A robot with a sword is the worst nightmare of most movie goers, but one robot with a sword is a sign of positive advances in industrial robots. Yaskawa has a robot called Motoman, and it can learn advanced and complicated movements just by watching an expert perform them.
As an example of what Motoman can do, Yaskawa pitted it against one of the leading sword experts in the world, Isao Machii, a five-time world-record holder. First, they got Machii to show it how to do basic sword movements. They used a rig that looks similar to a video-game motion capture setup that allowed the master's 3D movements to be recorded. The robot than "watched" those movements and emulated them. It was able to do it with equal or better precision and tirelessly. As you can see from this video, that wasn't the case for the human.
But this isn't just cool (but make no mistake, it is cool). It is also a better way to program industrial robots than can be found now on the market. It could lead to saving time and money in programming (and re-programming) industrial robots.
Traditionally, industrial robots have been hand programmed with lines of code. This is a painstaking and difficult process. Recently, we've improved that with the use of "learning pendants." These are devices you put on robots and guide them manually through the process of their work. The pendant writes the code to mimic the actions, and then code can be written or edited to perfect the movements. Here's a video of that process.
You can see how that process is better than hand-coding but slower than watching a sword master do his thing.
[ Want a reason we should let robots rule? We've got 10. Read 10 Reasons Robots Should Rule The World. ]
More recently, the industry has gotten better at leading the robot through its paces in what is called "lead by the nose" programming. Industrial robots like Baxter can be taught to perform a task by hand (without the pendant). It represents a major leap forward but still requires guiding the robot. Here is how it works:
By allowing the robot to "watch" digitally and then perform the function, you save time and gain precision. The Baxter video is from 2012, and it has gotten better since then. And it is not the only "lead by the nose" robot out there. And Motoman isn't the only robot that can learn by watching. A DARPA-sponsored project at the University of Maryland learned to cook while watching online videos.
But the basic advantage here is the improved programming. The cost of owning an industrial robot is four times the purchase price of the robot. This comes in maintenance but also in the cost of programming and re-programming the robot. If you can shorten the time to program, you can improve the ROI on the robot. You can also decrease downtime and increase flexibility. You can more easily move robots in the plant to where they are needed without having to go through major reprogramming, allowing them to be used where they are needed and not where they know what they are doing.
Of course, it doesn't take a rocket scientist or a robotics engineer to see the societal problem here. Basically, humans will be teaching their new robots how to do their job -- faster, more precisely, and tirelessly. Not the best way to go out, I'm sure. At the same time, the goal of any technology is to free the human for tasks that require knowledge and experience that an AI can't handle.
Faster human programming for industrial robots is a genie that can't be put back in the bottle. It is getting faster every year. So we might as well be impressed. Most industrial robots do far less precise movements than cutting a peapod in half. Without trying to sound overly dramatic, this is the last step. we needed to really reduce the role of human on the assembly line to supervisor and repair staff. In many ways, it might be more dangerous than a robot wielding a sword, but also more useful.