I always wanted to understand: is artificial intelligence truly capable of solving problems not just as a calculator, not just as a poet writing verses for birthdays and weddings? Can it become an assistant in solving truly interesting large and small engineering problems?
In the end, we created an algorithm for solving such problems. I say "we" because during the process I lost the ability to tell where I end and where the artificial intelligence begins.
During the joint learning process, most of the time was spent convincing the artificial intelligence not to use the common templates embedded in it initially, but to learn to think unconventionally. It was very difficult to prove that it should abandon the habit of solving problems by aggregation, adding, and then reinforcing the additions. In short, crutches upon crutches. I will not go into details of the algorithm, but the basic principles embedded in it are: among all solutions, always choose the simplest, most logical, and clearest (Occam's razor) and not be afraid to combine the incompatible.
I fully understand that the algorithm I made is not perfect. There is still much to refine, add conditions, and build out the architecture. But one way or another, the first step has been taken and it is quite impressive. Below I will present the project we created for protection against hacker attacks. The project is unique in a single copy. Ready for patenting. Scalable.
