A surgeon watches as the delicate precision of the Da Vinci surgical robot is demonstrated at University Krollwitz Hospital, Halle, Germany © REX, Action Press

The jobs might not be coming back.

Computer productivity has defined the workplace of the twenty-first century. Artificial intelligence and robotics allow businesses to replace more and more workers with computer agents, not only increasing the pace of automation but also expanding it into fields previously thought "safe." Professionals such as doctors and lawyers are watching helplessly while AI takes over the routine functions of their jobs, leaving many to wonder what fields are still safe for someone who just wants to make a good living.

For centuries economists have largely agreed on the impact of technology that improves productivity, allowing businesses to produce more with less staff. Although in the short run it causes layoffs, over the long term an economy as a whole reliably grows and more jobs open up than were disrupted. This issue, studied since the Luddite Rebellion, has been well understood for several hundred years. That may now be changing.

A recent study by the Pew Research Center has found a new split among economists on whether technology will continue to be able to replace the jobs that it destroys. Today's artificial intelligence revolution may be different, many researchers argue in that it may destroy jobs more efficiently than we can replace them.

And that's a scary thought.

The relationship between productivity and jobs

The traditional relationship between labor and technology has been a positive, if contentious, one. When new devices improve productivity, companies can maintain the same output with fewer people, so they lay some off. Although this leads to temporary unemployment, it also improves incomes for the remaining workers and management, who go out and spend their extra money. This creates jobs elsewhere, and the combined productivity of these new jobs along with the new technology means that the economy as a whole ends up both producing and consuming more overall.

In other words, the pie gets bigger and increases everyone's slice with it. As Professor Donald Grimes, an economist with the University of Michigan, wrote:

"Think of tractors and all of the other agricultural productivity gains. We are now producing far more output with a small fraction of the number of farmers that we used to have. Or of the productivity effect of the assembly line. And in terms of IT, think about how spreadsheet programs reduced the number of 'book keepers.' Or how the internet allowed people to buy their own airline tickets, thus eliminating lots of travel agency jobs.

"In terms of the big picture these productivity gains raised the incomes of the people who continued to work in these professions, and/or made the owners of the machinery richer, or helped to make airline tickets cheaper. When people spent their extra money they created jobs in other industries and the people who lost their jobs were re-employed in new jobs. Over time these productivity gains raised our wages and our standard of living."

This model makes two assumptions, however: First, that we all share in the profits of increasing productivity, and second, that laid-off workers will have other industries to move into. Both of these assumptions may be falling apart.

Is this time different?

"Automation is Voldemort: the terrifying force nobody is willing to name," said Jerry Michalski, founder of think tank REX in the Pew survey. "We hardly dwell on the fact that someone trying to pick a career path that is not likely to be automated will have a very hard time making that choice. X-ray technician? Outsourced already, and automation in progress. The race between automation and human work is won by automation..."

As the pace of automation has increased over the past 10 years, replacing checkout clerks with fussy touch screens and video stores with Netflix (NFLX) queues, employment has tumbled. While that part is normal, unlike past revolutions those jobs don't seem to be coming back.

"There's always that issue, is it different this time?" said Professor Allen Sanderson, an economist with the University of Chicago. "There's been some debate among economists, not only about why the unemployment rate hasn't dropped back to 5 percent, but is 6 or 7 percent going to be the new normal? [And] related, why haven't we returned to the 3 or 4 percent growth rate that the U.S. has enjoyed over the last several years?"

Employment continues to struggle, and the two major factors that drive growth in the wake of a technology revolution seem to have simultaneously failed. Worker shares in the new economy aren't growing in proportion with productivity, when they grow at all, as the well-documented income stagnation nationwide has shown. In fact, according to the Bureau of Labor Statistics, for two of the past four years hourly compensation has actually fallen on average.

Far from enjoying greater purchasing power with which to create new jobs economy-wide, the average American has only as much or even less pocket money than before. If we can't spend more, we can't replace newly-obsolete positions with growth elsewhere.

Not that there may be any safe places left.

As Michalski alludes in his quote to Pew, one of the signature features of this age is not only the pace but the breadth of automation.

In combination with outsourcing, increasingly the question has become not which fields are falling victim to new technologies but which ones aren't. Graduates of the last ten years have seen this firsthand, as the jobs most commonly targeted for automation are entry level positions performing the mundane tasks.

The data backs up these anecdotal experiences, which will only get worse as computers become increasingly capable of simulating human judgment. A 2013 study out of Oxford University estimates that nearly half of American jobs could eventually be performed as well or better by a computer.

Already historically professional jobs, ones thought to require a human decision-maker, have proven vulnerable to computers. Someday, perhaps soon, a computer may even be the one assigned to write about the automation of journalism.

Ultimately, any job that boils down to taking in data, processing it and turning around a finished product may find itself vulnerable. As Oxford's researchers point out, this describes a disturbing percentage of the modern workplace.

More from MainStreet