NPR’s Planet Money has delivered a remarkable look at the work that goes into making a typical T-shirt. The project includes a Mississippi cotton farmer who uses 13 people and 26 machines to produce 13,000 bales of cotton a year, which can be turned into 9.4 million T-shirts.

The numbers are stunning, and a little frightening amid concerns over how technology is eliminating jobs. Instagram famously sold for $1 billion with only 13 employees. Being a significant part of the world economy no longer requires lots of human employees. In fact, with the costs of employing humans, it’s preferable to have as few as possible.

So what’s the future for the middle class and the average worker?

Tyler Cowen, an economist at George Mason University, addresses this in his latest book “Average Is Over.” He compared future jobs to freestyle chess competitions, where humans and computers team up. Humans lean on computers for their computational powers, but overrule them when they know the machine is missing something. He explained to the New Yorker:

Think in terms of this future middle-class job: You read medical scans, and you work alongside a computer. The computer does most of the judging, but there are some special or unusual scans where you say, “Hmm, that’s not quite right—I need a doctor to look at this again and study it more carefully.” You’ll need to know something about medicine, but it won’t be the same as being a doctor. You’ll need to know something about how these programs work, but it won’t be the same as being a programmer. You’ll need to be really good at judging, and being dispassionate, and you’ll have to have a sense of what computers can and cannot do. It’s about working with the machine: knowing when to hold back, when to intervene.
Or take business negotiations. In the early stages of negotiation software, on your smartphone, there may be programs that listen to the pitch of a voice, or that test for stress. You’ll just ask the program, “Was he lying? Was he eager to do business with me?” Maybe the computer will be right sixty per cent of the time. That’s useful information, but it’s still going to be wrong a lot.