Machine learning, especially the development of deep neural network learning algorithms, is changing the way people discover knowledge. As the oil and gas industry is shifting to unconventional oil and gas exploration and development, the evaluation and interpretation model based on limited petrophysical parameters is difficult to meet the complex lithology and structure of unconventional reservoirs, which poses a great challenge to the traditional logging evaluation technology. Oil & gas artificial Intelligence (Oil & Gas AI), based on oil and gas big data, machine learning algorithms, oil and gas application scenarios, has greatly promoted the application and development of AI technology in various professionals of oil and gas industry. According to the data–driven petrophyical knowledge discovery, and the research idea of the “data–algorithm–platform–knowledge–application scenario”, firstly we analyzed the inherent attributes, principles, quality control, hardware requirements, learning model selection, testing, and performance evaluation implementation process for the machine learning algorithm. The tree graph of the applicability of the machine learning algorithm in logging is summarized, especially the relationship between the application potential and machine learning algorithm in oil and gas logging. These applications include simulation methods for data correction, petrophysical analysis for data calibration, logging data quality control, integrated evaluation, and reservoir monitoring. The study case shows that machine learning algorithms in lithology identification and reservoir evaluation, classification, mechanics, and reservoir evaluation based on the data link across multiple physical properties of petrophysics compared with traditional well logging method, which break through the limitation of experimental conditions and physical properties and has interdisciplinary and comprehensive characterization, had obvious advantages and potentials in well logging technology.