Are We Ready to Use AI Technologies for the Prediction of Soil Properties?

Artificial intelligence (AI) has become a hot topic for different professions in which geotechnical engineering is no exception. It is anticipated that AI could perform tasks, solve complex problems and make decision by mimicking intelligence or behavioral pattern of humans or any other living entities. Attempts have been made to study and adopt AI technologies in geotechnical engineering. In this paper, a dataset of marine soil in South Korea is re-analyzed using different commonly adopted AI algorithms. The soil’s compressibility is considered as the dependent variable (i.e., to be predicted) while other soil index and physical properties are regarded as the independent variables. The data are split into the training and validation set. While an algorithm learns from the training set, its prediction performance is examined using the validation set. Then, the Bayesian model class approach has been used to explain the potential problem of the use of AI algorithm to predict soil properties. At the end, by using this study as an example, the author discusses from a partitioner’s perspective how AI could affect our professions. In particularly, the question “are we ready for using AI to predict soil properties” is addressed.