Gigaom AI Minute – October 17

Transcript

Isaac Asimov famously posed his three laws of robotics, designed to protect people from robots. The first one is that a robot can never hurt a person, or, by an action, allow a person to be hurt. The second one is that a robot must obey a human at all times, unless it conflicts with the first law. And the final one is that the robot must protect itself, unless that conflicts with the first or second laws.

People point out the famous trolley problem where the robot has to choose whether to drive its occupants off a cliff or run over somebody in the road, shows where the laws break down. And this is true, but it's often argued that that's an extreme case, and how many times does that really happen?

But in reality, these three laws don't get us very far at all, because the concept of harm is always associated with a risk. So for instance, should a robot allow you to kill yourself? Well, you could say no, that's you coming to harm. But should the robot allow you to hang glide, or skydive, or for that matter, get a second helping of dessert? Should it allow you to smoke cigarettes? Should it allow you to indulge in any behavior that there is any statistical possibility that you could come to harm?

And this is why, these laws, while they're great as a set up for fictional stories, don't really help us all in the real world.