The company’s initial focus on Canada is partly because the country has a large amount of survey data in the public domain, including narrative field reports, outdated geological maps, geochemical data from borehole samples, airborne magnetic and electromagnetic survey data, lasers Radar readings and satellite images that span decades of exploration.
“We have a system where we can ingest all this data and store it in a standard format, quality control all the data, make it searchable, and be able to access it programmatically,” Goldman said.
Once all available information has been compiled for a site, KoBold’s team will use machine learning to explore the data. For example, the company might build a model to predict which deposits have the highest cobalt concentration, or create a new regional geological map that shows all the different rock types and fault structures. Goldman said it can add new data to these models as they are collected, allowing KoBold to adaptively change its exploration strategy in “near real time.”
KoBold has used the insights of machine learning models to obtain its Canadian mining rights and develop its site plan.Its collaboration with Stanford University Earth Resources Forecast CenterSince February, an additional layer of analysis has been added to the mix in the form of an artificial intelligence “decision agent”, allowing the development of the entire exploration plan.
Jef Caers, a geoscientist at Stanford University who oversaw the cooperation, explained that the digital decision maker quantified the uncertainty in the results of the KoBold model and then designed a data collection plan to reduce this uncertainty in turn. Just like a chess player trying to win the game with as few moves as possible, the goal of AI is to help KoBold make decisions about potential customers with the least amount of energy-whether the decision is to drill a hole in a specific location or walk away.