A View from David Ewing Duncan
Robots Evolve to Deceive
A Swiss insect expert creates tiny robots that evolve to work together in groups–and learn to trick “outsiders.”
A bevy of seven-inch-high s-bots in Switzerland would make Machiavelli proud, though their conduct might make us humans ponder our own behavior, and that of any future intelligent machines we may create to evolve on their own.
Programmed to live lives of just two minutes, these mechanical beings work together over the course of hundreds of generations to find “food” and to avoid “poison.” The experiment, by Swiss entomologist Laurent Keller, is designed to help us better understand how social creatures evolve to communicate. Working at the University of Lausanne, Keller’s team built cute devices with wheels, cameras, ground sensors, and a programmed “genome” that dictated responses to their surroundings.
Some s-bots had blue lights they could flip on and off to signal others. Their robo-ecosytem contained small trays that looked like ashtrays with Christmas lights that were designated as food or poison. If the s-bots found the former, they could “mate”, sharing their “genome” programming so that it is passed on to the next “generation”; if not, they died off, and so did their “genes.” The idea was to simulate events that would take thousands of years or longer for honeybees or humans to develop, and compress it into about a week–which for these bots was 500 generations.
The results: the bots sharing genes learned to help one another perpetuate their “DNA,” using their lights to signal to their clan when they discovered food. When outsiders with different genomes were introduced, members of the clan sometimes blinked their lights far away from the food to draw the strangers away.
Keller and his team did not expect this level of sophistication in the bots’ communication. They concluded that kinship and the imperative of the group to survive spurred a group dynamic that included helping one another and deceiving outsiders. Other researchers are planning to build more bots to test how other forms of group behavior might have evolved.
I’d like to know what gives rise to altruism, tolerance, and collective ambition on the one hand, and mass movements, fundamentalism, and cruelty on the other. Also, a personal curiosity of mine: why is it that some humans in history follow paths that lead to their own demise or destruction, something that the historian Barbara Tuchman called the “march of folly”? For instance, what group dynamic led the Trojans to pull the Trojan Horse into their city when they had reason to suspect that it might be a trap? And why do countries and people launch or persist in wars that are damaging or catastrophic?
This robot project comes at a time when we humans are tinkering with altering our behavior with psychotropic drugs, brain implants, and other technological manipulations. I wonder if the s-bots, too, will one day develop the need for their own behavior modification should they evolve to have certain humanlike neuroses. But if this ever happens, we’ll be long gone, I suspect, having been pushed aside (perhaps deceived by blinking blue lights?) by bots who have evolved to be as glorious and as dysfunctional as we are.
Listen to David Ewing Duncan talk about s-bots on NPR Talk’s Biotech Nation in the next couple of weeks on his weekly segment with host Moira Gunn called “BioIssue of the Week”. Go here for details.
Read the Keller study in Current Biology (restricted access).
A Science magazine news story about the study (MIT library access).
Watch the robots in motion.
The Lausanne team’s website on social communication.