Early one morning, as light danced through the forest canopy above, Fedrick Eshiloni reached into the ground and scooped up fists of ochre-colored earth.
The scenery hardly resembled a hub of innovation: In this wooded stretch of Zambia’s northwest, home to reedy swamps and termite mounds the size of houses, locals still move goods by oxcart. But the 22-year-old, dressed in a blue workman’s uniform and accompanied by a team of prospectors, was performing a critical first step in an emerging high-tech quest to find the metals key to powering a clean energy future.
After a day collecting bags of soil, Eshiloni and his colleagues would haul their samples to a makeshift camp, where they are dried, sieved, and tested for traces of 34 chemical elements. Even tiny amounts offer clues whether ores containing copper and cobalt, both critical to the production of electric vehicles, lie below.
(The electric travel revolution has arrived. Read about it here.)
These first steps aren’t so different from the way mining companies have explored since the mid 20th century. But what happens next amounts to a high-stakes test of new data-driven techniques that some believe could radically transform mining—and help limit global warming in the process.
Unlike conventional prospectors, this team from KoBold Metals, a Bill Gates-backed California start-up, is using the data that they assemble—from soil samples to airborne surveys to reams of historical documents—as building blocks for a suite of complex geological models that are powered by artificial intelligence. KoBold and its backers are betting that AI will more effectively predict where ores have formed, and ultimately unlock new, deeper deposits.
KoBold isn’t the only mining company that’s embracing big data to aid the next generation of discoveries. But its high-profile financiers and focus on the metals needed for the green energy revolution are drawing attention to an emerging raw materials bottleneck that risks thwarting global efforts, including deals negotiated at the United Nations Climate Change Conference in Scotland, to build a less carbon-intensive world.
According to the International Energy Agency, keeping global warming “well below” 2 degrees Celsius (3.6 degrees Fahrenheit), the central goal of the 2015 Paris Climate Agreement, will require unprecedented growth in the production of commodities like copper, cobalt, nickel, and lithium. All are essential building blocks for solar panels, wind turbines, power lines and, above all, battery-powered electric vehicles, which are less carbon intensive than their gas-fueled counterparts, especially where electricity is generated from renewables.
By 2040, the IEA projects that meeting Paris targets will require annual global sales of more than 70 million electric cars and trucks, which together will demand as much as 30 times the quantity of metals that’s used in their production now.
Shifting to a green future is not without contradictory complications—at least for the immediate future. Although new technologies and tighter regulations have made mining less environmentally destructive, the extraction and processing of metals still contaminates water and soil, encroaches on habitats, and emits pollutants and the same greenhouse gases that have caused a warming climate to begin with.
Emissions linked to the minerals used in green energy technologies, however, are a small fraction of those generated by the fossil fuel-powered systems they’re designed to replace. Over time, as the adoption of electric vehicles accelerates, more battery recycling could make the search for new battery metals less imperative. Some of the green transportation burden could be shouldered by other solutions still in development—such as cars fueled by hydrogen—or technologies not yet imagined.
For now, though, analysts stress there’s no substitute for digging rocks out of the Earth.
Keeping warming to below 2 degrees Celsius using existing technologies will require “massive additional volumes of metals,” says Julian Kettle, senior vice president of mining and metals at Wood Mackenzie, a global energy consultancy. “There’s simply no way around that.”
Rocks under the hood
Founded in 2018, KoBold derives its name from cobalt, a lustrous bluish-silver metal that helps drive performance of the lithium-ion batteries that revolutionized consumer electronics when they were introduced in the early 1990s. The same batteries are used on a much larger scale to power electric vehicles, and cobalt gives them greater range, longer lifespans, and better protection against fires by reducing corrosion.
Its supply, though, is especially precarious: Nearly 70 percent is sourced from the Democratic Republic of the Congo, where a history of labor abuses and corruption have heightened the urgency to find deposits elsewhere. Automakers are also seeking cobalt alternatives—the metal, after all, is an expensive commodity—though the performance limitations of today’s cobalt-free batteries makes cobalt demand likely to accelerate.
Other metals KoBold seeks to unearth could soon feel a supply crunch as well. Gerbrand Ceder, a materials scientist who researches batteries at the University of California, Berkeley, believes nickel faces the greatest risk of long-term shortages, in part because it’s the most viable cobalt substitute.
Analysts also foresee a scarcity of copper, which is used in a range of green technologies, including electric vehicle motors, wiring, and charging infrastructure. A typical battery powered automobile uses three times as much copper as its gas-guzzling cousins.
These supply constraints have emerged, in part, because discovering viable metal deposits has become harder. That’s largely because the most accessible ones have been extracted: In Zambia, Africa’s second largest producer of copper, the ores mined today were either “sticking out of the ground” or just below the surface when they were found, according to David Broughton, a geologist with 25 years’ experience in the region who advises KoBold and others.
This doesn’t mean there aren’t deposits lurking deeper in the earth: The interaction between rocks and fluids that formed them more than 400 million years ago occurred well beneath the surface. But unlike the oil and gas industry, which has gotten markedly better at accessing hard-to-reach places, mining exploration has not made a major technological leap in decades. As a result, odds of success are dismal. By most industry estimates, fewer than one percent of projects in areas without extensive prior exploration result in commercially viable deposits.
