of professions are those to which the costs to automation are unreasonably high. Extremely specialised occupations such as professional wine-tasters, or even industry specific consultants such as fast food franchising experts, are ones where the technology to replace what they do is readily available, but the returns to doing so very low. It would cost a considerable amount of money to train an AI software to know all the region-specific requirements for fast food franchising, so much so that the return on that investment compared to just relying on human labour would be low. The high startup cost to automation is one of the reasons why, aside from within a few tech savvy multinationals, or at the startup level, traditional workforces still predominate. It costs too much to change the infrastructure used. However, minimum wage increases and other regulations which raise the costs of employees can make the cost of automating one’s workforce look relatively more attractive.
The final major barrier to automation is social; people do not like new things. There is a general reluctance among the population about using driverless cars, despite them being much safer than human drivers, though this reluctance will probably decrease over time. Similarly, strikes become more common in the face of public sector automation. The recent Southern Rail strike is an example of employee fear to changing work environments, people generally are unwilling to sacrifice their own benefits for overall efficiency improvements. Think of the backlash that would occur if one tried to automate the civil service bureaucracy, or the NHS. As Artificial Intelligence slowly creeps into people’s everyday lives, however, the changes seem less daunting. While social barriers may slow the pace of automation, they will not halt it altogether.
If the implementation of these new technologies is inevitable, what does this mean for an economy, and what should be done about it? In the long term we can see that the productivity gains and lower costs to production are a net benefit to all, and we can reasonable predict that new industries arise, even if we can’t exactly predict what these will be. However, in the short term, as certain skilled occupations disappear, those with professional qualifications would end up competing with unqualified workers, lowering wages, and potentially pushing some people out of the labour force altogether.
Three main policy proposals would help in minimising the damaging effects of this change: education reform, removing barriers to market entry, and a simplification of taxes and benefits. The current education system was designed for an industrial age of stable routine work, and is failing to adapt to modern times. Changes to curricula need to be more dynamic and should feature more input from parents and businesses alike rather than academic experts. The rise of after schools robotics clubs suggest that parents are working to improve their children’s chances in the new economy, and policymakers should be more receptive to their concerns. Simultaneously, a need for lifelong education rather than a single period before entering the workforce, aids in allowing for people to find new jobs if their old one ceases to exist.
Removing onerous regulations on starting a business, and the various financial and labour market regulations that make access to capital harder, and hiring and firing workers a nightmare, would improve the dynamic ability for the economy to adapt to changing technology. By making it easier to access capital and start a business, new jobs can appear more rapidly. Relaxing labour laws would allow those looking for work to work in a variety of positions, developing new skills, thereby making it easier for them to find a job they can settle into. By making this process harder, the government lengthens periods of unemployment.
Finally, complex tax and benefit systems distort incentives for both entrepreneurs and employees alike. By making tax codes difficult, the transactions costs for reaping the benefits of new technologies rise, thereby delaying or distorting their implementation. They add to the hesitance of starting a business, or to the capital requirements of accessing the necessary legal and accounting services. These difficulties slow down the adaptive processes of an economy to reap the productive advantages of new technologies. Similarly, complex and bureaucratic benefits systems harm those looking for new work, and disincentivise the retraining and experimentation needed to find a new job. These should be replaced with a single transfer payment, either like a negative income tax or a basic income, which would allow people to invest in themselves and seek new and better jobs.
To ensure we can reap the long terms gains of technological advancement, it is important that a sound policy framework is adapted to make the transition as painless as possible. What the long term benefits entail will be discussed in the next post.
First published on IEA’s blog.