Description: This PhD project develops new ways for artificial intelligence to learn from the many different types of data that geoscientists work with. Rather than building a separate model for each task or dataset, the project trains a single system that learns a shared “language” connecting these diverse data sources. This shared representation allows the model to reason across data types in the way an experienced geoscientist might, drawing on multiple lines of evidence to interpret what lies beneath the surface.
Benefits: Modern geoscience generates enormous and varied datasets, but each is typically analysed in isolation. By learning a common representation across data types, this research makes it possible to combine evidence more naturally and apply the same underlying model to many downstream questions, from mineral exploration to environmental monitoring. The approach reduces the need for large labelled datasets in any single task, helps surface patterns that span multiple data sources, and supports more transparent, evidence-based reasoning. This will accelerate discovery of critical resources and strengthen Australia’s capacity to extract value from its growing geoscience data assets.
