So what's being done to keep us safe?

What's being done now isn't enough," says Blay Whitby, an artificial intelligence expert at the University of Sussex in England. "It's surprising that the field of 'computer security' is only 10 years old." But things are changing, and researchers are showing increasing interest in making robots safer. One approach, which seems simple enough, is to program them to avoid contact with humans altogether. However, this is more difficult than it might seem. Telling a robot to navigate a cluttered room without considering what its limbs or other parts might encounter along the way is quite difficult.

"In the future, regulating robot behavior will become increasingly difficult, as they will have increasingly powerful learning mechanisms," said Gianmarco Veruggio, a roboticist at the Institute for Intelligent Automation Systems in Genoa, Italy. "As a result, their behavior will become impossible to fully predict, since they will no longer follow predefined paths but will learn new ways of moving."

Then there's the issue of unpredictable failures. What happens if the robot's motors stop working, or a system fails while performing open-heart surgery or serving you a cup of hot coffee? "Of course, you can reduce failures by adding redundant systems," says Hirochika Inoue, a roboticist at the University of Tokyo who is currently an advisor to the Japan Association for the Advancement of Science. "But that's no guarantee. It's impossible to be 100 percent safe in technology," says Dr. Inoue. "That's because no matter how meticulous you are, you can't outrun the unpredictable nature of human behavior." In other words, however sophisticated your robot is at avoiding people, humans wouldn't always be able to avoid it and might trip over it and fall down stairs.