Robots learn to deceive through digital evolution
To study the evolution of communication, researchers at the Ecole Polytechnique Fédérale de Lausanne in Switzerland, pitted hundreds of robots against each other in a fight for survival.
The robots […] were placed by their designers in an arena, with paper discs signifying “food” and “poison” at opposite ends. Finding and staying beside the food earned the robots points.
After each iteration of the trial, researchers picked the most successful robots, copied their digital brains and used them to program a new robot generation, with a dash of random change thrown in for mutation. [Wired.com]
In the beginning, the robots mostly rolled around the arena flashing lights randomly. But soon robots learned to follow the flashing of lights, knowing that other robots most likely would be where the food is.
But space being tight around the food, not all robots could stay close to the food, so robots learned to not attract attention by flashing when close to food. Consequently, some even evolved a repulsion to flashing lights.
While deceptive, they’re not really Decepticons. To cute.