Since the beginning of large-scale industrialization, automation has led to massive, widespread job losses. Whole cities like Detroit fell victim to this when the car industry replaced large numbers of humans with robots. Today, many of the hardest-hit places are barely a shadow of their bustling and blossoming past selves.
Adding robots to the economy displaces workers. A study from the University of Chicago found that adding just one machine per 1,000 workers to the economy causes the employment rate to decrease by at least 0.18 percent. This may not sound like much, but in a country the size of the United States, it corresponds to about half a million layoffs. That’s the equivalent of a medium-sized city of workers without a job.
Those workers face an immediate, direct crisis as they have to find new jobs or maybe even new fields entirely. Their unemployment also has downstream effects on the broader economy as they cut down on spending for the duration of their time out of the workforce. Unemployment affects everyone who participates in the economy.
With the fourth industrial revolution in full blast, the fear is palpable yet again. Artificial intelligence has the potential to automate even very complex tasks, to say nothing of the simple and repetitive ones. That labor-saving potential only sounds great before you think about the social consequences. No city wants to end up muddled in bankruptcy and decline. No workers want to lose their livelihood, boring and repetitive though some parts of their jobs may be.
But the news isn’t all bad. Without robots in the automotive industry, the innovative products made by companies like General Motors or Tesla wouldn’t even exist in our wildest dreams. Without the robots in tech factories, the device that you’re reading this article on wouldn’t exist either. And without an AI-based recommendation algorithm, you might never have stumbled upon this article.
Stories like the decline of Detroit are gut-wrenching. But there’s reason to believe that history won’t repeat itself when it comes to the AI-fueled revolution.
Companies That Adopt Robots Hire More Workers
On a macroeconomic level, the logic seems simple: If AI makes workers obsolete, then adopting it will make unemployment rise. At first sight, a study from France confirms that suspicion. The authors found that a 20 percent increase in robots in a given industry leads to a 1.6 percent decline in employment there. Of course, robots are a way more general term than AI, but we can assume AI would lead to similar results.
The reality is different, though, when shifting perspective from the national economy to individual corporations. Somewhat counterintuitively, companies that adopt robots hire more workers. Admittedly, this data might be a bit misleading because companies with stronger growth can afford more robots sooner, which makes them scale even faster.
There are, however, a couple of compelling reasons to believe that robots help companies expand their human workforce. The French study suggests that, if workers and robots share the workload, then the value added per worker goes up.
For example, consider a company that employs five workers to manufacture a product worth $100. On average, every worker contributes $20 to the final product. After experiencing some growth, this company buys some robots and now only needs two workers per product because the robots are doing the rest. As a result, the remaining two workers contribute $50 each to the final product. Since this is a dramatic increase in efficiency, the company might expand its activities and hire additional workers, who can also now contribute $50 each thanks to the improved efficiency — compared to $20 in the pre-robot stage. This mechanism increases a company’s labor demand, so it might decide to hire more workers to expand its palette of products and services.
Robot-adopting businesses also grow their headcount because providing products and services usually becomes cheaper when they’re not handmade. Sure, buying machines is an upfront investment and some maintenance work is necessary. Robots don’t get tired, sick or need to go on vacation, however. In the long run, this efficiency pays off. The reduced cost makes it easier for these companies to snap up more market share in their industry because they can offer higher quality products at the same price or keep the same quality but offer lower prices than the competition.
Research from Canada, Denmark and Spain comes to similar conclusions. Although no such study has been undertaken in the U.S. as yet, companies need to invest in robots if they want to stay competitive, even if employment in their sector goes down as a result.
We should keep in mind, however, that every AI-system can be viewed as a robot, but not every robot contains AI. Investing in AI might lead to more corporate growth just like general robotics. Fortunately, though, unlike other robots, AI might not lead to an industry-wide decline in employment.
A Dutch study comes to a similar conclusion as the French, Canadian, Danish and Spanish ones, namely that more automation — meaning more robots — leads to more unemployment. Despite this, computerization, under which we can safely count AI, doesn’t have an effect on unemployment.
In other words, although physical robots do eliminate jobs to some extent, computers and AI don’t have the same effect. What this study doesn’t show, though, is the link between AI and economic growth. But other studies have.
AI Is Poised to Fuel Explosive Growth
Those readers who are old enough to have witnessed the rise of the internet (the author was too small back then) might remember how it raised similar concerns about unemployment. Even though it became widespread a mere 20 years ago, it has already created millions of jobs since and comprises 10 percent of the U.S. gross domestic product. Perhaps more importantly, the internet hasn’t been responsible for the decline of cities like automation in the car industry has.
Currently, 63 percent of CEOs believe that AI will have an even larger impact than the internet, according to PwC. In their global AI study, PwC estimates that by 2030, the global GDP will have increased by 26 percent due to AI alone. That’s more than the current GDP of China and India combined.
PwC isn’t the only optimist in the room. The World Economic Forum estimates that 97 million new jobs will be created through AI by 2025. They also estimate that some 85 million jobs will be lost in the same period, but that’s still a net surplus of 12 million jobs.
