AI, Robotics - The Impact On Jobs

This article comments on the impact of AI, Robotics, and 4IR upon Jobs. It is the third article in a series that addresses the impacts of AI, Robotics, and 4IR

It adds to comments on employment impacts included in my first article, of this series. The second article in this series can be found here.

I briefly comment on some of the assumptions that underly many forecasts about the future of employment in a new world of AI, Robotics and 4IR

Industrial and Technology Revolutions Disrupt Employment

Robert Allen Professor of Economic History recently published a study that considered the impact of automated weaving technology on the textile industry, during the Industrial Revolution. At that time, weaving was possibly the largest worldwide employer. Professor Allen noted that it required a highly skilled workforce. Automated weaving machines were quickly adopted; the impact was felt from England throughout the Mediterranean to China - skilled weavers were displaced by machines. Although new jobs were created to operate the machines, those jobs involved a much lower level of skill, and were significantly fewer than the number of people displaced.

Technology was soon adopted to automate other hand trades. The rising pool of unemployed had the effect of lowering the average wage.

Meanwhile, before the Industrial Revolution, India and China had the largest manufacturing sectors. However, with the rise of the Industrial Revolution in the West, India and China suffered effective de-industrialisation. Prior to the Industrial Revolution they exported manufactured goods to Europe; following the Industrial Revolution they were exporting mostly raw materials. 

Returning to the present day, we need to understand that most of the new technologies have significant disruptive potential to the Means of Production, Value-Chains, and Value-Streams, and the associated social and economic structures. For example, 3-D, and 4-D printing have the potential to collapse global supply chains. Some European companies that exported their manufacturing to China, have subsequently returned production to Europe, enabled by 3-D Printing. The European jobs that were lost when production moved to China are permanently gone; the Chinese jobs have gone as well!

Taking a historical perspective we can conclude that the introduction of new technologies does not necessarily yield employment growth; does not necessarily involve new higher skilled jobs; and when viewed on a global scale, the impact on employment of new technologies can, at best, be a zero sum game.

The Transition from Agriculture to Manufacturing

Further, in the West, from the mid nineteenth century to the 1970s there was a link between education and rising productivity, and employment and wages growth. It seems that during the period from 1850 until the 1970s, education enabled the economically productive realisation of latent human capability in a positive relationship with productivity and employment growth. However, from the 1970s that link appeared to have broken.

The break in the link between education, technology, and productivity may have resulted from the industrialisation of Asia. Trade flows have reversed. Asia is now exporting manufactured goods to the West; and the West has experienced a degree of de-industrialisation.

That is, global centres of employment growth shift in response to the industrialisation of new technologies; industrialisation in one geographic region can adversely affect employment in another.

In the current era that link between education, rising productivity, rising employment, and wages growth could only be re-established if there is a new regime of latent human capability that could be rendered economically valuable and productive. 

However, it is highly improbable that humans have an unlimited pool of economically valuable latent capabilities; in which case, it may be unlikely that the link between education, rising productivity, rising employment and wages growth will be re-established.

Forecasts of the Impact of AI, Robotics, and 4IR on Employment

There have been several published forecasts of the impacts on employment arising from the global deployment of AI, Robotics, and other 4IR technologies. I would like to comment on several recent forecasts.

The WEF has forecast (Future of Jobs report 2018) that 133 million new jobs will be created, while 75 million jobs will be lost.

Apparently, these figures are derived from employer surveys.

However, the IMF has concluded that there will be a significant employment and wages decline.

The IMF have based their conclusions upon alternative best-case and worst-case economic modelling.

Is there a way of reconciling these two opposing forecasts?

We could look at an analogy: the commercial development of a major gas discovery. The commercial development of a gas discovery would involve an initial setup phase, followed by an operationalisation phase. 

The setup phase involves significant infrastructure development requiring a large workforce. Many of the people employed in the setup phase would be contract labour, and specialists who typically work on contract also.

However, once the infrastructure has been developed and operational processes have been embedded many of the people involved in the setup will no longer be needed. Fewer people are required to support ongoing operations.

We could apply this same reasoning to the WEF forecast of 133 million new jobs. 

