From a robot , artificially created, one would expect it to always tell the truth. But automata can also learn to tell lies, a new Swedish study shows. The engineers Sara Mitri and Dario Floreano, together with the evolutionary biologist Laurente Keller, decided to do the test with 100 groups of 10 simple robots to which they assigned the task of “looking for food” in an area controlled under a series of rules, such as lighting a blue light when they found the “food” to warn their companions. The artificial brain of these robots consisted of 14? Neurons? with 33 connections or “synapses”. After 100 laps, the robots that had scored the most points went on to the next round.
As the tests progressed, the researchers found that those who got the most food kept their location a secret (they didn’t turn on the lights). And what is more surprising, that from the 50th generation on, some of them sent the signal in areas where there was no food, to confuse the rest.
“It’s evolution and natural selection ,” suggest the authors, who believe that conflicts of interest can favor individuals in nature who suppress or misrepresent information .
