The machine learning part II

The machine learning part II

The thing what is a problem with the artificial intelligence and the teaching process, where the machine learn things is to make teaching or programming process easier. And the thing in the teaching process is that the teacher or programmer must handle the huge masses of information, for making computers do something. When we are thinking about my favorite thing, the robot car, we are facing the thing, that the programmer needs to find out multiple details about the things, what the car will notice.

And then the programmer must give the instructions to the computer, what the car should do in every situation, what it might face on the streets. And this causes problems. The thing is that the car must notice people, who are stepping suddenly on the way and push the brake pedal. But there is one thing, what we must realize, and that is, what if somebody would try to rob the car. And this is the thing, where the car needs to defend itself and people inside it, but if the person would rise hand, and the computer thinks, that the finger is the gun, would the case turn very horrifying thing.

But when we will think about the interface of this kind programming tool, what is capable to create orders, how they act in the situation, we must think the brand new solution. This new type of programming tool might use multiple interfaces and allows many ways to handle the objects. And one version would be the tool, what looks like drawing program. Objects can copy to the memory of computer using similar cutting tool with picture handling tool.

And the thing is that the object will be surrounded and then on the screen is written the instructions, what to do in that situation. Of course, this kind of programming tools must have many different ways to input the data to the system. During the process, the programmer must first make the instructions, what to do, and then the car would send to the highway for recording the necessary reactions in certain situations. When the robot car faces the problem, what is not stored in memory, it would ask help to solve it. And then the driver can take control and record the reaction. This kind of things might be useful when the car is driving a certain route, but there is a problem with this kind of things.

The programmer must realize that the same object has different forms in the real world. And the thing, what makes more problematic for programming the reaction for the car, how it should do in the traffic lights, than making programs for killer drones. The killer drone needs only instructions, what is the enemy, and if the sensor would tell if the signal tells that the target is enemy, would the killer drone shoot the target by using missiles.

The transmitters of the own troops tell to drone, that they are own side, but the drone can open the fire if the signal is not recognized as the own side troops. But the car what drives independently needs many other ways to do than shooting the missiles. And that's why the programming the self-moving car is more complicated than making the drones. The thing is that the killer drones would just need advice, how to shoot targets, but the automobile cannot handle situations in the same way. That car must follow the instructions and orders to operate by following the law and avoiding risks.