Autodesk chief Andrew Anagnost, whose software firm is all for more automation, told CNBC that “some aspects of automation are going to eliminate certain jobs [so] what we have to focus on is what jobs are going to be created.”

Those jobs, he said, would be in building, programming, deploying, and maintaining automated systems.

Over the last few years, many people have been alarmed at the job-destroying implications of automation. Of course, this has been a recurring fear since the “Luddites” were smashing textile machinery in early 19th-century England, if not before, but relatively recent advances in robotics and machine intelligence have brought the issue to the fore again.

A seminal piece of research here was a 2013 study by Carl Frey and Michael Osborne of the Oxford Martin School, which suggested that 47% of U.S. jobs were at risk of automation in the near future. Other research has provided similar predictions, and the National Bureau of Economic Research also published a paper this year that suggested each new industrial robot replaces around six human workers—although, given the relatively low uptake of such machines in the U.S., the overall effect remains low for now.

However, the key question here is whether automated systems can create enough new jobs to offset those they destroy. There’s also the question of whether people who have lost their jobs to automation can be effectively retrained to cope, or whether they will simply be left behind by the march of progress.

Treasury secretary Steve Mnuchin may have declared earlier this year that the issue is “not even on [the White House’s] radar screen,” but it could have a serious political as well as economic impact.

Last month, the Oxford Martin School released another piece of research that suggested that the effects of automation may have even helped swing last year’s U.S. election. “The support for Donald Trump was significantly higher in local labor markets more exposed to the adoption of robots,” they wrote. “Other things equal, a counterfactual analysis shows that Michigan, Wisconsin, and Pennsylvania would have swung in favor of the Hillary Clinton if robot adoption had been two percent lower over the investigated period, leaving the Democrats with a majority in the electoral college.”

“Our study suggests that automation has been the real cause of voters concern,” said Frey, one of the paper’s authors. “The prime victims of the ‘Robot Revolution’ want anything but the status quo. The populist rebellion in America, Europe, and elsewhere, has many causes, but workers’ losing out to technology is seemingly the main reason.”