Our modern lives, from the smartphones in our pockets to the electric cars we drive, rely heavily on metals buried deep within the Earth. Yet, the industry tasked with finding these essential materials is struggling. Discovering new deposits has become significantly harder, precisely when our demand is at its peak. This is where innovative technology, particularly AI, is stepping in to change the game, promising to locate minerals much faster and more efficiently.
Key Takeaways
The demand for critical minerals like lithium, copper, and cobalt is soaring due to the shift towards electric vehicles and advanced technology.
Traditional mining exploration is underfunded and technologically stagnant, leading to a decline in discovery rates.
New AI-powered methods can predict mineral locations thousands of times faster than conventional techniques.
This technological advancement aims to make mining safer, more environmentally friendly, and economically resilient.
The Growing Need for Critical Minerals
Everything around us, from the buildings we inhabit to the devices we use daily, comes from either growing or mining. As we move towards a more sustainable future, the need for electrification is paramount. This means more electric cars, robots, drones, and aircraft, all powered by batteries. Our schools will require more computers, and data centres will need advanced chips to support AI. All these advancements depend on a steady supply of raw materials like lithium, copper, cobalt, and nickel. To meet these demands and build a truly circular economy, projections suggest we'll need to build over 400 new mines by 2040.
The Mining Industry's Exploration Challenge
Before any mine can be built, the raw materials must be found. However, the mining industry has been lagging behind other discovery-driven sectors like pharmaceuticals and technology. While those industries invest heavily in research and development, mining spends very little on exploration – less than a penny for every dollar returned to shareholders. This underinvestment means the technology used for finding new deposits has barely changed. In fact, over the last 30 years, we've become ten times worse at discovering new ore bodies. The easy deposits, those near the surface and easily visible, have largely been found. Now, we need to look deeper, and that's where the real challenge lies.
A New Approach: AI-Powered Prediction
Contrary to popular belief, we don't lack the mineral deposits themselves; we lack the information about where they are located. Imagine having a crystal ball to see exactly where the best rocks are. Since we don't have one, the next best thing is to make accurate predictions. This is precisely what companies like KoBold are working on. They are developing machine learning technologies to predict various properties of the Earth's subsurface, such as the concentration of copper, rock density, and fracture patterns, all while rigorously quantifying the uncertainties in these predictions.
Transforming Exploration with Data
Traditionally, exploration involves flying aircraft over vast areas to collect data on the Earth's magnetism and gravitational field, which can hint at what lies beneath. However, interpreting this data is complex. There can be an infinite number of possible geological scenarios that fit the same measurements. The old way of doing things was to pick one possibility and ignore the rest, often leading to suboptimal mine designs and decisions. The new approach involves collecting all possibilities consistent with the measured data. This is achieved by simulating the physical response of different rock arrangements. AI can learn the relevant physics of the rocks and perform these simulations 10,000 times faster than conventional methods. This allows for better data collection and more accurate predictions about where to explore next.
Quantifying Uncertainty for Better Decisions
When exploring, if a rock body is denser than its surroundings, you might drill through its centre. But with hundreds of thousands of possible solutions, the smartest move is to gather data where uncertainty is highest, systematically eliminating possibilities. This maximises the information gained for every dollar spent. This process is repeated to quantify uncertainties. Even after discovering an ore body, defining its exact size and shape remains difficult. Imagine drilling and finding five percent copper; predicting the concentration for someone sitting next to you, or across the room, or in the next city, is a huge challenge. We've only sampled a tiny fraction of the Earth's crust, often with significant gaps between sampling points.
Building the Mine of the Future
This new technology is already making a difference. In Zambia, for example, it's helping to design and develop a mine based on predictions, even with limited rock samples. The AI approach allows for a clear understanding of where to collect more data, where to drill next, and when it's appropriate to stop drilling and start building. The industry typically designs mines based on a single model. However, tools like KoBold mine are being developed to consider numerous possible mine designs against various ore body shapes. This leads to better decisions regarding the amount of ore to extract, waste produced, water usage, and overall cash flow. It also helps in planning permanent infrastructure like shafts and tunnels, maximising ore recovery while minimising waste.
A Safer, Greener, and More Resilient Future
Beyond profitability, better predictions lead to a safer mine by identifying weaker rock formations. It also contributes to a more environmentally sustainable mine by reducing impact and a resilient mine that can support local communities and businesses through fluctuating commodity prices. The Mingomba project in Zambia is set to be a mine of the future, developed by a global team, including local talent. As our lifestyles evolve and demand for these materials grows, the mining industry must transform. By embracing better technology and responsible practices, we can build better mines for a sustainable future.