As lower-skill jobs are swept away by technology, economists are wondering what kinds of skills workers will need — and how the economic benefits of automation will be distributed.

Perhaps the hottest topic in economics these days is how far robots and artificial intelligence (AI) will go in replacing humans in the workplace — and how quickly the changes will unfold. For decades, technology has been replacing mid-level jobs in the United States and other advanced economies. But the pace has quickened, with the effects billowing through not just automotive factories and electronics facilities but also law offices, operating rooms, and the roads we drive on.

Some economists, including MIT’s Erik Brynjolfsson and Andrew McAfee, have argued that, in addition to eliminating jobs, technology will also open the way to new ones. In a paper last year, MIT’s Daron Acemoglu and Boston University’s Pascual Restrepo anticipated that the rate of automation would taper off and eventually encourage “the creation of new complex tasks.” But will that process occur rapidly enough? What kinds of skills will new workers need? And how will the economic benefits of automation be distributed?

A new paper by Acemoglu and Restrepo raises some concerns. So far, they say, new jobs have done little to offset the number of job losses in manufacturing (even after they adjust for losses due to trade or imports from countries such as China). For men without college degrees, this is bad news — the number of industrial robots is growing. But it’s also worrisome for people with higher skill levels. Not only is there no guarantee that the new jobs will be located in the areas where the losses have occurred — there’s no assurance that jobs that are considered “safe” will be safe for long.

At a recent seminar at MIT’s Initiative on the Digital Economy, Columbia University economist Jeffrey D. Sachs, who heads the Earth Institute, predicted that many cognitive activities that are highly valued today will lose out in the coming decades to smart machines. This would have consequences (and not necessarily positive ones) for jobs in fields such as software development and medicine — and perhaps even Sachs’ own profession of economics (where, he reckons, “at least three-fourths of what I do can be done better by a machine”). “Until now,” Sachs says, “it was labor that was being replaced.” But increasingly, it will be more skilled activities performed by people with college degrees.

The essential question going forward, says Sachs, isn’t what the new jobs will look like. (In fact, he says, “There’s too much emphasis on jobs.”) As capital replaces labor and productivity increases, he says, the critical choices will be how income and aggregate wealth are distributed, and how people spend their time. Workers in Sweden spend several hundred fewer hours on the job per year than their counterparts in the United States, Sachs notes, “and they are happier than we [in the United States] are.” Technology, he says, could increase well-being for everyone, but it won’t happen unless we do something.”