AI in Mining: Myths and Facts

AI helping underground mines

Can Artificial Intelligence be applied in the underground mining sector? Or is it a myth?

Maybe because it’s so new, and for so long has been in the realm of science-fiction, but Artificial Intelligence is surrounded by a number of myths. So too is Machine Learning, AI’s little sibling.  

But it shouldn’t be the realm of myth. AI may seem intimidating, but it won’t take over your mine. In the contrary, efficiency is what AI and ML target so that operations are optimized.

To help bust some of those myths, we’ve put together a guide to address some of what you may have heard about Artificial Intelligence and Machine Learning, and where reality truly exists. 

Myth: Machine Learning and Artificial Intelligence require huge investments only big corporations can make.

Fact: In underground hard rock mines, it’s easy to pick up quick wins on small projects. AI can do things as simple as sifting through the maintenance data you already have to help predict when haulage trucks will break down. 

Myth: AI and ML will replace all the people in your operation with algorithms

Fact: The technology doesn’t replace people at all. What it does is enhance workers’ ability to do their jobs with better information. Read our blog about ML algorithms do just that! 

AI improving Underground Hard Rock Mining operations

Myth: Working with AI is like working with a sentient person.

Fact: AI and ML algorithms are far, far away from human-like consciousness. There are ways in which algorithms are able to mimic some particular, narrow functions of the human brain. For example, they can play and learn to excel at games like chess or GO. But an algorithm designed for chess is never going to be able to have a conversation with you, and a conversational robot isn’t going to play chess. AI-driven technologies are designed to perform specialized tasks. And even if they can sound human, and can learn certain human thinking patterns, they aren’t nearly at the stage of full blown sentience.  

Myth: Machine Learning can work with any quality of data without problems.

Fact: While ML algorithms can work well with data that includes a lot of noise, there is a limit. If data is of poor quality, no matter how sophisticated the algorithm, it can’t be used. You can have mountains of data that is missing key information. Consider a mountain of data from a fleet of trucks, but the data is missing a timestamp—so the algorithm doesn’t know what day or time the data is from. Or maybe the data doesn’t identify what sensor it’s from—is it transmission temperature, or tire pressure? Automatically collected data by machines also tends to be more reliable than data manually entered by humans, particularly if it’s been over a long period. 

Myth: Artificial Intelligence and Machine Learning are dangerous.

Fact: Like any other piece of technology, AI is fallible. When using AI—or any technology—you should do so with a healthy skepticism. It’s important to counter-verify anything output from an AI or ML system. Like with any other new technology, you should make sure there is a person in the loop of the system. Not only will you prevent errors, but you will enhance that person’s ability to do their job. It has the added benefit of removing the repetitive, perhaps boring, tasks from many jobs, leaving the more interesting, in-depth analytical tasks for the user. 

Myth: Machine Learning is perfectly objective.

Fact: There are many ways bias can find its way into Machine Learning. Classic ML techniques rely on the engineering that created them. And the biases of the people that built those algorithms can shape and bias the system’s output. On the Deep Learning side of ML, data is given to the algorithm in the rawest possible form. In that case, the system can only be as objective as the data it is given. Any information given to an ML system by a person is subject to the biases of that person. The algorithms learn from what they are given. 

CIM addressed the power and pitfalls of predictive algorithms applied to mining last year through this article – an interesting read for more perspective.