New York Times tech columnist Kevin Roose says we've been approaching automation all wrong. "We should be teaching people ... to be more like humans, to do the things that machines can't do," he says.
There are three categories of work that I think is unlikely to be automated in the near future. One is "surprising work." So this is work that involves complex rules, changing environments, unexpected variables. AI and automation really like regularity. They like concrete rules, bounded environments and repetitive action. So this is why AI can beat a human in chess, but if you asked an algorithm to teach a kindergarten class, it would fail miserably because that's a very irregular environment with lots of surprises going on. So those surprising jobs are the first jobs I think are relatively safe. The second category is what I call "social jobs," jobs that involve making people feel things rather than making things. So these would be the jobs in social services and health care, nursing therapists, ministers, but also people who perform sort of emotional labor as part of their jobs, people like flight attendants and baristas, people we don't typically think of as being "social" workers, but their jobs do involve an element of making people feel things. And the third category of work that I think is safe is what I call "scarce work." And this is work that involves sort of high-stakes situations, rare combinations of skills, or just people who are experts in their fields. This would include jobs that we have decided are unacceptable to automate. We could replace all of the human 911 operators with robots. That technology exists. But if you call 911 today, you will get a human because we want humans to be doing that job when we're in trouble. We want a human to pick up the phone and help us to deal with our problems.