Artificial intelligence and machine learning to enhance critical mineral deposit discovery

Abstract

The application of machine learning (ML) in mineral exploration has garnered significant attention and investment, yet greenfield mineral deposit discovery rates remain unchanged. This limited success stems from challenges such as low data quality outside existing mines, inconsistent sampling, limited interdisciplinary collaboration, and the unique complexity of geoscientific problems. Unlike traditional ML applications, mineral exploration demands a focus on subtle variations within finite search spaces, requiring an exploratory rather than accuracy-driven approach. Effective implementation necessitates collaboration between data scientists and geoscientists, leveraging ML as a tool to test hypotheses and analyse diverse datasets. However, reliance solely on ML overlooks the critical role of human creativity in generating and evaluating novel search strategies. Broader adoption of statistical methods, integrated spatial models, and innovative data preparation techniques can address the inconsistencies in exploration datasets. Furthermore, subjective modelling approaches, such as Delphi methods, can complement ML by incorporating expert judgment to overcome predictive limitations. By combining technological advancements with human expertise, the mineral exploration industry can enhance discovery success and achieve long-term sustainability. There is an important short-term requirement to secure the supply of critical metal resources, as their supply from existing mines and brownfield exploration is finite and commercial recycling of critical metals is still in its infancy.