Beyond Automation: responsible AI workflows in geotechnical modelling and monitoring

Overview

Geotechnical engineering is entering an era where machines can see more, compute faster and detect patterns long before humans notice them — yet we remain unsure how far to trust them.  This lecture explores that tension.  We examine the reliability limits of AI in a domain defined by sparse data and high consequence, the legal and insurance frameworks that keep liability firmly human, and the cultural forces that slow adoption even when the technology works.

Through examples from current practice, from automated pipe-defect recognition to 3DGS based deformation monitoring and machine learning assisted back analysis, we show where AI genuinely strengthens engineering insight, and where caution is still warranted.

Finally, we look toward the possibility of more autonomous workflows and what must be solved before such systems can operate responsibly beneath the surface.  

Turing asked whether machines could think; for engineers, the harder question is how to verify and validate what they produce.

Programme