Artificial intelligence is making mining more efficient, but it comes with its own hurdles, executives at some of Canada’s largest gold miners told a session at the Prospectors and Developers Association of Canada’s annual convention this week.
AI or machine language learning has the advantage of being able to crunch large data sets and improve its output from experience. It’s being applied to geological figures to suggest drill sites and to run autonomous haulage trucks, for example. It’s set to generate millions or billions of dollars in savings.
Yet, data must be vetted and operations monitored by humans while questions of stock market compliance and security remain to be smoothed out. It requires skilled programmers and expensive technology, and removes employment from remote communities who depend on mining for job creation and northern development.
On the positive side, Agnico Eagle Mines (TSX: AEM; NYSE: AEM) says it’s testing an AI application that may increase production at Canada’s second-largest mine by output, Detour Lake in northern Ontario, by 5% a year or 35,000 ounces. It sees potential in repeating the technology at other mines to compound the impact.
On the caution side, there’s an AI skills shortage, Andre Leite, vice-president for Ontario operations, told a conference panel on gold mining technology in Ontario.
“It’s not something that is commonplace,” Leite said. “You need the skill set to be able to get the value out of it.”
Appropriate skills
Kinross Gold (TSX: K; NYSE: KGC) reported similar potential and drawbacks. AI used in automation should create incremental gains and its help in geochemistry can be worth hundreds of millions of dollars, but mines need appropriate skills, said Luke Jalsevac, vice president responsible for studies and advanced exploration at the Great Bear project in northern Ontario.
“You can have the brightest of the brightest on the tech and AI side, but if they don’t understand mining then we run into a lot of issues,” he said. “It’s not intuitive to understand how challenging it is, whereas if you come from making cars, you can control your inputs, like Tesla. We still are just based on data points that we’re extrapolating.”
Kinross plans to complete a preliminary economic assessment at Great Bear, 500 km northwest of Thunder Bay, late this year. Surface construction work may start at the same time for a mine to produce 450,000 oz. to 500,000 oz. per year. The company is using AI to build algorithms to analyze geochemistry with 50 to 100 different variables and 300,000 samples.
“We run into issues where we’re just not able to manage that much data and this is where the AI piece, the machine learning piece, is coming in and we’re finding a lot of benefits,” Jalsevac said.
“If we mine effectively, efficiently, but we don’t understand the geology that’s a huge value destruction because we’re just mining the wrong things more efficiently. One of our goals is to use AI and machine learning to help us better understand our geology. We’ve been pretty successful in that regard.”
Virtual reality
AI helped Iamgold (TSX: IMG; NYSE: IAG) engineers design a compact mill at the Côté project in Ontario by powering 3-D imagery.
“We have a mill that’s pretty compact, it’s really fit for purpose,” Michel Payeur, vice-president of technical services, told the panel. “We were able to model it that way in virtual reality.”
Located 125 km southwest of Timmins, Côté is a 60:40 partnership with Japan’s Sumitomo Metals. It is to pour its first bar this month and begin producing 495,000 oz. a year from the third quarter as Canada’s third-largest gold mine by output.
Payeur said one of the challenges is showing AI’s work to industry regulators.
“The issue with AI that we have today is a little bit more on transparency,” he said. “From a governance perspective, it becomes a challenge to try and use these technologies that have a certain amount of opacity to them and to bring them in progressively into the portfolio and to rely on them as you have to disclose information.”
Agnico, the largest miner by market value after Newmont (NYSE: NEM, TSX: NGT) and Barrick Gold (TSX: ABX; NYSE: GOLD), has been applying machine learning to core logging and is considering more uses, Leite said.
“If you put 20 geologists in a room, you’re probably going to get 20 very different models out there, so this allows us more consistency on that process and with large datasets allowing us a much higher capacity to process information,” Leite said. “We’re looking at applying some of the consistency you need in core logging, resource modeling and transitioning to mineral processing.”
Labour shortage
The skills gap mentioned is compounded by a labour shortage forecast at about 20,000 workers over the next decade across all its sub-sectors, according to the Ottawa-based Mining Industry Human Resources Council. Agnico is preparing to bring in a dozen mechanics from Mexico to work at its Macassa mine in Kirkland Lake.
Stagnant population numbers in Ontario’s north and challenges convincing workers to move to say Kirkland Lake from even Sudbury, plus lack of housing on top of the skills shortage makes the situation complex, Leite said. People may not want to work underground, but skilled employees in the gaming industry might compete among others to start at a $60,000 salary while in mining they would easily earn more than $100,000 a year, he said.
“It’s about making that awareness, especially to the parents,” he said. “The kids, sometimes there’s this vision of what they want to be in versus what the reality is, but in northern Ontario there is a lot of youth there.”
Kinross, which is returning to Ontario with Great Bear, is re-engaging with schools to build interest in mining and may link them with a training program at the Round Mountain mine in Nevada, Jalsevac said.
Payeur said the biggest challenge for the industry is not if AI, automation or other new technology will erase jobs in northern communities that may depend on mining for employment and development.
“The biggest challenge for the industry is how to reach out to students today to get them to enrol in mining programs,” he said. “To convince students today that mining isn’t something from the heyday of uncle whoever, and isn’t low tech, uninteresting, dirty and dangerous.”
See some hope here. Reality may bite but I see some people who are embracing reality. They are more likely to succeed. AI may be able to analyze large data sets but if the AI programmer does not know the business the result will be failure. Ultimately, because of their ability to judge data sets, humans will out perform AI.