Frey and Osborne’s estimates cover about 138 million Americans’ jobs. Moenius and his colleagues found that Las Vegas, Riverside, and El Paso all had high numbers of office and administrative-support jobs, food-preparation and -serving jobs, and sales jobs, and thus had the most vulnerability to automation. Moenius estimates that 65.2 percent of jobs in Las Vegas, 63.9 percent in El Paso, and 62.6 percent of jobs in Riverside are susceptible to automation in the next two decades. The automation of transportation and material-moving jobs also contributed to the potential job loss in these places, as well as in Greensboro, North Carolina, where 62.5 percent of jobs are susceptible to automation.

The jobs that the Redlands analysis places new focus on are slightly different from the types of jobs academics once thought would be easily automatable. That’s because before the Frey and Osborne study, scholars had predicted that routine jobs were the most likely to be automated, but Frey and Osborne suggested that advances in computerization have made it likely that non-routine jobs will be automated, too. The power of machine learning means that programmers with large data sets can use them to make machines smarter, allowing them to do non-routine tasks; for example, oncologists are using data from medical journals and patient records to automatically create treatment plans for cancer patients. “It is largely already technologically possible to automate almost any task, provided that sufficient amounts of data are gathered for pattern recognition,” the authors write.