Now the obvious knee jerk reaction is well you own a robot and you’re responsible for

everything that it does. But as these devices become much more autonomous it’s not at

all clear that that’s really the right answer or a good answer. You go out and you buy a

great new robot and you send it down the street to go pick you up a Frappuccino down at Starbucks

and maybe it’s accidental but it’s standing at the corner and it happens to bump some

kid into traffic and a car runs the kid over. The police come and they’re going to come

and arrest you for this action. Do you really feel that you’re as responsible as you would

be if you had gone like this and pushed that kid into traffic? I would argue no you don’t.

So we’re going to need new kinds of laws that deal with the consequences of well-intentioned

autonomous actions that robots take. Now interestingly enough there’s a number of historical precedents

for this. You might say well how can you hold a robot responsible for its behavior? You

really can actually and let me point out a couple of things.

The first is most people don’t realize it. Corporations can commit criminal acts in dependent

of the people in the corporation. So in the Deepwater Horizon Gulf coast accident as an

example BP oil was charged with criminal violations even though people in the corporation were

not necessarily charged with those same criminal violations. And rightfully so. So how do we

punish a corporation? We punish a corporation by interfering with its ability to achieve

its stated goal, make huge fines as they did in that particular case. You can make the

company go out of business. You can revoke its license to operate which is a death penalty

for a corporation. You can have it monitored as they do in antitrust cases in many companies.

IBM, Microsoft I think have monitors to make sure they’re abiding by certain kinds of

behavioral standards. Well that same kind of activity can apply to a robot. You don’t

have to put a robot in jail but you can interfere with what it’s trying to do. And if these

robots are adaptable, logical and are learning. They’ll say well I’ll get it, you know.

I can’t do that because my goal is to accomplish something in particular and if I take this

particular action that’s actually going to be working against my interest in accomplishing

that situation. So rehabilitation and modification of robot

behavior just as with a corporation is much more logical than you might think. Now another

interesting historical precedent is prior to the Civil War there were a separate set

of laws that applied to slaves. They were called the slave codes. And slaves were property.

But interestingly enough the slave owners were only held liable under certain conditions

for the actions of their slaves. The slaves themselves were punished under if they committed

crimes. And so we have a historical precedent for the kinds of ways in which we can sort

this out so that you are not in constant fear that your robot is going to bump into somebody

and you’re going to go to jail for 20 years for negligent homicide or whatever it might

be.