The lure of big data
KoBold’s goal is to “reduce the uncertainty of what’s under the surface,” says Josh Goldman, the company’s chief technology officer. A more rigorous application of data and a boost from the evolving field of artificial intelligence are key to improving those odds.
Techniques of AI, including automation and machine learning, have already aided the climate fight by enabling better tracking of emissions, more sophisticated climate modeling, and the development of smart grids and other energy-saving devices. AI applications in mining have primarily focused on improving extraction from existing operations, though there’s gathering momentum in using them to aid the search for new deposits.
Today, companies ranging from tech behemoths like IBM to more specialized outfits like Canada’s Minerva and GoldSpot offer AI tools or services geared toward exploration. KoBold, though, is one of the few that also invests its own capital in projects—including its efforts in Zambia and others in Canada, Greenland, and Western Australia.
The company’s technological wizardry consists of two complementary systems. Connie Chan, a partner at the venture capital firm Andreessen Horowitz, which invested in KoBold in 2019 along with Gates’ Breakthrough Energy Ventures, likens the first to a “Google Maps for the Earth’s crust and below.”
Building it is a geological treasure hunt. Not only does KoBold collect its own data—drawn from rock and soil sampling, and measurements like gravity and magnetism taken from a helicopter. It also scours the historical record, using tools of machine learning—where computers draw insights from data too complex for humans—to extract key information from old maps and geological reports, which can run into the millions of pages. On some projects, KoBold forms joint ventures with established mining companies, which provide data of their own. BHP, the world’s most valuable mining firm, is a partner in Australia.
KoBold uses all this information to build and train a suite of analytical tools it calls “machine prospector.” While they won’t unearth metals directly, they can give geologists a better idea of where to look—or where not to. One tool key to KoBold’s work in Zambia helps identify mafic rock—which can fool explorers into thinking they’ve found copper—and therefore prevent costly failed drilling.
Another tool, in use in northern Quebec, where KoBold hopes to find nickel, copper, and cobalt, guides its research team to the most promising rock outcroppings for sampling, hastening the search. “You can actually get through an area of a couple hundred square kilometers in a season,” says David Freedman, a KoBold geologist who spent last summer traversing the tundra.
How well will machine learning work?
Machine learning tools developed by KoBold and others have already made geologists’ lives easier; as Freedman notes, there’s no wind, rain, or mosquitos when planning out a prospecting route from behind a computer. Nonetheless, these techniques are still in their early stages and their ability to drive major finds remains an open question.
Antoine Caté, a geologist and data scientist at SRK, an international consultancy, believes machine learning models have the potential to “dramatically improve” success rates in exploration—in part due to their ability to detect patterns among datasets with more variables than the human brain can process. Still, he cautions such tools are only as good as the information that’s fed into them: If an algorithm is built with substandard data, it will be ineffective at best and at worst lead prospectors down false paths.
AI also doesn’t eliminate the need for human ingenuity. “These tools are amazing for diagnostics,” Caté says. “But at the end of the day you still need a skilled person to take the information and make something out of it.”
KoBold’s Goldman agrees. The need for robust data, he says, is why KoBold’s sleuthing is so meticulous. Still, he admits the company’s tech could take time to fulfill its promise, and how much it might speed up discoveries of deposits is uncertain.
Chan, whose firm has helped fund tech giants like Airbnb and Instagram, believes the wait will be worth it. The mining industry’s exploration struggles and the urgency to find more battery metals, she says, means a software-driven approach is long overdue. “If anyone can show they’re more effective at choosing the right places to look—that’s incredibly valuable.”
Even if machine learning techniques prove successful, though, it might not be enough to prevent future shortages. Better exploration is only part of the picture: To meet the Paris Agreement’s two-degree goal, Wood Mackenzie estimates that the mining industry will need to invest more than $2 trillion in mine development over the next 15 years—a colossal increase from the roughly $500 billion it committed in the 15 years prior. Scaling up will also require action from governments. Too often, Kettle says, policymakers stimulate demand for green technologies while also enacting regulations that make it harder to mine the materials needed to power them.
In Zambia, embracing the green energy revolution is a national matter. A new government elected in August is seeking to turn around an economy ravaged by debt, and minerals, which account for three-quarters of exports, are essential to that puzzle. Minister of Mines Paul Chanda Kabuswe says the “looming boom” in battery metals has the potential to bring Zambia “immense benefits.” Still, no major deposits have been discovered in decades. To secure supply for the long run, the mining industry will need to do better.
Humphrey Mbasela, a Zambian geologist helping KoBold make sense of the soil in Mushindamo District, believes the big data approach will help. For too long, he says, explorers have been “chicken scratching”—too focused on searching near the surface, while the big prizes lie deeper.
“The resources are there, undercover,” he says after a day trudging through woods and fields gathering samples. “They’ve just not been unraveled.”
"help" - Google News
December 04, 2021 at 01:16AM
https://ift.tt/32Z0T6b
Electric cars are powered by rare metals. Can AI help find them? - National Geographic
"help" - Google News
https://ift.tt/2SmRddm
Bagikan Berita Ini
0 Response to "Electric cars are powered by rare metals. Can AI help find them? - National Geographic"
Post a Comment