Robots evolve and learn to lie

May 11th, 2009 | By admin | Category: Featured Articles
Floreana and his colleagues outfitted robots with light sensors, light-colored rings and wheels and placed them in habitats furnished with bright "sources of food and other" poison. " The first recharge their batteries and other download them. Their neural circuitry was programmed with 30 "genes" of program code elements that determine the level of detection of light and how they respond to these stimuli. The robots were initially programmed to light emitor at random and move randomly to detect light. To create the next generation of robots, Floreana recombined genes that were better, those who managed to load their batteries from their "food source". The resulting code (with some additional changes in the form of random changes) was installed back in the robot to turn in what,

in essence, were descendants of the above, and put them back in their artificial habitat. "We created a situation common in nature: foraging with uncertainty," said Floreana. "You have to find food, but do not know how each food if you eat poison, you die." They were allowed to "eat", "Play" and "dying" to four different colonies of robots. For generation 50, in three of the four colonies the robots have learned to communicate by turning on their lights to warn others when they have found food. The fourth colony, however, evolved to give life to robots "cheats" that illuminate to tell others that the poison was food, while they move to the source of "food" and are loaded with it without issue or a flicker. Some robots, though, are real heroes: cautioned and died to save other robots. "Sometimes," said Floreana "in nature is an animal that emits a cry when she sees a predator, it devoured him and the others escaped, but I never expected to see this in robots."
Tags: , , , ,

Leave Comment