Future of Machine Learning in Geotechnics (FOMLIG), 5–6 Dec 2023, Okayama, Japan

This report presents the key talking points in the First Workshop on the Future of Machine Learning in Geotechnics (FOMLIG), that include data infrastructure, geotechnical context, computational cost, and human judgment. On the first point, it was argued that further growth in data sharing needs stronger demonstration of value to practice and protection of data privacy. On the second point, significant progress has been made in addressing site specificity (site recognition challenge). On the third point, it is costly to interpret monitoring data in the context of machine learning guided observational method (MLOM) because the 3D domain influencing the geotechnical structure is large, real-time dataset is very large and its attributes are complicated, data fusion remains challenging, and computation speed must support real-time decision making. Real-time machine learning-based decision support is clearly not useful if it is not providing the engineer with sufficient lead time to adjust the construction process. On the fourth point, the capability of generative AIs such as ChatGPT to act as an intelligent companion to an engineer in decision making is exciting. The role of human judgment in human-machine teaming is unclear, but for human-machine teaming to be effective, a deliberate approach is needed to build trust between the human and the AI/robot partner.