Speaking at the Future of Australian Mining in Sydney, Dr Penny Stewart, Managing Director of Petra Data Science, said that bringing transparency to the region is something her company is working on.

"If people don't understand what's going on behind the scenes, they don't trust it, nor do they trust it," she said.

“With these techniques, everything is good, everything is good.

“Early adopted people grabbed it and did a great job.”

However, the problem with this black box approach arises after those early adopters continue to move forward.

Those who follow them may be less willing to accept what the black box spits out to them than their ex.

They eventually returned to their past methods and lost the fruits of artificial intelligence and machine learning.

"We have had this experience in a lot of companies, and these companies have excellent work," Stewart said.

“In one case, Geo has had a great success. He published his paper on it.

"Then he left and the people who took over were more conservative. They stopped using it even though it was working for them all the time."

Data 61 Senior Research Engineer Dr. Dave Cole said at a recent CSIRO Resource Innovation Showcase that it is important for miners to be critical of machine learning and artificial intelligence results.

“If you give me some data, I can give you a model,” he said.

“But you need to have humans do some geological work in the area to drive that criticism.”

In another way, pay attention to what is in the input box to better understand the results.

“We have tools called shape diagrams,” she said.

“They show you the characteristics of the model. It gives you the relevant importance of model input.

“Some things that people think are important are not necessarily that important.

“Those show what is really driving them to optimize.

“Sometimes, when you show the plot, there’s a real 'wow, that’s so fun'.

“There are then some graphs showing the relationship between different parts of the geology.”

This is important, Stewart said, because engineers and geologists tend to study one-way processes.

"If you try to increase the yield of the crusher, you think if hard rocks pass through, it will slow down the crusher," she said.

“But it might be that the softer rock slowed down.”

It is in the correct combination of hard and soft rock that the optimization can be found.

"This is an example of the complexity that machine learning can bring to you," Stewart said.

"People know that this exists. They see it empirically. But how do you build a model to capture that?"

Stewart said that with the popularity of machine learning and artificial intelligence in the mining field, the need to make things more transparent is also growing.

She said: “Geologists want to know what they get.

“Now, as we become more mature and larger in the market, a lot of our cooperation with our customers adopts this glass box method.

“It also allows people to interrogate models.”

Dingo is another company that processes a lot of data to help miners optimize results, although its approach focuses more on improving equipment uptime.

Paul Higgins, the company's executive chairman, said the market has reached a point where many customers are over-hyped but are wary of some of the big investments that have been made in the past that have nowhere to go.

"Dingo's approach is to work with clients with special issues, which is important enough for them to give us some data," said the company.

“Our field experts and their field experts have come up with something valuable.

“We do see that we need to take customers on the journey.

"Otherwise, their default position is, at the end of the day, if these things don't work, I'll have to go back to what I know. If you ask me to do something different than this, I need to understand because of the risks Very big."

Another part of Petra's work is to apply its machine learning and artificial tools to platforms miners already use.

"If engineers can use these tools in normal workflows, it will be easier for them to use them," Stewart said.

“Especially last year, we spent a lot of time building partnerships to make them look like they always are.”

One example is Petra's collaboration with the environment to put its tools in.

"In the processing plant, the output from Petra's Maxta goes into the pie, which the operator uses to control the factory's screen.

“We can send our output to their control screen.

“The feedback we have received is that it is getting more and more difficult to find senior engineers in the field of mine planning.

“If they have a tool that will help them get the job done faster, they will be happy because they are understaffed.”