The neural network was able to solve the famous three-body problem, which further confused scientists

An international team of scientists from the Universities of Edinburgh, Cambridge, Santiago and Leiden reported controversial success in solving the famous three-body problem. A classical problem in celestial mechanics, which supposedly has no solutions, was solved by a neural network many times in a row, albeit in a simplified form. But a study of her decisions led to the conclusion that great minds deceived themselves.

Brief information: the problem of three bodies is one of the problems of celestial mechanics, which consists in determining the relative motion of three bodies interacting according to Newton's law of gravitation (for example, the Sun, Earth and the Moon). The problem is extremely difficult and, as it is considered, in general, it is unsolvable.

Before the advent of supercomputers, none of the mathematicians seriously tackled the three-body problem, with the exception of a few special cases. All known solutions to date are based on serious constraints that simplify the initial conditions. Scientists decided to move away from them and developed a neural network to find solutions to the problem in its purest form. To speed up the process, she was assigned a supercomputer as an assistant, which performed a lot of routine calculations, solving equations compiled by a neural network.

In this case, the neural network behaved like a creative person - it went through and checked the options for solutions at an intuitive level, and not through a step-by-step analysis. More precisely, it was thought so, but when the creators of the system saw how easily it solves the problem, they began to doubt. After a long analysis, they came to the conclusion that the "creative" solutions of a neural network differ little from the results that a supercomputer can produce, operating by a simple enumeration of options.

This looks like a new paradox. The neural network had freedom of choice, but in the course of solving the problem, it independently came to the same conclusions as mathematicians of past eras, began to think like them. Does this mean that the human mind, in principle, cannot solve the problem of three bodies? Or does its solution just boil down to the obligatory simplification of the initial conditions to the norm in which a person is accustomed to exist and think?