These are the jobs increasingly at risk from automation

The twin trends of globalization and robotics — or globotics — will usher in a period of unprecedented disruption that could displace workers at the fastest pace in history, argues Richard Baldwin, international economics professor at the Graduate Institute, Geneva, in his new book, The Globotics Upheaval: Globalization, Robotics and the Future of Work.

Like factory workers who lost their jobs to automation, white-collar and service workers are now in danger of being displaced en masse, said Baldwin, also editor-in-chief of policy website VoxEU.org. He recently joined the Knowledge@Wharton radio show on SiriusXM to discuss this trend and how workers can protect themselves.

An edited transcript of the conversation follows.

Knowledge@Wharton: If automation first diminished blue-collar jobs, wasn’t it inevitable that technology would affect white-collar jobs?

Richard Baldwin: I think it is a natural, but what I would rather focus on is technology. What is really driving this opening up of office jobs and service jobs is that digital technology is getting really good. It is especially important for people working in offices, because that is really about information. Basically, you are moving information around and processing it, and digital technology has really changed the ability to do that over longer distances. Once things get arranged so you can work from home, it is not that big of a jump for somebody to work from a farther away country, charging a much lower salary.

The biggest arbitrage or gain for your employers will be to hire people who can do more or less what you can, but for a tenth of the price. Increasingly, these international freelancers or ‘tele-migrants’ are going to be in developing countries. But I think there will be a bit of a time-zone thing. Quite a few people in South America will provide services in North America, and people in Africa will provide to Europe, and Southeast Asia to East Asia, for example. That is where these tele-migrants are going to be lining up.

Knowledge@Wharton: One of the things that you bring up from a historical perspective is the slowing growth that we have seen in general. Many people will talk about it in the short term because of what we went through a decade ago with the financial crisis and the recession. But you talk about it going back to the early 1970s, when we really started to see the rate of growth slow down.

Baldwin: Growth definitely slowed down in all of the advanced economies around 1970, by about half. The U.S. was growing 3% to 4%, and it started growing 1% to 2%. Japan was growing at 6% to 8%, and it started growing 2% to 4%, and so on. But [no one] is 100% sure what caused that. The employment in U.S. manufacturing peaked right about that time, and it has been declining ever since. It was basically robot arms replacing human hands. The key there was that they invented the computer chip, so you could put a computer on a robot arm and a robot could then do lots of stuff that you used to need human hands for. It started automating away jobs.

Knowledge@Wharton: In what sectors do you see the greatest concern of job loss?

Baldwin: In my book, I talk about automation of service-sector jobs and professional jobs and globalization, or at least tele-migrating stuff. I think the easiest way to do it is just to look around your office and see who is telecommuting. Which parts of which jobs can be done without actually being in the office? Those are the ones that are going to go first. Also, I think it is important to remember that this is not about occupations as a whole, it is about tasks within occupations.

Look at the tractor. … It was a very good tool that changed the nature of the job the farmer was doing, and it meant we needed fewer farmers to do the work. But it didn’t eliminate the occupation of farmers. When you think about people in accounting or IT, or people who manipulate data online, that sort of stuff, those are the jobs that can be automated most quickly. Lots of teams have remote workers, and those remote workers are the ones who are going to go first.

Knowledge@Wharton: You say that this upheaval not only will impact our economies, but it will impact our political systems as well. How so?

Baldwin: I am not predicting we are going to have a huge upheaval. I am just saying it is not at all unlikely that we will have one. Estimates of how much job displacement will happen go from scary and super-fast to reasonable over a long period of time. Honestly, I don’t have anything to add to those experts that I surveyed, and I don’t think they really know. You are just guessing about this complex future.

But it is a serious possibility that the displacement will happen very fast, and that these people working in offices and professional jobs will join hands with the people who have been hurt by competition by China and robots in factories. We could have a mighty upheaval, something like the yellow vest [protests in France], but just much larger and much faster. That is what I am worried about.

Knowledge@Wharton: Amazon is a huge employer worldwide. Is it one of those companies where workers could be replaced?

Baldwin: Amazon is an interesting case. They basically are in the old business of ordering online and delivering. So much of that is very, very physical, and they use tons of AI to make it easier to find customers and whatnot. It’s the warehousing kind of workers in Amazon that are being replaced. Those are more like factory jobs, so that I don’t think is so disruptive. What I am thinking about is people who work in an office, say, dealing with phone subscriptions.

If I email my phone provider, which is Swisscom in Switzerland, that I want to change my subscription to allow for [my traveling to] the U.S. for the next 10 days, then there will be a human at Swisscom who opens up my email, reads it, tries to figure out what I want, and opens up one database to change my subscription, closes that, and opens up the financial billing database to change my billing.

