Robots evolve sharing, altruism, might not be plotting our downfall

Illustration for article titled Robots evolve sharing, altruism, might not be plotting our downfall

Usually when we hear about robots learning a new skill through evolutionary programming, it makes us fear for the coming robopocalypse. But just this once it appears they may be learning something friendly, too. A program in Switzerland has put simple robots through hundreds of generations of simulated evolution, and they've learned how to share.


Altruism seems to be counter-intuitive from an evolutionary perspective, because it means helping people who won't necessarily improve your chances to survive. But there is an evolutionary logic, described in what's known as Hamilton's rule: you'll help people in your family, because they pass on your genetic material, too.

It's a model that's widely recognized but difficult to test, due the requirements of watching hundreds of generations to see if altruistic traits get selected for. The team created robots that decided how much "food" to share with the robot equivalents of clones, siblings, cousins and non-relatives. The researchers found that the more these robots shared with close relations, the more their genetic material was "passed on," mirroring Hamilton's rule. Scientists believe this could be used to design better swarming robots, which constantly need to share tasks with each other.


Maybe our new robot overlords aren't so cruel after all.

Share This Story

Get our newsletter


Corpore Metal

Seems obvious to me that when sapient artificial life gets built one day, it be just as much a mixed bag as we emergent organisms are. Some of it might be violent but others may not be.

In fact, maybe humanity won't be wiped out by some single AI bent on genocide. It might be wiped out just as an accident in a war between the AIs that want to kill us and the AIs that want save us. We'd be their version of civilian deaths maybe.