Researchers at the University of Emouri in Atlaa, Georgia trained a neural network to discover new laws in physics, and surprisingly, the system was able to do so.
According to Phys, the research team succeeded in making this unique achieveme using empirical data related to a mysterious state of material called dusty plasma. Scieists then observed how artificial ielligence provided amazing and accurate descriptions of strange forces that had never been fully understood before.
Artificial Ielligence was able to correct some of the incorrect assumptions of the plasma theory
The developme of such a system shows that artificial ielligence can be used to discover unknown laws that corol the ieraction of particles in a system. In addition, artificial ielligence modifies the old assumptions in plasma physics and paves the way for the study of complex and multidisciplinary systems in completely new ways.
“We have shown that artificial ielligence can be used to discover new tips in physics,” said Justin Burton, a writer at the University of Amarai. Our method of artificial ielligence is not a black box. We know how and why this system works. The framework it offers is also applicable and can poteially be used in other multidisciplinary systems and open new paths to discover. “
The researchers began their work by combining real tests and a carefully designed artificial ielligence model. They keyed their study with dusty plasma. This state is found ranging from the cosmos, from Saturn and the moon’s surface to smoke from forest fires on the ground.
Despite its widespread presence in the universe, the precise forces operating between the particles in the dusty plasma have not yet been understood properly. The reason is that this system is unnecessary; That is to say, the force that puts one particle on the other is not necessarily responded to equal force from the other particle.
To deal with this challenge, the researchers built an advanced 3D imaging system to view the moveme of micoplastic particles io a plasma enclosure. The results of this study were then used to teach a custom neural network. Unlike most artificial ielligence models that require huge data, the network made by the University of Amati was taught with a set of small but rich data and was designed to consider iernal physical laws such as gravity, air resistance and iercoineal forces.