AI, Robotics, and other 4IR technologies will change the means of production. This implies a transition period from the current production paradigm to new means of production. That transition will require a significant number of specialists. We can reasonably assume that these are specialists in AI, Robotics, 3-D Printing, Robotic Process Automation, IoT, etc. They will typically be engaged on contract to setup new infrastructure, and processes. But, once the final optimisation of embedded processes has been ‘completed’ then they will not necessarily be required to oversee daily operations.

Other specialists may be employed to implement training for new workers; and retraining for displaced workers. But, over time, their numbers may decline also.

We should be skeptical of the WEF forecast of a gain in employment of 50 million and not interpret this as a long term absolute gain in employment. 

Given that this forecast is derived from surveys, I suspect that it reflects estimates for the initial setup and transition period. The long term ongoing gain, if any, would be much less.

Further, IDC estimates that spending on cognitive systems is likely to increase by over 50 percent in 2018. 

And a recent Accenture report (2018) noted that over 60% of executives expect an increase in roles requiring collaboration with AI over the next three years; with 58% planning to use AI to augment roles in their organisation. 

The same Accenture report stated that 76 percent of executives have already used AI to augment tasks over the past three years, and 90 percent have used AI and other technologies to automate tasks.

However, only 3% of employers plan to invest in related training and re-skilling programs.

This suggests that the focus is on automation and productivity gains; there is no planning for high skill job redesign, skills uplift, or retraining. 

This just supports the conclusion that the WEF forecast of new jobs are simply guesses of labour needs to effect the transition to a highly automated mode of business; the WEF forecasts are not the product of long range planning and modelling of labour needs.

A McKinsey study (2017) suggested that by 2030, 75 million to 400 million workers (3 to15 percent of the global workforce) will need to change their occupation.

This same report noted that income polarisation in the West will increase; high skilled workers with in-demand skills in AI, Robotics and other 4IR technologies will attract higher wages, while low-skilled workers, and those without AI, Robotics, or other new technology skills will decline.

The McKinsey report noted that employment growth could be maintained only if the displaced workers found new employment within a year.

That seems unlikely, given that the current focus is on automation and jobs displacement, and considering the significant scale of jobs disruption (up to 15 percent - 400 million - of the global workforce). 

Many of the people displaced will already be advanced in their careers, and age; this is a significant factor that will delay re-employment of these people.

The only comparable transition of such a large movement of the labour force was the shift from agriculture to manufacturing at the beginning of the 20th century. That transition was enabled by education in a positive relation with productivity and employment growth.

In the move from agriculture to manufacturing there was an increase in the number of jobs. Why? Because Manufacturing required human capabilities not used in agriculture, such as; planning, design, forecasting, and other specialist skills. That is, latent human capability was activated and made economically valuable and productive.

The McKinsey report, and other surveys addressing employment impacts, project employment growth based upon historical evidence that education (re-training displaced workers) will drive productivity and employment growth. However, the link between education, productivity and employment growth was broken by the 1970s.

It is not clear that there is a new regime of latent human capabilities that could be assessed as economically valuable, and which could re-ignite the link to productivity and employment growth.

I think that any new jobs created will arise from existing human capabilities, but with a higher level of cognitive demand.

If middle aged people comprise the largest part of the displaced workforce it just seems unlikely that they will find new employment within a year; that will lead to a rising pool of unemployed. 

As mentioned earlier, the regional shifts in industrialisation, and de-industrialisation driven by the disruptive potential of many of the new technologies will affect employment outcomes, resulting in an ever changing ledger of gains and losses across geographic regions.

The McKinsey article notes that an underlying assumption of employment growth is that productivity gains will be re-invested.

However, a rising pool of economically disenfranchised former middle-aged workers, and the growing global concentration of wealth may adversely impact the re-investment of productivity gains, economic demand, government taxation revenues, and demands for government social support.

Technologies are advancing in their capabilities and sophistication, and are being targeted at almost every field of human endeavour, and are gradually incorporating capabilities that only humans had.

Now, businesses do not create jobs unless there is a business benefit in doing so, and will naturally use technology to displace jobs if the business case is positive in that direction. Over time, new technologies will encroach upon and incrementally incorporate human capabilities, or these technologies will facilitate the collapse of Value Chains. The business case for human employment will tend to an increasingly marginal proposition. 

If people have a future role, it will likely be highly specialised, advanced analytical, or creative. But, that is not going to guarantee employment for the majority, at all.