In 1900, 41 percent of the U.S. workforce was employed in agriculture. One hundred years later, that share had declined to 1.9 percent. Over that interval, the jobs that were easy and cheap to mechanize were mechanized, and now we are left with a handful of jobs that machines find extremely difficult to do. Machines can’t make strategic decisions about which crops to grow, and as a general rule they can’t fix themselves, so that leaves a significant role for managers and mechanics. Until recently, machines were also really bad at doing things like picking heads of lettuce and other delicate crops, as this requires a deftness of hand and an attention to detail that machines lack.

This is why the agricultural sector continues to have an appetite for less-skilled labor, which has been a huge driver of the recent comprehensive immigration reform effort. The idea is that because native-born Americans will never pick cucumbers — or at least because they will never pick cucumbers at a wage that would make for affordable cucumbers — we need a steady supply of less-skilled, low-wage workers to keep farms that grow cucumbers and lettuce and other delicate crops viable.

Now, however, a number of innovative firms have developed machines that use sophisticated sensors and an enormous amount of raw computing power to do jobs that had once been beyond the reach of machines. The reporters Gosia Wozniacka and Terence Chea recently described a Lettuce Bot that can “thin” a lettuce field in the time it would take twenty workers to do the same. Though the Lettuce Bot and machines like it remain expensive, there is every reason to believe that prices will fall. These picking machines are not quite good enough to pick fresh-market fruit, but they’re getting there. The reason these machines are being developed is the same reason agribusiness interests have been agitating for a substantial increase in less-skilled immigration: the supply of workers willing to work the fields is not big enough to keep wages extremely low, and so farms have been desperate for low-cost alternatives.

A similar desperation drives technological innovation across the U.S. economy. Consider one of the most exciting developments of recent years, Google’s self-driving cars, which promise to liberate human drivers from drudgery. Of course, another way to liberate oneself from having to drive is to hire someone else to drive for you. Plenty of city dwellers hire taxis from time to time, but personal chauffeurs are beyond the reach of all but the most affluent Americans. This would be less true if, for example, Congress decided to create a special visa for would-be chauffeurs from the developing world, designed to welcome every able-bodied woman and man willing to drive a car for the minimum wage. To be sure, hiring a personal chauffeur would still be expensive, particularly if you intend to provide a wide range of fringe benefits. But the demand for self-driving cars would presumably be much smaller than it is without this steady stream of low-wage chauffeurs. Limiting the supply of would-be immigrant chauffeurs has made life somewhat less convenient for those of us who’d welcome being driven around, whether by a machine or by a living, breathing person, but there is good reason to believe that technological innovation will get us to something like the same happy outcome.

The fact that technological innovation is spurred by labor scarcity isn’t in itself a reason to oppose less-skilled immigration. But it is a useful reminder that a modern economy is a dynamic system, and that the current structure of the labor market is not set in stone. As we debate immigration reform, it is important that we keep this in mind. Jobs that are currently unattractive to workers with a high school diploma or more might become more attractive to them as mechanization drives a reduction in mid-skilled jobs, thus undermining the notion that there are jobs that Americans won’t do. And jobs that have traditionally been done by less-skilled immigrants might be rendered obsolete by machines like the Lettuce Bot.

Frank Levy and Richard Murnane, two of America’s leading experts on how technological change is shaping the labor market, have just released a new report sponsored by the center-left think tank Third Way, “Dancing with Robots: Human Skills for Computerized Work.” Their basic premise is that as machines get better at performing routine, structured, rule-based tasks, workers will have to focus on tasks that exploit the flexibility of the human mind. Levy and Murnane find that in occupations subject to computer substitution, most of which are found in the middle of the skill distribution, the number of jobs has been growing more slowly than in occupations that are not at risk of computer substitution. They project that this trend will intensify in the years to come. This spread of computerized work has, in their view, contributed to the deterioration of the labor market position of less-skilled and mid-skilled workers. This development, in turn, appears to have contributed to declining marriage rates and family formation among the non-college-educated. The end result is that even as education becomes more important, a growing share of children are being raised in households that aren’t giving them the foundational skills they need to flourish academically.

Levy and Murnane end “Dancing with Robots” with a call for increased investment in human capital, with an emphasis on early childhood education and career and technical education for students who aren’t bound for college. Their prescriptions are interesting and worthwhile. Yet I was struck by the fact that Levy and Murnane made no reference to less-skilled immigration, as one implication of their work is that future technological developments threaten to further undermine the relative position of less-skilled U.S. workers, and that this in turn will have knock-on effects on the children of those workers. If this is indeed the case, is now the ideal time to welcome a new wave of less-skilled immigrants, many of whom will struggle to adapt as computerized work continues to spread?