The last two decades have brought a tidal wave of real time instrumentation and monitoring to underground construction projects, with thousands of sensors providing data every few seconds. Real-time access to geotechnical/structural monitoring and TBM data, with remote graphical access to data, is now standard practice worldwide. In theory, this provides wonderful medicine for the epidemic of geotechnical risk brought about by massive levels of uncertainty in ground conditions and ground behavior. This presentation addresses ways in which information engineering, including AI machine learning, has been and can be further used to advance tunnel design and construction across numerous projects in the US. The presentation will particularly focus on learning, prediction, forecasting and risk reduction using TBM data. AI-driven advances discussed in the presentation will include characterizing the ground conditions using TBM data, updating the geological model using TBM data, TBM condition monitoring and anomaly detection using pattern recognition techniques, and optimization of TBM operations to reduce cost. Results from implementation of procedures from multiple projects are presented for each aspect. The prospects for TBM automation are also discussed within the context of these advances.