How to teach AI?

 How to teach AI? 

If we want to make AI that operates smoothly in everyday life, we must make one thing clear to ourselves. The thing is that the AI must handle many variables so that it can operate independently. And it's hard to program those things into the AI’s source code. Even the best programmers forget something. So the answer can be the robot that sits in the classroom and learns like a human. This helps to create all the necessary variables for the AI. The robot acts as a medium between humans and the AI. 

If we want to make self-developing AI, we must use two AIs. The first AI makes a requirement of what it needs for the other AI. Then another AI creates source code. Then, that system must connect the AI to the source code of the first AI. And after that, the system reboots the servers. Before that, the system must simulate how it controls the virtual robots. 

The AI’s self-development happens as in natural organisms. The AI must detect what it needs, and then the biggest problem is how the AI can create code that involves the data it needs. Source code is the DNA of the AI, and there must be something that tells the AI what a certain code sequence makes. The answer to that problem would be the digital twin, the other AI that it can use to simulate what certain source code makes. The best way to get the code is simply to ask for it from some other AI. However, that means the other AI must have access to that source code. 

That requires that the AI tell its problem to other computers. And those other computers give it the code that it needs. But there are legal and other kinds of problems for that kind of self-developing model. So, how do we improve our skills? If we don’t know something, we search for data on the internet or visit the library and read about it in some books. Or, maybe we can take some courses on the topics. 

The thing in AI self-development is how the AI describes the problem to itself? When we talk about teaching AI, we can imagine a situation where the robot is sitting in the classroom and listening to how the teacher explains something. This is one version of the AI training. The AI can also search how some coders make solutions and then copy that code into its digital twin. If we want to teach AI to drive cars, we can use the same method as we use when we learn to drive cars. The driver must drive a car that records the actions for the AI. Then the AI asks necessary questions, and the human accepts that process if it's safe. 

We must remember that there are many variables that the AI should notice in the traffic. It would be very hard to describe all of those variables and traffic signs for the AI by writing them into its source code. The answer would be that the programmers make the base model for the AI. And then AI fills that model. And one of the most interesting ways to learn how to drive is to put the human-shaped robot into a driver’s school. The robot must know that it must avoid humans, know the traffic light, and other regulations. Then the robot that the AI uses for its developing platform or transmitter that transmits data to the AI will finish that training.