On Friday, during a conversation with Mike Allen of Axios, the newly minted Treasury Secretary Steven Mnuchin said that there was no need to worry about artificial intelligence taking over U.S. jobs anytime soon. “It's not even on our radar screen,” he told Allen. When pressed for when, exactly, he thought concern might be warranted, Mnuchin offered “50 to 100 more years." Just about anyone who works on, or studies machine learning would beg to differ.

In December of 2016, about one month before President Trump officially took office, the White House released a report on artificial intelligence and its impact on the economy. It found that advances in machine learning already had the potential to disrupt some sectors of the labor market, and that capabilities such as driverless cars and some household maintenance tasks were likely to cause further disruptions in the near future. Experts asked to weigh in on the report estimated that in the next 10 to 20 years, 47 percent of U.S. jobs could in some way be at risk due to advances in automation.

The Obama administration is certainly not the only group of experts to believe that the impact of machine learning on the labor market has already started. In a conversation earlier this month, Melinda Gates cited rapidly advancing machine learning as part of the reason that the tech industry needed to tackle its gender diversity initiatives immediately. In 2016, a report from McKinsey found that existing technologies could automate about 45 percent of the activities that humans are paid to perform. Even Mnuchin’s former employer, Goldman Sachs, believes that a massive leap forward in terms of machine learning will occur within the next decade.

To be sure, most experts agree that the impact of advancing artificial intelligence won’t be felt equally. It’s less likely that machines will suddenly be able to replace the entirety of a human’s workload, but instead, that machines will become able to perform more and more individual tasks—and eventually to solve more complex problems. But without planning and intervention, such as retraining efforts, this could create an even more stratified workforce, where only the most educated, highly skilled, senior workers have stable work. And that would have disastrous implications for an already troublesome economic inequality gap.

While Mnuchin’s estimates of how long this may take feel wildly off base, his response seems intended to further an increasingly politicized dilemma: who, or what, is responsible for taking jobs away from American workers. President Trump spent the better part of a year convincing American voters that it’s offshoring and immigrants that are responsible for the decline of certain industries and the loss of jobs that have come along with it. Foreign workers and cheap pay, not automation, are the real culprit of the job loss sweeping across the Rust Belt and former factory towns across the country, according to the GOP. This narrative persists despite the fact that research has shown it not to be the case. A 2015 study from Ball State University found that around 88 percent of factory job loss in the U.S. was attributable to increased productivity—via improvements to things such as machinery and automation—not offshoring or trade, which played a much smaller role.

While he may have been toeing the party line with his response, Mnuchin’s assertion that he is not at all concerned with the impact of automation on the American labor force is vastly out of touch with the concerns of actual American workers. Munuchin seems to believe that there’s no reason to even start worrying about machines replacing human labor until about 50 years from now. But nearly two-thirds of Americans think that by that time, most human jobs will already belong to machines.