AI – Driving Future Exploration

Quminex Abstract: New Data-Driven Approaches to Aid Exploration Decision-Making 
Speaker: Alan Baxter

Early-stage mineral exploration represents the highest-risk segment of the mineral supply chain, often consuming 12 of the 16 years typically required to derisk a prospect. While the industry frequently attempts to apply “big data” algorithms to accelerate decision-making, mineral exploration actually faces a small, fragmented data reality. Regional datasets vary wildly in resolution, multi-modal coverage is inconsistent across prospects, and well-documented deposit training labels are globally scarce. 

Two main modeling paradigms dominate current industry discussions: deep learning and classical machine learning. Deep learning models rely on data-intensive, raster-based frameworks that require vast, consistent datasets to generalize effectively. When forced onto sparse exploration data, these models risk memorizing specific training targets rather than learning transferable geological rules. Furthermore, because deep learning optimizes via abstract, self-derived features, its outputs remain an uninterpretable “black box” to geoscientists. 

Conversely, classical machine learning operates effectively on sparse, non-raster datasets but relies on explicit, human-derived inputs. While historically viewed as less sophisticated, feature engineering inspired by deep learning can bridge this gap. By pairing human-designed, explainable features with robust classical models, geologists and exploration managers gain an objective framework to evaluate model outputs, understand trade-offs, and derisk prospects with confidence. 

VRIFY Abstract: Targeting 4.0: Data-Driven Prospectively Mapping in Manitoba’s Superior Province
Speaker: Ed Nelles

Mineral exploration increasingly demands that geoscientists not just integrate complex, multi-layered datasets to identify targets, but also communicate results in a simple and consistent manner.  

VRIFY has developed a suite of tools to support this reality, including a geoscience-led prospectivity mapping software. VRIFY Predict uses pretrained vision transformers to assess exploration potential and generate targets for multiple commodities from a wide variety of datasets and known mineral occurrences. 

In this session Vrify will use a portion of Manitoba’s Superior Province as a case study, assessing prospectivity for multiple commodities using publicly available data, including lake sediment geochemistry, ground gravity, and airborne magnetics. A data-driven workflow within the software converts these heterogeneous inputs into high-resolution, probabilistic prospectivity maps within a unified processing framework. 

The process is grounded in data but is also flexible and iterative, enabling geoscientists to apply their geological expertise through layer selection and data augmentation. This approach lets that expertise shape the outcome in a rapid and repeatable manner with the support of advanced AI. 

The results derived from the workflow are explainable and will be showcased in context, in a consistent, visual format that supports rapid review, iteration, and communication, helping teams move more efficiently from data to decision.