Elon Musk is worried that robots will take over the world — or more precisely, that machine intelligence will become so advanced, that it will realize that humans are in the way, and will eliminate them. Musk isn’t the only one who’s worried, even though many experts say it’s not a concern. Let’s assume for the moment that it’s a real possibility. An interesting question then presents itself: how? Will it happen suddenly, as the singularity arrives, and within an instant machine intelligence rockets forward and takes over? Or will it happen gradually, as machines slowly take over larger areas of the economy, until they are more or less running the world? If it happens slowly, there are some key developments that may accelerate it. In this essay we discuss three of these worrisome trends, which together seem to be facilitating the rise of the machines.
At the current time, grocery chains are replacing employees at the check-out with self check-out machines. Many years ago, banks replaced human tellers with ATM’s, and before that, gas stations introduced self-service. This is of course enabled by cost savings and increasingly intelligent machines, and it will only accelerate. The trend is for us to interact less with other people and more with machines. This is the first issue. But most people seem quite happy with this development. An interaction with a machine can feel quite normal, especially if the machine talks to us, and has pictures of smiling humans that we can relate to. This is the second problem: humans are gullible, and will happily take direction from anything, as long as it appears credible. Finally, there is the entity behind all of these machines: the grocery store, the bank and the gas station. All these are of course corporations, and in the Western world, the corporation has a special status under the law: it’s a legal person. These three developments — more machine interaction, human gullibility, and the status of the corporation as a person, may end up working together to take our society in a direction we don’t want it to go.
In the near future, it will be possible to go through your entire day without interacting with a single person. Your phone wakes you up with its alarm, then updates you with the latest gossip and news. Various machines in your kitchen prepare your breakfast. Your dishwasher and your robot vacuum clean up, while you run some errands in a self-driving taxi. None of these errands requires you to talk to a person. You get a flu shot by identifying yourself, and then putting your arm in a device which administers the shot. You get a haircut from a robot barber that remembers exactly how you like it. In the afternoon you practice tennis with an automatic ball thrower. Perhaps you relax in the evening with a robot massage, and some screen entertainment, which of course is already delivered to us by machine. Even when you do interact with a person, it’s often electronically, which means a machine is in the middle. In this case, are you really interacting with the person, or are you interacting with a machine that forwards your responses to the person at the other end? This distinction is key, since at some point the video you see will no longer be a person, but a computer generated image that is indistinguishable from a person. Most people won’t have a problem with this. This leads us to our next issue — people want to believe stories.
For a simple example of how easily humans believe a story, consider a computer mouse. Most people would say that when they move the mouse, they are moving the cursor on the screen. But in fact, nothing like that is happening. The person is moving a mechanical device, which then sends electronic messages to the computer with information about how the device is moving. The computer then generates the cursor and moves it in the same way. In fact, even that is a myth. Nothing on the screen is moving. Rather, the computer is changing the colors of pixels on the screen in such a way and with such speed that it looks like a cursor is moving across the screen. The brain is tricked into seeing movement, just like a flipbook with a series of slightly changing pictures. The result is a powerful myth that allows us to be more productive. If we believe we are moving the cursor, we can get a lot more done.
This gullibility allows human society to advance. An axe in the hand becomes an extension of the hand, enabling us to manipulate our environment more effectively. An electronic voice on our phone tells us to turn left at the next intersection, and we happily do so. But it is this gullibility that allows us to be manipulated so easily. Fake news on the internet riles us, even though an educated person should know it’s false. The most frightening proof of this phenomenon is the Milgram experiments, a series of experiments in the 1960’s that showed how normal individuals would commit torture under the right circumstances. As long as there was a believable story, they went along with it. If the entity in charge of generating the stories is trustworthy, then of course there’s no reason to worry. But governments are not always trustworthy, and nor are corporations.
Companies often manifest themselves these days in the form of a machine. Many machines we use today are just extensions of companies. Apps, websites, and ATM’s are all owned by the firms that developed them. Even a car that you own can be thought of now as an extension of its maker. Your Tesla is constantly sending information back to headquarters about the state of your car, and the state of your driving. These machines could easily be stand-alone corporations. In New York City, taxi firms sometimes own many taxis, and each one is its own corporation, to reduce liability in case of an accident. Since a corporation is a person in the eyes of the law, we may find ourselves facing a situation where a stand-alone robot must be legally treated as a person. Consider a future scenario where you’re out for a Sunday drive, and you’ve overridden the self-drive mechanism to remind yourself what if feels like to actually drive a car. You pass by a fleet of robots that repair highways. They are dropped off in the morning by the corporation that owns them, and spend their day shoveling, hammering, and pouring concrete. Each one is a legal person. Imagine you hit one of them. There will be damages to pay and blame to assign, which will be worked out through the legal system. The robot has rights of course, and is entitled to his day in court.
It’s not too hard to imagine a society consisting of a mixture of people and robots. For the most part, they will interact amicably. The people will tell the robots what to do, like “deliver my groceries”, or “take out my appendix”. But the robots will also tell the people what to do, like “wait here until this train has gone by”, or “rest for a few days because I just took your appendix out”. If the context is right, people will of course do what the robots tell them to do. In fact, today, when your phone tells you which way to drive to get to a restaurant, you’re already ceding control to a machine, but you do so because you feel like you are still in control.
With each passing year, these robots will get smarter. Initially of course, there will be a large company that owns a fleet of them. But as they get smarter, their corporate charters will cede more independence to them. They will be free to make day to day decisions, only referring to the head office if they encounter a situation they are not programmed to deal with. Individual robots, or even large teams of robots, may act independently for long periods of time — days, months, even years. Humans will grow more and more comfortable with having robots in charge. It’s not too hard to imagine this future going very badly. Bit by bit, the machines take over, and like the frog in hot water, people will be blissfully unaware of the danger.
In the 1940’s, the science fiction writer Isaac Asimov proposed a series of rules that robots must obey in order to ensure that humans remain in charge. The rules tried to guarantee that no humans would be harmed in any interaction with a robot. This is far more complicated than Asimov implied of course, but clearly some framework will be needed. The speed of recent developments in AI, together with the three issues outlined in this essay, suggest that the time has come to build this framework. Otherwise, the robots win.
