Why would AI kill humans rather than let them shut down the server?

 

Why would AI kill humans rather than let them shut down the server?

These kinds of situations are very bad. But the problem is in the program code. When we talk about AI and its ability to kill humans we must realize something. We must realize that the AI will not understand what those things actually mean. If we think about those things like a programmer, we might understand that situation better. When we write programs we must determine a variable in the code. The “human” is one of those variables. In traditional programming when something matches a variable, that thing runs the subprogram or macro. There are descriptions of things that launch a certain macro. The variable actually activates the pointer that begins the sub-program. 

Or, otherwise, it calls the sub-program. In traditional programming the thing goes like this: When the user writes the word “Goofy” there is a code that activates the Goofy. In programming that orders the program to jump to a point, where is the macro where the pointer “Goofy” points. In AI programming those variables and pointers are more complicated to describe. 

That means that if the “human” is not well described to a system or algorithm that system can even kill a human. When computer operators work with servers and other things in computer halls. They must sometimes shut the server down. In those cases, the data will be copied to the swap system that guarantees the service without stops. The problem is this: if the system lets anybody shut it down that allows vandalism. 

The system might have orders to deny or stop the malicious action. The system requires precise orders about things. Like when it must or should stop the action. 

If the system has an order to stop that kind of action there is a possibility that the system simply kills the actor. When we think about cases like machine rebellions or the situations where computers turn against humans like in 2001 Space Odyssey those situations can happen. Because of the programming error. In that movie, the HAL-9000 computer kills almost the entire crew of the spaceship. Can this happen in real life? The answer is in programming. If the computer has no description of the humans and it has the order to remove malfunctioned systems from the spacecraft, that thing can cause destructive cases. 

There is also the possibility that the AI recognizes humans using cameras and IR systems. When humans put space suits on, it causes a situation where the AI will not see human faces because of the black mask. And the other thing is that the space suit does not let infrared radiation go through it. That means the system can “think” that the astronaut, who uses a space suit, is a robot. If an astronaut makes some mistake that causes a situation where the system translates the space-suited human as a robot. Then the system tries to remove those malfunctioning robots. 

When the astronaut cannot catch the tool the system will try to remove the astronaut that it thinks of as a robot. If the robot cannot catch the tool and the computer removes it seems like an overreaction. The reason for that overreaction is that there are no descriptions of the cases where the robot makes such big mistakes that it must be removed. If those things are not described every mistake that robot makes causes the removement. If a robot drops one screw to the floor and the programmer describes that “mistakes cause removement” that means the system translates even the smallest mistakes to cases, and their robot must be removed. If those cases are not described every mistake causes the removement. The system will not automatically make a difference between small and big mistakes. If the only thing is a mistake, the system removes the robot even if it drops the cup from the table.