By Kevin Murnane

Which box does this go in? CREDIT: MIT CSAIL/YOUTUBE

Last week I wrote about how science infuses science fiction in the Syfy network series The Expanse. Now comes a story about how the researchers at MIT’s CSAIL (Computer Science and Artificial Intelligence Laboratory) are turning science fiction into science.

The researchers created a system that allows a robot to correct its mistakes in real time when a human observer notices that a mistake is being made. The observer just sits there and watches; she doesn’t physically interact with the robot or anything else. If the observer recognizes that a mistake is occurring, the robot changes course and does the right thing.

How in the world did the researchers at CSAIL achieve this? It’s ingenious, really. EEG (electroencephalogram) data are gathered from the observer who wears a cap of electrodes that measure patterns of electrical activity in the brain. One distinctive pattern, called an ErrP (Error-related Potential) signal occurs whenever a person recognizes that either she or someone else is making a mistake. The EEG data are analyzed in real time by a classifier program that is trained to recognize ErrP signals. When an ErrP is identified, a message is sent to the robot telling it to change its course of action.

Brain-controlled Robots

The researchers tested their system as follows. (The study described in the research paper is different from the one depicted in the video.)

The observer sat across a table from a Baxter robot made by Rethink Robotics. Situated on the table between the two were two LED-equipped targets, one to the right and the other to the left. Baxter's task was to reach for a target when its LED lit up. The robot’s arm rested midway between the two targets and when the LED came on it started to reach toward one or the other. If Baxter reached for the wrong one and the EEG signal from the observer included an ErrP, the robot received a message that switched it to reaching for the other target.

Read Forbes article