We likely have all seen recent articles about how Artificial Intelligence (AI) is going to change the mining industry. I have been wondering if AI is a real solution or just a great buzzword. My original skepticism has diminished somewhat and let me explain why.
At a booth at the 2019 PDAC I had a chance to speak with a publicly traded company called Albert Mining (referencing Albert Einstein’s intelligence). They are providing exploration consulting services by applying a form of AI and have been doing so for many years. The company has been around since 2005 but were not using the term AI to describe their methods.
These days the term “AI” has become very trendy. Currently IBM Canada and Goldcorp are using Watson and AI to further their exploration efforts on the Red Lake property. GoldSpot Discoveries is another recent player in the mining AI field. It appears Goldspot offers something similar to Albert Mining but they extend their platform to include picking projects, picking teams, and picking investments. That’s a lot of analysis to undertake. Albert Mining is focused solely on mineral exploration.
Here is what I learned
Albert Mining’s system, called CARDS (Computer Aided Resource Detection System) uses pattern recognition and multi-variate analysis to examine a mineral property to look for targets. The system requires that the property has some known mineralization hits and assay samples. These are used to “teach” the software. Both positive hits and negative hits are valuable in this teaching step.
The exploration property is sub-divided into cells and data are assigned to each cell. These data attributes could be derived from geophysics, geochemistry, topography, soil samples, indicator minerals, assayed samples, geological maps, etc. I was told that a cell could contain over 700 different data attributes.
The algorithm then examines the cell data to teach itself which attributes correlate to known mineralization and which attributes correlate with barren areas. It essentially determines a geological “signature” for each mineralization type. There could be millions of data points and combinations of attributes. Correlation patterns may be invisible to the naked eye, but not to the computer algorithm.
Once the geological signatures are determined, the remainder of the property is examined to look for similar signature hits. Geological biases are eliminated since it is all data driven. The newly defined exploration targets are given a ranking score based on the extent of correlation.
Some things to note are that the system works best for shallow deposits, unless one has some deep penetrating geophysical surveys. The system works best if there is fairly uniform data coverage across the entire property. The property should also have generally similar geological conditions and as mentioned before, the property needs to have some mineralized assay information.
This exploration approach reminds me somewhat of the book Moneyball. This book is about the Oakland A’s baseball team where unconventional statistics were used to rank players in order to find hidden gems.
Are geologists becoming obsolete?
I was told that many in the geological community tend to discount the AI approach. Either they don’t think it will work or they are fearing for their jobs. Personally I don’t understand these fears nor can I really see how geologists can ever be eliminated. Someone still has to collect and prepare the data as well as ultimately make the final decision on the proposed targets. I don’t see the downside in using AI as another tool in the geologist’s toolbox.
Albert Mining’s stock price has recently gained some traction (note: I am not promoting them) because junior mining news releases are starting to mention their name more often (Spruce Ridge Resources and Falco Resources are some examples).
Probably years ago if a mining company said their drill targets were generated by an algorithm, they might have gotten strange looks. Today if a mining company says their drill targets were generated by AI, it gives them a cutting edge persona. Times have changed.