Little bit of theory :)

Appearance of the AI's code

(I tried to describe this page as concisely and simply as possible, so that you understand what I want to represent you in general)

The very first drawing that shows how AI can see certain algorithms. They are interconnected by chains, and depending on what we want to enter into the code, AI will activate neural connections. Suppose we have a certain task that AI can handle, if we get the same result when we enter information, it means that the AI algorithm is correct. Below you can see how AI thinks in terms of security. If condition A is triggered, then it blocks him and does not give access, if condition B is also true, then it will also block him and will not broadcast any information.

Task A - secured

In that picture above, you can see, the AI algorithm perceives task A and thus the algorithm allows it to be well-protected or strengthened without failures. All information that will pass through task A will have a clear path from start to finish

Both are secured

The algorithms are the same here, just task B is added here, which also has its own route. After the end of the algorithm, it will also be protected.

You may think that it is quite simple how the AI perceives this code, but this was only the base of the algorithms

Something more interesting lies in the perception of AI...

In this example, we only have the number 9, which was written by hand, and we decided to translate it into AI algorithms so that he understands what is written. Our field has a range of 28 x 28 so that IA can extensively determine what is there.

Here, as you can see, we reveal in width and height absolutely all options (cells) that are contained in the AI source code. Our "nine" is not found in all that cells, which can be seen in the image. With each new step, AI understands more and more what is written and what it needs to understand. Absolutely every dot, slash, etc. plays a huge role, which can soon lead to a distorted answer, thereby giving the user an incorrect result.

And now for a second, let's imagine that you can use not only numbers, but also, for example, murals. That is, AI will search for all source cells not for numbers, but for line shapes that the user enters. The representation of a number or letter as filled cells in a 28 by 28 field is the algorithm for understanding AI. It gets even more interesting, doesn't it?)