That being said, the distribution of these jobs might be unequal. In Sweden and the U.S., productivity is projected to increase by more than 35 percent by 2035 due to AI. For France and Spain, however, the gains might be less than 20 percent.
This effect might not be as enormous as it sounds, though. The World Economic Forum estimates that only 40 percent of the AI-related economic growth by 2030 will come from an increase in productivity; the other 60 percent will come from consumption. In other words, a little less than half of the growth will come because workers get more efficient or get replaced by AI. The rest of the growth will come from the fact that people will be using products with AI, such as social networks or streaming services. One caveat to this point is that we’re comparing two different time spans here, as one projection was for 10 years and another 15. Nevertheless, it’s clear that productivity is only a part of the equation that contributes to overall economic growth.
Despite legitimate fears, it seems as if the adoption of AI will lead to sustained economic growth, which in turn spurs employment rates. The important part will be to make this growth fair and equitable to as many citizens as possible both within and across nations. Within a nation, this means making sure that marginalized communities such as women, LGBTQ+ individuals, overweight people, people with disabilities and people of color get access to training and education in AI. We must also strive to eradicate currently existing algorithmic biases. On an international level, this means enabling access to computing resources for low-income countries and fostering collaborations and research across different countries.
AI-Based Management Might Be Unhealthy
Before you get overly enthusiastic about the findings by the World Economic Forum, though, consider that economic growth doesn’t mean more riches for everybody. At the risk of sounding overly socialistic and European (it’s OK, that’s where I’m from) let me say that workers’ rights are routinely squeezed to death by AI.
The Verge reports this clearly: It’s OK to use AI to take over boring and repetitive tasks. But what if you promote an algorithm to a middle manager? The algorithmic manager is fixated on KPIs and has no mercy or humanity. The machine’s only goal is to achieve desired results as efficiently as possible. The result is that it cuts breaks, raises objectives and might end up laying off workers.
We’ve already seen negative results from AI management. Workers routinely burn out under AI managers. Some even get severe physical injuries. AI-managed productivity metrics also don’t seem to be effective. For instance, algorithms are used to measure the productivity of software developers. Paradoxically, that measurement doesn’t seem to increase productivity, but rather destroy it.
If we’re employing algorithms not as lowest-level workers but rather as managers, we should at least teach them some human decency. For example, include some hard borders in an algorithm that specify how many breaks an employee is eligible for. Also, even without AI, it’s common practice to animate workers to outperform the best in the team. But how about animating them to just be slightly above average and implementing that goal algorithmically?
These kinds of approaches may sound like a waste of resources in the short term, but my intuition is that they will pay off eventually. In fact, if we don’t adopt a more moderate approach toward productivity, we’ll likely end up with swathes of job-, income- and futureless workers, not because they’ve been laid off but because they’ve been drilled beyond their breaking points.
It’s incredibly important to emphasize the capabilities of AI in creating new jobs. These, however, must be humane, or it’s a moot point.
How to Prevent an AI Apocalypse
I’m not raising the point of social equality because I’m a guardian angel. I’m just a regular, egotistical human. But I believe that, as humans, we profit if others around us are doing well. When unemployment goes down, that saves the government a ton of money and opens up new opportunities in the future. After all, maybe one or two of these low-level workers whose job is saved will one day be the Steve Jobs or Elon Musk of the decades to come.
And there’s another issue: If, like the World Economic Forum suggests, 97 million jobs get created through AI and only 85 million jobs get destroyed, there’s no guarantee that the 85 million laid-off workers will be in the pool of the 97 million lucky ones. Populations grow, people migrate, and so the 97 million jobs may be quite a scarce resource after all.
In order to give the 85 million laid-off workers a perspective, it’s imperative to re- and upskill them, ideally long before they lose their old jobs. AI can be part of the solution here since a learning bot could automate this curriculum of continued education. AI algorithms should also be created to match workers to new opportunities and to equip brand-new workers with the necessary skills, ideally even before they’ve left college.
AI will create more jobs than it will destroy. But we’ll need AI to ensure a fair distribution of these new jobs.
Invest in Your Workers
Luckily enough, governments are starting to understand what’s going on. France, for example, takes the AI-led world of tomorrow very seriously. In a report from 2019, France Stratégie proposes to train more AI experts and workers that really understand what AI means for tech, law, economy and society as a whole.
In addition, France Stratégie proposes “improving schemes for safeguarding career paths in the sectors and subsectors that are set to be heavily impacted by the risk of automation.” In other words, some smart decision-makers in business and government need to get together and map out plans for workers in a rapidly shifting landscape, sector by sector. I applaud this approach.
This requires a lot of foresight and nobody has a crystal ball. We’ll make mistakes. But if we, as a society, want to move forward, we need to address the fundamental fear of decline, of not keeping pace with shifting technologies. AI will bring prosperity to many and doom to few, but we need to act now to curb the doom and spread prosperity to every single worker.