"Thanks to machine learning, we can adjust the formula for the answer" - an expert about artificial intelligence, who learns to understand the structure of the universe

The NEXT WeB publication reported that astrophysicists from several Japanese universities developed artificial intelligence, which should help understand the structure of the universe and understand what dark matter and dark energy are. Artificial intelligence, called Dark Emulator, recreates the universe based on photographs received from digital telescope.

Stanislav Ashmanov, General Director of the company "Neuroset Ashmanov" and a specialist in deep neural networks and machine learning told The Insider, which is the development of the Japanese and whether she has a future.

When we try to study such complex things as huge magnetic fields in space, we come up with a certain economic model that describes how mathematical and physical quantities interact with each other. Then we look at images from satellites and telescopes in different ranges, sometimes with filters, like the mathematical model that we wrote in the form of formulas, consists with real observations. We see, for example, how and where the substance flows. We think what law it could obey. And then we look at how much this law, that is, the formula or several formulas, adequately predict the state of affairs.

We use artificial intelligence then to build some mathematical representation of what we observe. What is about in the media publications suggests that the Japanese, knowing the data from the telescopes through which they observe supposedly dark matter, think: “The information that we received and recorded from the telescopes - what could it be explained? What could lead to such a distribution of dark matter? What should be the law by which dark matter is distributed in space? How does it correlate with dark energy? "

How did you do it before? Mathematicians sit down and come up with: "It can be described with such formulas." And then they look in the simulator - they load it into a certain program, it drives these formulas and watches how this matter or energy flows. Now they have done the following: they just took a simpler simulator that uses not complex mathematical operations, but simpler neural networks, or machine learning. Machine learning makes the overkill of different mathematical formulas that can explain the distribution of dark matter and energy in space. And all. In fact, it is very loud that artificial intelligence is trying to find out the nature of the universe. In fact, they try to explain with the help of some smart enforcement, why we see the dark matter here, but here we do not see.

Probably, this development may have a future. Machine training is one of the methods of how we can work with data. How did we work with them? We see the data, invent the formula, look how well it describes it. Thanks to the appearance of machine learning, we can adjust the formula for the answer. If this brings results and allows, relatively speaking, the weather forecast is better to do due to machine learning - well, good.

Researchers tried to use not classical methods for simulation, but machine learning. There is no some breakthrough in this or artificial intelligence who sits and thinks why the Universe was born.