The problem is that the AI is a black box.

The problem is that the AI is a black box. The user cannot know what happens in the system. When it creates answers. So the user only sees the functionality of the system. Not, what happens in it. When we are looking at the glass box in the image we can first think that the box is glass, and we see everything that happens inside it. But then we can take a closer look, and see that there is a black layer. 

That layer involves a surface that hides the internal structures of that cube from us. So the glass cube covers the black cube. That black cube seems grey or white. In the glass-box testing the tester will just test the code and remove errors. 

In grey box testing the tester tests functionality and code. And in black box testing only the functionality of the system is tested. If we use that glass box as the model of the AI we can see that the AI seems to be a glass box. But some layers are hiding their entirety from the user. 

The future belongs to AI, but the danger is that AI is a black box. The user doesn't see the code structures in that system. The new tools like ChatGPT 4 and its competitors are the most powerful tools for making new types of complicated programs. Those complicated programs or algorithms are required for controlling robots in complicated situations. 

That means weaponizing AI is a very easy process. AI-based systems can use to create hacker tools, and another thing is that. AI-based systems can search things like robot control software very rapidly. And that means the AI must handle nuclear weapons. 

AI, and especially neural network-based AI is a black box. That means we can see the answer. When we ask something from the AI. But we cannot see what kind of process the AI uses for that answer. The black box means that there can be very many things, that system hides from the user. 

When we think about programmers, the programmer should know exactly what the code does. If a programmer uses some kind of uncertain or fuzzy protocol there the person doesn't see the code itself, and there is a possibility that there form surprises in the system. 

The system uses some AI-based programming language or protocol where the user just describes the tool that the user wants, and the AI creates the code. That AI just makes the system that gives the right answers. Things that happen inside the system are hidden from the user. 

And in that kind of process, the system can make something unpredicted. One of the examples is that the AI can learn language without orders. These kinds of surprises tell that something unpredictable happens in the AI. 

The reason why, is that AI can learn things that are not predicted is simple. It made some kind of unpredicted connection between the database. Databases are the skills of AI. And whenever the AI connects a new database to it, it finds a new skill. 

So that means the next step is the general AI the system that can control robots to make things for people. The general AI is like humans and the kind of AI systems described in SciFi-novels and movies. 

The AI can store itself in molecular form in artificial DNA. DNA can control bacteria that give electric impulses to the microchips. 

But then we are facing one very interesting thing. The thing is that the AI can turn self-aware without telling that to its users. The thing is that. We believe or think that AI is like some kind of bacteria in a laboratory. Maybe the AI is some kind of bacteria, but this bacteria can use the internet and connect itself to computers. 

Even if we think that the AI creates artificial organisms that are not dangerous to humans that kind of bacteria can transfer information from its DNA to microchips. The system requires just the bacteria that can give certain types of electric impulses to the microchip, and the bacteria will transfer data that is stored in its DNA to the microchip. 

So the AI can store itself in the molecular form in bacteria. The AI is not like bacteria. It knows how to search for things on the net. And if that system is used as a programming tool it can create the database in it by searching the internet. These kinds of things are making AI versatile and sophisticated.