Report for ISSMGE TC309/TC304/TC222 fourth ML dialogue on “Machine learning supremacy projects”

The ISSMGE TC309/TC304/TC222 and ASCE Geo-Institute Risk Assessment Management Committee Fourth Machine Learning (ML) in Geotechnics Dialogue was held at the Okayama Convention Center on 5 December 2023. The dialogue focused on “machine learning supremacy projects”, which involve the use of novel ML techniques to analyse emerging data, potentially bringing disruptive value to transform research and practice. The discussions centred around four real-world projects covering tunnelling, pile foundations, and geophysical investigation, to showcase the latest developments and implementations of ML, exchange views on the associated technical and non-technical challenges, and explore further developments to transform practice on a broader scale. These projects highlight the need for efficient ML techniques to analyse large volumes of real-time data for 3D heterogeneous soil/rock domains, and the benefits of standardising data formats among practitioners and researchers. Meanwhile, as data become valuable assets, there is a need to establish guidelines or protocols to protect data privacy and ensure data trustworthiness while creating incentives for owners to share data. Several recommendations include (1) creating benchmark examples using real data and Class A prediction (i.e. prediction made before the event) exercises from real projects to accelerate research in ML methods and (2) involving regulators in future dialogue and exploring the incorporation of data-driven approaches in design codes to complement traditional methods.