How Startups Are Getting in Their Robots' Heads

When I talk to roboticists, one scenario that keeps them up at night is the possibility of their robots making an unexpected movement and hurting someone. An industrial arm could accidentally wallop a nearby worker, a self-driving car could swerve off a road or a humanoid robot could fall on a toddler. Yikes!

Historically, a surefire way to guard against these risks was to rope off a robot in its own work zone, keeping people out of reach. But as artificial intelligence advances improve robots, these machines are increasingly working alongside people in factories, on the roads and eventually, in homes. That’s prompted roboticists to seek out new ways to understand robots’ decisions and anticipate their actions. But there’s a deepening divide over how best to do it.