“These things are hard to tip over,” geologist Wilson Bonner assures me as the four-wheeled all-terrain vehicle he pilots suddenly tilts sideways, throwing me toward the churning mud beneath our wheels. We climb the side of a densely forested hill in rural Ontario, Canada, on a cold fall day, heading toward a place that Bonner’s employer, the startup KoBold Metals, says represents the marriage of cutting-edge artificial intelligence with one of humanity’s technologies. oldest industries.
We indeed complete the half-hour hike relatively mud-free, eventually crossing a ring of broken trees and mangled brush to form a strip of bulldozed mud. A black pipe about as wide as my arm protrudes from the ground – the upper end of a hole almost a kilometer deep that was dug into the ground by a drilling rig the size of a truck that sits unused nearby. It’s not much to look at, but this hole could mark a milestone in the future of mining, an industry crucial to the global transition to renewable energy.
As the world begins to transition from fossil fuels to greener alternatives, the global race heats up to find the vast quantities of cobalt, lithium and other metals needed to build all the batteries of electric cars, solar panels and wind turbines that we are considering. need. But finding new mineral deposits has always been difficult and expensive, and it’s only getting worse. Most of the world’s easily discovered reserves are already exploited. Those that remain tend to be in remote locations and deep underground. Miners generally claim that only 1 in 100 exploratory drilling reveals anything.
KoBold Metals, a four-year-old startup, is one of a handful of companies trying to make the process faster, cheaper and more efficient by applying artificial intelligence. KoBold has built a massive database incorporating all the information it can find about the Earth’s crust – the equivalent of 30 million pages of geological reports, soil samples, satellite images, academic research papers and century-old handwritten field reports. A team of data scientists converts all this disparate information into something machine-readable – by scanning written reports with optical character reading software, for example, or by standardizing geophysical information recorded in different digital formats.
All of this is done through machine learning algorithms that identify geological patterns and other characteristics of places where metals have been found in the past. The algorithms can then be deployed against the full database to find promising locations with similar patterns that have not been explored, generating a series of maps showing where the target metals are likely to be found.
Backed by investors including venture capital firm Andreessen Horowitz and Bill Gates’ Breakthrough Energy Ventures, KoBold’s first exploration teams took hold last summer, prospecting in areas of Zambia, Greenland and Canada , including the Ontario site near Crystal Lake.
