Learning from the past, building the future: AI in geotechnical engineering through the FOMLIG 2025 NextGen initiatives

Artificial intelligence is reshaping engineering disciplines, and geotechnical engineering is no exception. The next generation of geotechnical engineers must not only learn to use AI tools, but understand where they come from, what assumptions they rest on, and where they may fail. This paper reports on three intertwined NextGen initiatives hosted at the 3 rd workshop on the Future of Machine Learning in Geotechnics (FOMLIG3 2025) in Florence, Italy. First, a Museum of AI showcased twenty landmark papers from Karl Pearson’s principal component analysis (1901) to GPT-3 (2020), tracing the historical foundations of modern AI alongside the intellectual heritage of Renaissance Florence. Second, a NextGen Student Competition challenged participants worldwide to predict shear wave velocity ( V s ) from piezocone penetration test (CPTu) data for offshore wind turbine foundation design, attracting 22 competing teams and six extended abstract presentations. Third, structured community exchanges captured practitioner and researcher perspectives on the challenges, opportunities, and future directions of AI in geotechnical engineering through open discussions. The perspectives gathered span data quality and physical modeling, education and skill development, ethics and governance, and industry practice and trust. Together, these initiatives offer a multidimensional portrait of AI adoption in geotechnical engineering and document the lessons learnt for the benefit of the broader community.