Someone Taught Robots How to Lie


Ronald Arkin and Alan Wagner — two names likely doomed to live in infamy. In a stunning display of hubris, the men, researchers from Georgia Institute of Technology, published a paper in the International Journal of Social Robotics that detailed their foolhardy experiment to teach two robots how to play hide-and-seek. As part of the project, the scientists taught one robot to shift its direction, hide in a different location, and leave behind a false trail, a system that worked 75 percent of the time to deceive the other robot. The other 25 percent of the time, the lying robot’s strategy was correct, it just failed to tip over the right markers. Says Wagner, “The results were also a preliminary indication that the techniques and algorithms described in the paper could be used to successfully produce deceptive behavior in a robot.” Everyone understands that Skynet can start differently every time, right?

Robots Taught How to Deceive [Wired]