Growing up in Beijing, roboticist Cong Wang had one skill he just couldn’t master: soccer. His high school’s soccer coach had played the sport professionally, yet he struggled to teach Wang how to kick the ball between the posts or even just to stop the ball with his chest.

“When he was explaining things they were so simple, but when I tried them it was out of my capability, no matter how smart I am,” Wang tells Inverse. “Whatever he taught me, I wouldn’t be able to do it. Because we were physically different. The tricks he was talking about were just beyond my mechanical capabilities.”

It’s the same basic problem Wang now tries to solve while working in his robotics lab at the New Jersey Institute of Technology. Instead of designing robots that are built to do one task perfectly, he and his fellow researchers want to do to a robot what his soccer coach eventually did for him: teach an entirely new skill to something that wasn’t specifically made to perform it.

While robots that can master the martial art of nunchaku as proficiently as, say, a talented toddler, might just seem like a mere novelty or set-up to a joke, there’s a lot of real-world application to this technological development, especially when it comes to making manufacturing more efficient.

For instance, robots that can learn to swivel their arms could open up a new frontier for robotics. Even though dismal forecasts about automation keep coming — in spite the rare bit of good news — developments like the one by Wang and his team signal another step toward more dexterous robots that could take on jobs performed by humans today.

And as his team describes in a pre-print paper available on arXiv, they decided to teach a robotic hand how to flip nunchucks. Wang and his students learned this relatively basic karate technique, then they used motion-reading sensors to demonstrate all the necessary moves to the robot, so that it could figure out how to replicate it.

For those unfamiliar with the finer points of martial arts, nunchuck-flipping isn’t something you need to be a karate master to perform.

“It’s not that challenging,” Wong says. “But it did take me two months myself to learn it.”

The robot didn’t need quite as much time, he admits.

“Oh, they’re fast — in hours, they’ll be fine,” he says. “It’s a machine, and they practice hard.”

Why It Matters

While this particular feat might seem a mere novelty — there’s only so much demand for nunchuck-flipping robots, at least for the time being — Wong sees the robot’s ability to learn and handle such unprecedentedly complex tasks means this opens up the final frontier of automation.

Industrial robots in Germany.

He points out that the construction of car bodies has been fully robotized since the 1970s, but even today the final interior assembly needs to be done by humans.

“The tasks require a lot of hand manipulation, a lot of fine motor skills, and a lot of handling of composite objects that are partly soft and partly rigid,” he says, all of which are tasks robots like this are now able to handle. “So our vision is with our technology in the future, those tasks can also be robotized.”

It’s not that robots couldn’t already do a task like flip nunchucks. As Wong notes, there are already robots that play table tennis, pitch tents, throw darts, run alongside humans, and so on.

“But the problem is a lot of those works are very case-specific,” he says. “It takes a whole team of researchers years of engineering just to do that particular application. So what we’re thinking is we can transfer human skills on this level to robots.”

A task simple enough for a human, but not to robots... until now.

Let’s say you wanted to automate apple picking. It’s a task that might seem relatively straightforward to a human, but Wang says it would take a whole team of Ph.D. researchers to replicate all the precise hand-eye coordination and careful handling of the fruits themselves that go into the task.

“However, if we can figure out how to teach a robot how to do this level of skills intuitively, just like teaching another human being, it just takes another apple picker to teach the robot,” he says.

What all that means for the future of human labor remains to be seen — this is the kind of innovation that makes the calls of tech entrepreneurs for a universal basic income seem that much more urgent. But for now, Wang’s robot remains in the lab, ready to add new tricks to its repertoire.