Since the Industrial Age, automation has shifted the kind of work that humans must do. And with the current AI boom, anxiety over employment is at an all-time high. Will new technology lead to mass unemployment? Or will a whole new set of jobs replace what we have today?

Of course, the answer is nuanced, and both arguments have elements of truth. But while AI will undoubtedly create new jobs, the technology will hurt workers in lower skill jobs, according to Moshe Vardi, a computer science professor at Rice University and a leading expert in AI.

As Vardi pointed out, manufacturing is actually at an all-time high. “It’s not that we have not lost manufacturing, it has just changed,” he said. “We are still manufacturing a lot in this country, but it has changed and uses 8 million fewer manufacturing employees.” Vardi talked to TechRepublic about why, in particular, manufacturing was the “sweet spot for automation.”

Two different schools of thought exist when it comes to the impact of automation on jobs, Vardi said.

“There is a school of the thought that very much believed that the economy is self-correcting, and if you automate something you should lower the cost,” said Vardi. “If you lower the cost, demand should increase. If demand should increase, you will need to hire more workers.”

But even economists, he said, think this view is overly simplistic. “When you lower the cost, demand will increase,” said Vardi, “but it’s not clear that it will be enough to compensate for it. It’s really impossible to predict the result of automation on the local basis.”

Vardi raised the example of ATMs. “ATMs is the one example where people thought there would be fewer tellers–but it turned out, we lowered the cost of operating the branch so much the banks opened more branches. And so we have hired more tellers,” he said.

What did take jobs away, Vardi said, is the birth of mobile banking. “When do you go to a bank?” he asked. “Never. Technology will kill these branches. The ATM did not, but other technology will ultimately kill many of the branches.”

Fintech, said Vardi, is particularly ripe for automation. “The tech community looks at banks as a big, fat, lazy target,” he said. “They have obsolete business models, very high services, and nobody loves their bank. The one thing that banks do have is the trust element right, and deposit insurance.” Still, said Vardi, “banks are going to have the fight of their life.”

New age of AI

So is this age of AI different, somehow, than previous periods of automation?

“People always need to move to something that they do better than machines,” Vardi said. “When people lost their jobs in agriculture, they moved to manufacturing. We had machines in manufacturing, but they had to be operated.”

Vardi calls manufacturing a middle-skill job. “When the capability of the machine went up and the cost went down, operating industrial robots is now $15 dollars an hour, so we don’t need so many people,” he said. In terms of the idea of bringing back manufacturing, Vardi said he doesn’t “know any economist who takes it seriously. We can try to think of how to have an economy where we encourage job creation, but manufacturing jobs are not coming back.”

Instead, said Vardi, “we have higher value kind of jobs, in higher value industries. The question is how do we get the workers to adapt to such industries?”

I asked Vardi what he thought of the argument that technology will create new jobs–like when the creators of Flippy, the burger-flipping robot, say that the robot will leave room for employees to do new, creative tasks.

“I think this is the propaganda of the creative class,” said Vardi, “that says, ‘Look, we get rid of the tedious jobs so you’ll be able do new jobs, interesting, creative jobs.’ But what jobs? What are these new creative jobs? So people give you example: Data analytics expert, data visualization. I said, ‘Okay, who gets these jobs?'”

The point, said Vardi, is that the conversation is mostly confined to the educated class.

“What are the largest corporations in the United States?” asked Vardi. “Let’s look at the top six. There’s one non-tech company there. It’s Exxon. The rest are Apple, Alphabet Google, Microsoft, Amazon, Facebook. What is the total market capitalization of these companies? Between them it’s about two and a half trillion dollars,” he said. However, these companies only employ half a million people. And most of those employees are highly educated.

With new employment trends, said Vardi, you need to ask three things: Are we creating enough new jobs? Are we creating them fast enough? And, what skill level do the new jobs require?

As AI gets smarter, said Vardi, the definition of what is “routine work” expands. One example is a decision whether to grant bail, which can now be automated using troves of data around recidivism. “We are defining things that used to be creative decision making,” said Vardi, “and saying no, it’s routine.”

“If machine can do it, it’s routine,” Vardi added. “AI is a moving target. We always call AI what we don’t know how to do yet.”

Many things we use in our daily lives used to be considered AI, Vardi said. “And now, we’re saying those things are just an algorithm.”

Also see…