Mining exploration becomes smarter with AI
Tech may provide next-gen standard for mineral discovery Metal Tech News Weekly Edition – January 15, 2020
Last updated 6/27/2020 at 5am
A promising concept in the mining industry, utilizing the technology of artificial intelligence to accomplish never before dreamed of precision in mineral deposit discovery.
While not an entirely new concept, as companies become more familiar with machine learning, the rate of success in its calculations can only grow.
Finding buried lodes of metals that are economic to mines has become more difficult with time. If you examine the past twenty years of exploration expenditure for gold, spending peaked in 2012 at US$6.1 billion. Since then, only four major gold deposits have been discovered.
Before 2012, the industry found an average of ten of these globally significant precious metals deposits per year.
With less capital being deployed and deposits being harder to find, the mining industry has largely sought out new metals deposits under the same system it had for decades.
But now a new method has been devised, powerful computers and abundant data have allowed researchers to develop programs to help them navigate a sea of information to pinpoint desired outcomes.
Controlling millions, even billions, of points of data, researchers can now streamline the mineral exploration system and improve the odds of making an economic discovery, all through the functions and power of artificial intelligence.
What is behind AI?
To understand how AI can help geologist find new deposits of the metals the world needs and wants, it is helpful to know how it works and what powers it.
First, there are several different methods that companies employ to operate different functions with different AI.
That's right, there is more than one way a computer thinks.
An ordinary computer does one thing at a time in a distinct series of operations known as serial processing.
A supercomputer can tackle complex calculations much more quickly by splitting problems into pieces and working on many pieces at once, which is called parallel processing.
Some companies use supercomputers as the incredible volume of data and necessity for correlation between that data needs to happen very quickly.
Voice recognition technology is a great example of necessary computational power as it breaks down sound digitally to sync up with matching voice patterns in its database.
Supercomputers, however, tend to be enormously expensive.
Shared or distributed computing, which interconnects multiple computers to function as an imitation supercomputer, is one means that has been developed to overcome this expense. Networked together, numerous computers communicate to handle singular problems.
Most interestingly of all, however, is using graphics processing units (GPU) to achieve parallel processing withing a single computer, a solution companies such as Google and Facebook are pursuing for AI.
Originally GPUs were designed as co-processors that operated alongside a computer's main central processing unit (CPU) in order to off-load demanding computational graphics tasks.
This means that GPUs already carry out the kind of parallel processing that gives supercomputers their name.
One of the reasons GPUs have emerged as the supercomputing hardware of choice is that some of the most demanding computational problems happen to be well-suited to parallel execution.
The prime example of this is deep learning, one of the leading edge developments in AI. The neural network concept that underpins this powerful approach – large meshes of highly interconnected nodes – is the same that was written-off as a failure in the 1990s.
But now that technology allows us to build much larger and deeper neural networks that achieves radically improved results. These neural networks power the speech recognition software, language translation, and semantic search facilities that Google, Facebook and many apps use today.
Simply put, artificial intelligence is just extremely fancy software, not unlike the solitaire game on your computer. But as it requires incredible processing power to function accurately it is, of course, in its own league.
It is this incredible processing power that is showing promise for finding the increasing elusive deposits and other metals.
AI in Geoscience
Currently, mineral exploration begins in the mind of a geologist, who must consider an enormous amount of geological, geophysical and other data to plausibly make the first pass.
This is a great deal of time spent calculating and determining the viability and potential of mineral fields.
Yet, just one out of 3,300 mineralized anomalies become a mine.
Speculation in mining could perhaps be best equated to the Gambler's Ruin Problem, the odds of finding an economic deposit are stacked against the explorer but the returns on an economic deposit are enormous.
So how does AI improve the odds for the explorer?
Now it could be argued that by giving a person the exact same data, they too could, logically, determine a hypothesis and narrow down the parameters for a mineral discovery.
A computer, however, can infer more with less and does it more quickly, much more quickly.
Take for example the fundamental reason any mineral deposit forms.
AI can link this reason to available geoscience data to determine the correlation. That correlation can then predict the likelihood of mineralization in new exploration regions.
There are just too many factors to be considered to accurately determine mineral field locations with just sheer willpower, it becomes impractical. There are over a dozen data layers involved with accurately determining potential mine sites.
And a data layer is any of the relevant data of a given geoscientific field, and even many other fields for that matter.
Geology, remote sensing, geochemistry, gravity, aeromagnetic and radiometric surveys, digital terrain, regional structure, known gold deposits, the list can go on.
Now there is a tool that geoscientists use to help them navigate the data and create a composite to help them better understand their desired outcome.
A geographic information system (GIS) is a technological tool for comprehending geography and making decisions. GIS organizes geographic data so that a person reading a map can select data necessary for a specific project or task.
A good GIS program is able to process geographic data from a variety of sources and integrate it into a determined outcome based on selected datasets, this is ideal in determining exactly where one might discover a mineral field.
With a GIS utilizing the vital data for comprising the layers involved in determining a potential lode source, you can see just how exhaustive sifting through such information is, it becomes nearly incoherent.
However, when input into a computer using AI, the game changes. There have been several companies that have begun to use AI to help with mineral discovery, and one appears to stand above the rest.
GoldSpot Discoveries Inc. is heralding the future of mineral discovery by utilizing artificial intelligence on a regional and localized scale.
GoldSpot employs its own algorithm to significantly decrease the risk of mineral exploration while increasing efficiency.
A machine learning algorithm means that a specific process is driven by a computer to link data. Much like human thinking, similarities, patterns and practice help us to determine what is what in the world.
GoldSpot is utilizing a powerful array of computers that can deduce possible mineral deposits with parameters that their geologists and engineers determine based off of GIS mapping and other factors.
The machine learning algorithm developed by this innovative company has successfully processed billions of data points and most successfully in one of Canada's most abundant mining regions.
In the Quebec's Abitibi region, the GoldSpot algorithm identified 86 percent of existing gold deposits but required as little as 4 percent of the surface area to actually determine the prospective ground.
This is an incredible achievement.
Their algorithm was able to deduce potential mineral resources with very little data. This is because the amount of data it had been fed worked like a calculated checklist and fact checker, comparing and determining, all nearly at the same time.
This could be compared to the gameshow Wheel of Fortune, where you guess what the phrase is by analyzing the other available data. That is what AI is capable of with the geological and other data used to find mineral deposits.
The future of all mining could become calculated and guesswork would become null as more data is accumulated. The accuracy will only continue to increase as factors change and incorrect data is filtered out.
With machine learning, mineral exploration has the potential to become an exact system and with companies capable of commanding powerful supercomputers to locate potential lodes, the guesswork would become redundant and mineral deposits could become a checklist.
The future that GoldSpot is creating is truly a bold one for the mining industry. Many old hats may find it vilifying but the future of artificial intelligence is as unbiased as its function and its prominence in mineral exploration will only continue to grow.