We’ve all heard how automation and artificial intelligence are going to replace human beings in the workplace, leave millions unemployed, many jobs entirely in the hands of robots, and fundamentally change the workplace forever. Many people are already aware that automation has begun to take jobs from humans on a large scale, with millions of jobs gone for good.
Even if we take an optimistic perspective, like the one advanced by Jason Furman and Robert Seamans in their study of artificial intelligence in the workplace, and assume that no general artificial intelligence that can replace large numbers of workers across all industries is on its way anytime soon, a controversial stance, we expect significant disruptions to the labor market. Advances in narrow artificial intelligence, the kind that can perform specific tasks well but cannot replace workers across multiple industries, and automation would still cause unemployment to rise above equilibrium levels for years, perhaps decades.
Furman and Seamans point out that those with the highest risk of having their jobs automated are those with the lowest levels of education. Forty-four percent of all tasks done by those without a high school education are deemed “highly automatable.” This would create a difficult situation, as the newly unemployed will have increasingly limited prospects for work within their educational range. Simultaneously, vast numbers of people would compete for the few jobs that remain for them to do.
Think about the gravity of this scenario: a stubbornly high unemployment rate that remains at unacceptable levels for years and primarily affects those who are already towards the end of the line when new jobs are handed out. What would be the economic effects? The social effects?
The recent focus on “Deaths of Despair” in areas of the country where economic opportunity has long vanished warn of drug use, alcoholism, higher rates of children born out of wedlock, and various other problems. It would be foolish of us to think that we would not see similar effects from long-term technological unemployment.
Again, this is an optimistic perspective. The invention of a general artificial intelligence would cause far more severe problems, for more people, across more sectors of the economy, likely for longer.
However, there is a clear route out of this potential economic disaster that doesn’t call for fundamentally reorganizing the economy, significant changes to our understanding of welfare, or trying to put limits on technology: massive retraining programs on a scale last seen in 1942.
At the start of America’s involvement in WWII, millions of women with little employment history, especially in manufacturing, were trained to build the “arsenal of democracy” in record time. So effective was their training that, in many places, efficiency and productivity went up as new workers took their stations. This mobilization of the workforce is credited with both winning the war and ending the Great Depression.
Among the various programs designed to smooth the transition for workers was the Training Within Industry (TWI) service. This program, managed by the Department of War, certified 1.75 million workers in basic job training. These graduates went on to provide on the job training to the inexperienced workers who entered the factories.
While it must be admitted that training people to do jobs machines can’t do will be more complicated than teaching people to work assembly lines, the fact remains that we have dedicated the resources needed to completely retrain vast numbers of workers before and we can do it again. If automation and artificial intelligence are going to be half as disruptive to the labor market as many people are suggesting they will be, then our response to it will have to be on a similar scale.
The need for large scale retraining programs is already accepted when it comes to other sources of unemployment. The decline in the demand for coal miners, the result of a variety of factors, including automation, comes to mind. Proposals for training miners to code are well known and seen as more than reasonable in many circles, though the program’s execution left something to be desired.
A program to battle technological unemployment would have to be even larger in scale. Its reach could not be limited to Appalachia’s coal-rich areas but would have to stretch from each metropolis to every hamlet. It could not focus on teaching one skill but would need to address a wide variety of skills, trades, and crafts.
While nothing on the scale of what we’ll need to deal with retraining currently exists, there are examples that point the way forward, which aren’t confined to the pages of a history book. The German system of retraining is a vital part of their economic system. Based on offering unemployed candidates extensive vocational education, a variety of studies agree that the program leads to better employment prospects and higher wages for the previously unemployed and works well for young people going into trades. It is much larger than any training program in the United States, with more than a million students in 2017.
This program works, and we should steal large parts of it before we have to.
Studies suggest that job retraining programs have both economic and social benefits for displaced workers, which we have plenty of right now. Even just improving outplacing policies leads to better outcomes for both the newly unemployed and society at large- even in the absence of the long term spike in unemployment described above.
Job retraining programs are going to become extremely necessary when automation hits the labor market with full force, but they can immediately do a lot of good, too. If nothing else, we should be helping the people who need to find work now and using what we learn from the effort to inform our response to automation later.
An adage says, “The best time to plant a tree was ten years ago. The second-best time is now.” We should have been retraining people to do the things machines can’t do for a while. We haven’t, but there is no reason we can’t start doing it. The benefits of doing it are substantial, while the price of taking no action is astronomical.