This week scientists unveiled a robot that can sustain injury to one of its six legs, think for a few minutes, and devise a more efficient way to walk—by essentially “limping” away as fast as possible. Jeff Clune and his colleagues accomplished the feat by endowing the robot with what they call a “simulated childhood” of possible motions, and letting the robot figure out the rest.

How far can this sort of robotic thought go? According to computer scientist Ashok Goel, if robots have unlimited capability to learn, “why would there be a limit to emotional intelligence?” After all, he says, humans aren’t born with a full set of emotional and ethical intelligence—children learn it by observing adults.