Artificial intelligence (AI) is better than meteorologists at describing complex natural processes, but it cannot create equations and test them against generally accepted laws of physics. This was stated by Roman Vilfand, scientific director of the Hydrometeorological Center of the Russian Federation. RIA News.
According to the expert, there are currently several areas where neural networks or machine learning have an advantage over humans. He particularly drew attention to ambiguous situations.
Vilfand explained that we are talking about processes where laws do not exist or have not yet been formulated. He also cited subgrid-scale processes as an example.
“This is where the neural network is very good,” the expert emphasized.
However, he explained that artificial intelligence does not yet understand the fundamental laws of physics. In this regard, artificial intelligence cannot conduct research the way real meteorologists and physicists do.
As the expert explains, the equations developed by scientists correspond to the fundamental laws of conservation of energy, conservation of momentum and others. However, such concepts do not yet exist in neural networks.
“A neural network cannot write equations, second-order partial differential equations,” Vilfand explained, pointing to the need to achieve “synergy and interaction” in the work of artificial intelligence and weather forecasters.
Formerly American company Amazon started Artificial intelligence identifying products in poems and ballads.