The world’s 15 largest economies are experiencing a shrinking fertility rate—an omen that signals aging populations and fewer workers, according to a report in The Economist.

How will the economic powers—including the U.S., China and India—maintain or grow production while also supporting an increasing number of retired people in the decades to come?

Advances in artificial intelligence will help make up for the shortfall, predicts Usama Fayyad, executive director of the Institute for Experiential AI at Northeastern.

“It’s a way to compensate for a shrinking labor force,” Fayyad says. “With knowledge workers taking increasing shares of the job market, AI may increase the efficiency of existing knowledge workers by anywhere from 50% to 80%. It means you can get the same amount of work—or more—with fewer people.”

Fayyad defines knowledge workers as people who are “manipulating information in a call center, at an office, at a newspaper” and other settings.

In this Q&A with Northeastern Global News, Fayyad draws from history to explain why he believes people will be helped more than harmed by the emergence of AI. His comments have been edited for brevity and clarity.

Are you optimistic about the future of AI?

I’m optimistic because higher productivity means we can probably maintain a standard of living with a smaller population that is probably doing higher-value work, and less onerous and less robotic type work.

Do you worry that machines will replace humans in the economy?

What we have seen with every one of these technology revolutions is that a lot more new jobs—higher-value jobs—are created.

The typical pattern is an elevation of quality of life with additional flexibility. People don’t think about it but today we only have a five-day work week of roughly 40 hours of work. That would have been unthinkable a century or two ago and out of the question before that. 

Historically, jobs have been changing right in front of our eyes. When the automobile took off, the economy around horses—the stable keeper and saddle maker and all that stuff—went away and got replaced by higher-value jobs. The automobile industry in the U.S. replaced horses but became one of the largest high-value employers for factories, suppliers, and later cab and truck drivers and transformed the entire transport economy.

What will be the key to people succeeding in the AI world?

People ask, “Am I going to lose my job to AI?” And my answer is, “Absolutely not. But you are going to lose your job to another human who uses AI if you’re not using AI.”

The trick is figuring out how to use these new tools properly. And that’s been the case throughout history for all jobs, right? Imagine an accountant who says, “No, I’m sticking to ledger books. I don’t trust these computers.” They’re probably not in business today. 

For the dentist who doesn’t make use of digital X-rays and high-tech drills, it’s the same thing. In medicine, if you don’t leverage these tools that help you do your job and provide better care more efficiently, you’re probably going to be replaced by someone who does.

What do you say when people express fear of being replaced outright by AI?

A good portion of the economy is based on workers doing knowledge work. A lot of the work is fairly repetitive and robotic. 

What machines can do well is take a lot of that repetitive work and accelerate it. We haven’t reached the stage where full automation is possible. 

But we have reached the stage where supervised automation is practical. So that means you can ask the machine to generate an initial draft of a sales agreement. On behalf of a lawyer, it can create a sales agreement to acquire a company or an asset. On behalf of a physician, it can make an initial diagnosis suggestion. Or it can create an initial program on behalf of a programmer or an initial design on behalf of a graphic designer.

All of that robotic work enables you to work faster and get more done. But you still need the human judgment, the common-sense reasoning, the ability to create new ideas and ask the right questions. Because machines are very far from being able to do that. We need skilled humans to check and change the initial work produced by machines for our convenience.

How do you compare AI to other technological revolutions?

An example I use is accounting. The accounting of 60 to 70 years ago was about manipulating these large ledger books, and you had to have good handwriting and mental addition skills. All of those things are completely irrelevant today. 

All of accounting today is pretty much done by computers. And we have more accountants now than we’ve ever had in the history of civilization. Instead of replacing accountants, computers essentially democratized the function, so that many more businesses and departments could afford accounting—which gives accounting its higher value, right? It’s a lot less dependent on the bookkeeping side and a lot more about contributing to business tracking, measurement, finance strategy and all that good stuff.

You could think similarly about the agricultural and industrial revolutions. Originally, all the work was being done by humans. Then you started using animals and then machines, and it completely changes what it means to be a farmer. You end up having fewer farmers but their productivity is way higher and this productivity enables the emergence of new jobs in equipment production, maintenance, leasing, management, etc.

What is your fear as we develop AI?

Every technology is a double-edged sword. For every good use, there’s probably a bad use. The car is a very useful device but it can also be an extremely dangerous weapon. So we have created rules and requirements and a whole system of enforcement around operating cars. That’s how society effectively copes with potential bad uses. 

The area I worry about a lot would be putting these AI systems in place where the principle of human intervention is not possible. You fully trust the machine. You trust the algorithm. You’re basically saying that the thing can make its own decision.

My worries are not about the machine becoming clever enough to say, “I don’t need humans.” I don’t see any signs of machines getting anything that resembles an understanding or a consciousness where they develop their own sense of ego and selfish agendas. My worry is about the machine being too stupid to have enough common sense or reasoning to understand the damage it could be doing with automated decisions that impact the lives and well being of people. 

I think AI is going to lead to mostly good stuff. But every once in a while you’re going to get something you haven’t seen before in this journey to evolve the machines. That’s why we need to transition all of this into a kind of engineering that focuses on safety and doing it responsibly. At the Institute for Experiential AI, we care a lot about responsible AI and keeping the human in the loop. The science of leveraging human intervention and capturing the right data and context and how to practice AI responsibly, we think, is where the future of AI is.

Ian Thomsen is a Northeastern Global News reporter. Email him at i.thomsen@northeastern.edu. Follow him on Twitter @IanatNU.