Until very, very recently you absolutely needed to have a human doing that because the computers couldn’t read the email and understand what I wanted. But now there is a whole thing called robotic process automation, which is kind of like digital knowledge workers. The computer opens up the email, reads it, understands what I want, opens up the database to changes of subscription, closes it, changes it to national database, all without humans and 100 times faster and with fewer errors than a human. It is replacement of jobs like that which I think are going to go fastest.

Knowledge@Wharton: You do say that some of these jobs though will be sheltered at least in the short term, correct?

Baldwin: When you think about which jobs will be replaced and which ones won’t, what you have got to focus on is what artificial intelligence cannot do. There is a bunch of reviews of workplace capacity today I can do. There was a very good one done by McKinsey Global Institute last year. If you line up the capacities where AI is very good and less good, the most human tasks are the ones it can’t do. Motivating people, managing people, providing creativity, dealing with unknown situations, applying ethics — things like that require a human touch or human talent. Those are the things that AI can’t do.

This new way of computers learning to think is all based on machine learning, which is programming computers in a radically different way. When we programmed computers before 2016, you had to know step by logical step what it should do in every single situation. You were just writing down a set of instructions for the computer to follow. Now with machine learning, they don’t do it that way. They take a million observations of, let’s say, a cat face, and 10 million not cat faces, and they estimate an enormous statistical model using super amounts of power for it to guess. It uses hundreds of thousands of clues to guess what is a cat, what is not. That model is so complex that even the AI scientists don’t understand what exactly it is using to identify the cat.

That is how our brain works for many things. I can tell you how I calculated a 15% tip, but I cannot tell you how I can recognize the thing I am looking at out my window is a car instead of a bus. That new capacity all depends upon that big data set. The question has to be clear, and the outcome has to be clear. Now, think about your job. What parts of your job are the questions not clear and the outcomes not clear? That is what is going to be sheltered by AI.

Knowledge@Wharton: You specifically mention journalism. How will that be affected?

Baldwin: There are already a number of programs that are robo-journalists. They are used routinely in reporting sports scores and stock market results, and especially election results when there are thousands of news stories that come in at the same time. They have a template where there is a great big database on election results, and then this AI machine generates stories for each and every district in a very quick way. The same is true with the sports scores and the stock markets. They take data from a general feed and turn it into a story using artificial intelligence.

Knowledge@Wharton: But that industry and others still would require human thought to process some of the work, correct?

Baldwin: Almost every job has something where it requires a real human to be there. What I am trying to push in my book is that people ought to look at their own job, their own list of chores, and see which could be automated by one of these machine translation things, which could be replaced by somebody on a Skype screen sitting in the office next to you. What you ought to focus on is getting good at the stuff that neither of those can do.

In the jobs of the future, we will be doing what tele-migrants can’t, and we will be doing what AI can’t. So, we ought to think about what they can’t do and focus on building talents in things that they can’t do.

Knowledge@Wharton: This could lead to a lot of people wanting to change jobs if they believe that theirs will be automated at some point, correct?

Baldwin: Absolutely. There are jobs where this is going to come faster and sooner and harder, and those are not the jobs you want to be in. But the idea you should move out [of a job] is a good idea, it is a good thought. But when I think about it, it has got more to do with what we should be getting our children to do and train for. We want to make sure that they don’t train for jobs that are very, very quickly going to disappear. But people who have jobs, you’ve got to think about moving into different things, sheltered jobs.

Knowledge@Wharton: Should education be adjusted to accommodate these changes so that young people will be better prepared?

Baldwin: Yes, that is the last part of my book. One of the key rules about getting ready for this is, the old rules don’t work. The old rules for dealing with globalization and automation were get more skills or education and training. Almost universally in Europe, the United States, in families all over the world, they say a kid has got to get more education so they can survive and thrive in this world of globalization. The reason that worked was, essentially, globalization and automation only work in things like manufacturing and farming and mining — industries where you actually do things.

But the more education you get, the more likely you are going to end up in a profession where automation was not working and there was no globalization because of technological barriers. That was not a bad idea for the last time, but this time you are going to have to be a little bit more subtle. You can’t just get more skills, it is going to have to be, which skills. In particular, we are going to have to think about more human skills, softer skills. Of course, everybody will have to have minimum of digital fluency and literacy, but mostly young people already have that. That will be the table stakes in the future market.

After that, managing people was much less replaceable than, for example, drawing architectural plans or looking through legal documents and trying to find evidence. Those are things that robots are starting to get very good at, so just getting more education is not enough. You have to focus more on the human skills.