1 min to read
The danger of AI
The danger of artificial intelligence isn't that it's going to rebel against us.
In movies, when something goes wrong with artificial intelligence, it is often because it refuses to obey and rebels. Nonetheless, in real life, the AI we create is not smart enough to even have the idea of rebelling. It clearly does not measure up to the human brain in terms of computing power.
So what exactly is the danger of AI? Well, when solving problems, AI does not understand the nature of what it is working on. This is a big problem since AI does exactly what we ask it to do, regardless of the consequences. It can achieve what we ask it to do without obtaining the results we expected. For instance, a team of programmers tried coding AI that learns to run as fast as possible and here is what they got :
Since the AI program does not feel pain like we do, it started doing silly walks and somersaulting in order to move faster. It did exactly what it was asked for and not what we wanted it to do. That is because we often tell AI to do the wrong thing and realize it when it goes wrong.
Moreover, image recognition AI does not understand what it is looking at and often gets confused when it sees new objects. In 2016, there was a fatal accident when somebody used Tesla’s autopilot AI in city streets eventhough it was trained on the highway. That person crashed into a truck because the car failed to brake. What happened? The AI recognized the truck as a road side and decided to drive underneath. It did exactly what we asked : “drive under what you recognize as road signs”. However, the recognition system was not trained enough.
These accidents happen all the time with AI because it can be destructive and at the same time not be aware of it.
