Remember that 2017 study by McKinsey that claimed that as many as 30% of global workers—some 800 million people—could be displaced by automation by 2030?
The threat that automation poses to humans, says Erik Brynjolfsson, director of the Stanford Digital Economy Lab and a senior fellow at Stanford University’s Institute for Human-Centered Artificial Intelligence, has been greatly exaggerated.
Brynjolfsson’s academic research and books examined our future working relationship with machines, and he continues to explore how advanced technologies, especially machine learning, change economies as well as workers’ jobs and lives.
In an interview with Workflow, Brynjolfsson shared his insights on these fast-evolving issues. Here are edited highlights of the conversation.
Give us an update on the big question: Are the robots still coming for our jobs?
I certainly don’t think that we’re facing mass unemployment or the end of work anytime soon. What we’re seeing is a massive transformation of work. There are multiple ways technology affects employment and wages. Substitution, or replacing a human with a machine, is the one most people focus on. But that’s just a small part of the story.
You can also use technology to complement humans, making them more valuable. There’s a lot of evidence throughout history that complementing has been more important than substituting. One way to see that is the fact wages have gone up over the past couple hundred years, implying that human labor has become more valuable, not less. A big part of that is because we have better tools to work with. You can do more with a bulldozer than with a shovel.
A third option is reengineering. Not just doing more with existing jobs, but inventing entirely new ones. In the long run, that’s probably the most important one.
How do we encourage more complementary human-machine teaming?
These choices can be made by CEOs and by engineers. I gave a keynote at the International Conference on Learning Representations, one of the main AI conferences, exhorting technologists to think about complementing rather than substituting.
But government policymakers also have a lot of influence. Right now we tax labor much more than we tax capital. One of the first laws of economics is when you tax something, you get less of it. So we are basically discouraging entrepreneurs from coming up with ways to use technology to employ people and encouraging them to substitute labor. It doesn’t seem like an intentional policy goal, but that’s what we’re doing.
What does the latest data say about which occupations automation puts most at risk?
My colleague Tom Mitchell and I looked at approximately 18,000 individual tasks related to 900 occupations, ranking each as to their suitability for machine learning. With the exception of airline pilots, who are pretty vulnerable to automation, low-wage occupations like cashier are more likely to have tasks that could be replaced by AI.
Every company has opportunities to use technology more widely, it’s just that some are leaping out ahead.
But while almost all of the jobs had some tasks suitable for machine learning, we did not find a single occupation where machine learning just ran the table and was able to perform every task. It doesn’t matter whether you’re a nursery school teacher, truck driver, radiologist, economist, whatever. Very few jobs will be entirely automated, but many will be transformed.
Which jobs are least likely to be automated?
There are three big categories. The first is creative work, out-of-box thinking that can’t be described by rules in advance. There’s a lot of work like that: design, inventing a new business or new product, a new piece of art or literature, scientific discovery. We don’t have machines that are good at those things.
The second is social/interpersonal skills. Most of us wouldn’t want a robot to teach nursery school or give a halftime pep talk to a soccer team. Persuading, selling, caring—all these are human skills that don’t require advanced degrees but do require high emotional intelligence.
The third category is fine motor skills. I’m a little less bullish on that one. Today if you go to a warehouse, you’ll see machines moving around the big things, but humans are still doing the pick and pack. How long that will continue, I don’t know. But the first two will be important for decades.
A recent study by researchers at Columbia and Berkeley found that larger firms that have invested heavily in AI enjoy the greatest growth. Do you concur?
Yes. We just released a paper on that topic. Most of the transformation is happening in the top 10% of firms, and that’s true across all industries, not just high tech. We’re getting this kind of ‘winner take most’ effect where a few firms are dominating the adoption of new technologies and seeing the biggest market gains from them.
[Read also The economic ROI of automation]
What separates leaders from the rest of the pack?
It’s more a matter of management taking action than of any inherent disadvantage. Every company has opportunities to use technology more widely, it’s just that some are leaping out ahead. This tends to happen a lot whenever there’s a diffusion of new opportunities.
Nine-tenths of the investment to transform a company is not in buying technology. It’s in organizational and human-capital change—reinventing processes, creating new business models, education, and training. The sooner the other companies act, the less likely they’ll get left behind.
What should we be concerned about when it comes to automation?
I don’t see AI replacing most jobs. There’s way too much work that needs to be done that only humans can do. But a common mistake made by CEOs, policymakers, and ordinary citizens is to hold onto the old jobs. America has never been successful by trying to stay in one place. The focus needs to be on reskilling towards new jobs and having a flexible enough workforce to make that happen. The more people try to hold onto old jobs, the worse off the economy will be in the long run.