Prosthesis

You Can Correct This Robot's Mistakes Using Only Your Mind

News: The Curiosity Podcast is here! Subscribe on iTunes, Stitcher, Google Play Music, SoundCloud and RSS.

Technology is amazing, but it has its downsides. When new stuff comes out, it takes time to learn how to use it—we're still not entirely sure how Snapchat works, to be honest. That's exactly what's been hindering our ability to control robots with our minds. Even though we have the basics of that technology, humans need to learn how to think in order for robots to understand those thoughts. In 2017, researchers announced a project to teach robots how to better interact with humans, instead of the other way around.

Related: We Already Have The Technology To Control Electronics With Our Minds

Advertisement
A feedback system enables human operators to correct a robot's choices.

Learning Through Its Mistakes

Reading minds is no easy task, but a team from Boston University and MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) may have found one sure-fire thought that robots can detect. It should be familiar to anyone who has published anything on the internet: "Um, I believe you made a mistake." The researchers call the signals that our brains produce when they notice a mistake "error-related potentials," or ErrPs. So instead of humans controlling a robot with a specific task—for example, looking at one of two objects that each correspond to a different instruction—they just have to watch a robot perform its normal functions, and any mistake it makes should be instantly detected through ErrPs.

Related: Brain-To-Brain Communication Has Already Happened (Sort Of)

In a video demonstration of the system, a human operator wears an electroencephalography (EEG) cap designed to measure their brain signals. Across the table from her is a red robot with a friendly smile on his LCD face who grasps objects and places each in one of two bins labeled "paint" and "wire." When the robot places paint in the paint bin, the human doesn't notice a mistake and the robot continues sorting. But once it positions an object over the wrong bin, the human produces an ErrP that should make the robot figure out the right thing to do in real time.

EEG brain signals detect if the person notices a mistake.

Real-World Potential

Right now, the system isn't all that reliable. The robot's ability to identify ErrPs was only slightly better than 50/50 chance—it failed almost half the time. It should be noted, however, that when the robot ignored a brain signal the first time, the human produced a much stronger signal in response. When researchers analyzed those "secondary errors" afterward, they were able to detect them with much more accuracy. In the future, we may be able to achieve that accuracy in real time.

Related: The Allen Brain Atlas Is A Super-Detailed Map Of The Brain

Although it might sound overly simplistic to think that we could control things as complex as driverless cars with a signal this simple, binary choices like the one the researchers used are the basis for computers as we know them. Coauthor Andres F. Salazar-Gomez suggests that an interface like this one could be useful for those who can't communicate verbally, helping them spell words through a series of binary choices. CSAIL director Daniela Rus goes a few steps further: "A streamlined approach like that would improve our abilities to supervise factory robots, driverless cars, and other technologies we haven't even invented yet," she told MIT News.

Is there something you're curious about? Send us a note or email us at editors (at) curiosity.com. And follow Curiosity on Facebook, Instagram and Twitter.

Love getting smarter? Sign up to our newsletter and get our best content in your inbox!

Watch And Learn: Our Favorite Content About Brain-Computer Interfaces

Mind-Controlled Quadcopter Demonstrates New Possibilities

Advertisement