Artificial intelligence in geophysics: Opportunities and risks

The article briefly reviews some artificial intelligence methods successfully used to process and interpret logging data and for seismology and geothermy. The possibilities of artificial neural networks, the Support Vector Machines, the Random Forest method, and genetic algorithms are highlighted. The basic information about the advantages and limitations of artificial intelligence tools is given. AI is not self-sufficient for geological and geophysical research. It is important to adapt its algorithms to work with large volumes of geophysical data. If the algorithm has too high computational complexity, calculations can be simplified by manually processing the input data or using conventional software. Sometimes, several algorithms are used to solve a single problem. In such cases, each network is trained several times. When comparing the results with approximately equal control errors, a computationally simpler neural network is chosen. For the purpose of better orientation in the computing world, information is provided on the computational adaptation of artificial intelligence to geophysical data. Attention is drawn to the possibility of financial risks associated with the use of an insufficiently powerful network when modeling a particular dependence.