Mineral Exploration 4.0
How would discovery rates change if we were more open with data? How is machine learning and artificial intelligence changing the way we explore and target deposits?
These are questions that were discussed during Mineral Exploration 4.0 at Prospectors & Developers Association of Canada (PDAC) 2019. Hosted by OZ Minerals and Unearthed, the discussion arose as OZ Minerals launched their open data Explorer Challenge, live from February to May 2019. For the crowdsourcing competition, OZ Minerals has opened over two terabytes of their private exploration data from the Mount Woods project and put up a A$1 million in prize pool. Their hope is that a global community of data scientists and innovators can help them find the next economic deposit in their tenement.
The panelists at the PDAC event included Richard Holmes, Head of Exploration and Growth for OZ Minerals, Brenton Crawford, Founder and Director of Solve Geosolutions, Steve de Jong, CEO of Vrify and former CEO of Integra Gold and Tim Dobush from the Frank Arnott Award Committee.
Here is a summary of the discussion:
Sharing data with the world: value add or erosion of competitive advantage?
Historically, the exploration and mining industry has been very closed with their data. Data itself has been viewed as a competitive advantage. Exploration companies have tightly held on to knowledge of how they go about finding ore bodies, fearful that if shared openly, this would unlock the key for others to find deposits.
Has this focus on risk prevented the assessment of how much additional value could be added from sharing data? Are we slowing down the discovery rate, not only for ourselves, but also for everyone else by being so closed?
Richard shared OZ Minerals’ case for opening over two terabytes of their private data. Mount Woods, a large project surrounding the Prominent Hill mine, has been explored by OZ Minerals for around the last 10 years. Large amounts of data have been acquired as the project was assessed, largely for another IOCG (Iron Oxide Copper-Gold) deposit.
As owners of this project, OZ Minerals see only a value add opportunity by sharing this data. They believe that by providing others who have different experiences, knowledge and background with access to the data, they have a high chance of finding something different in that data set.
If you own the ground, what is the risk? The argument that someone could find something in the data, wait for you to drop the ground, and then drill it, is a weak one at best.
Ultimately, by being closed with data, contrary to gaining a competitive advantage, we have most likely missed out on unlocking a large amount of value.
Why aren’t others following the lead?
Back in 2000, Goldcorp CEO Rob McEwen put forward the company’s exploration data to the public in a similar incentivised competition.
This helped Rob and his team turnaround the company and their Red Lake project from a struggling junior to a leading gold miner, with an additional CA$6 billion worth of gold identified by the competition.
Considering this success, it’s surprising that the next time that industry opened up their exploration data wasn’t until 2015. Steve de Jong and Integra Gold opened up the data for two historic gold mines they had recently acquired. Recognising the need to work through 75 years of historic data, Steve saw the opportunity to use the collective power of the crowd to work through this data much faster and indicate the best locations to drill.
How do we go about being more open with data?
Geologists that have worked in Australia, Canada or Scandinavia are likely familiar with the large amount of public, open data that is provided by these jurisdictions’ geological surveys. This is typically a combination of data collected by the geological surveys themselves (i.e. regional geophysics), and privately acquired data that has been released due to regulations which mandate that data becomes public over time. Australia has arguably led the way with the regulation side, with clear processes and frameworks for submitting and releasing data.
Most of us understand the huge value that this resource provides us. The ability to review this data often dictates the projects we acquire and the ground for which we apply for licences.
Governments have a clear incentive to push for policies like this, as it enables them to demonstrate the prospectivity of their land, which attracts exploration companies and boosts the local economy. Globally, the exploration budgets for Canada and Australia are on a par, and far outweigh those of any other nation. This is arguably in part due to their open data policies. Globally, awareness and communication of examples of how open data has driven economic development at a government and private level, should see an uptake in the approach.
But it’s not just about opening data, it’s about how you access it
An expert in data management, Tim shared the need to think not just about making our data open, but about how people are going to access and use the data. We need to be thinking about how we design the structures and services that make data accessible. By default, as regulators, governments are custodians of data and design processes around its submission and release. It doesn’t necessarily follow that they are best equipped to manage and promote the data. There is a need for more collaboration between industry, academia and governments in order to ensure that, as we build more access to open data, it’s done in the way that best meets industry’s needs.
How does machine learning play a part? What’s the hype and how does it really work?
Machine learning (ML) and Artificial Intelligence (AI) are some of the most overused buzzwords right now. This means that there is a lot of hype around them, along with a general lack of understanding and regular misuse of the terms and their applications. Brenton, a geologist and data scientist whose business applies machine learning to exploration targeting, kicked off the discussion by debunking some of this hype.
“The difference between machine learning and AI? If it is written in Python, it’s probably machine learning. If it is written in PowerPoint, it’s probably AI” — Matt Velloso.
Applications of machine learning to exploration are often thought of as black box approaches, because people don’t really understand them. A considerable amount of the work involved is getting relevant and clean data sets, which is one of the biggest challenges and barriers to applying ML to exploration targeting. Historic data is full of errors, is messy, and often redundant because it isn’t relevant to the question that we are asking.
Often, we may not have access to that many layers of useful data. A typical scenario is a project where only magnetics and gravity and some limited chemistry is available, and at a low resolution, which does not give many variables to train a model on. Where we do have lots of data (and geologists love collecting a lot of data), we often don’t consider why we are collecting it, so we may not collect it in a useful or consistent way. Consistency is a huge problem, particularly with chemical assay data. The high level of variability across the data in methodology, elements and detection limits makes it very difficult to use in ML approaches. Consistency is arguably more important than getting that extra 0.01ppm lower detection limit.
Practitioners, such as Brenton, do not claim that ML is a silver bullet for finding the next economic deposit, rather that it is an approach that allows you to combine many layers of data and identify key relationships between them, in ways that we haven’t previously been able to.
We need to be considerate of how we use and collect data to use this approach successfully. When we do, machine learning combined with domain knowledge will be a powerful new tool for geologists in assessing their projects.
What’s the link with open data?
Algorithms are trained, or learn, from known data points in order to predict where new ones may occur. A typical ML approach is to combine multiple layers of data surrounding a known deposit and look for similar signatures and relationships elsewhere. The more known deposits a ML algorithm is able to train on, the better that model will be at predicting the location of unknown deposits.
There is a very clear benefit industry-wide to share data in order to build the effectiveness of these models.
The ability to interact with diverse skill-sets and abilities
Machine learning is an evolving field, and is complimentary to other statistical techniques we have previously used in prospectivity mapping, including PCA, clustering etc. Another one of the main drivers for OZ Minerals presenting their data to the public is to access a diversity of approaches. Geologists, working in their field for many years, have tried and tested methodologies, but often do not have access to data-driven tools and approaches used in other disciplines and industries.
By pushing data outside of the organisation, OZ Minerals is able to get the best of many answers to their problem, from a huge range of approaches and recombine these to assess a way forward.
How is machine learning and open data changing the investment landscape for exploration?
A key driver of change is communicating early successes and opportunities. Steve shared his experiences running the Integra Goldrush challenge, and the difficulties of communicating this opportunity to the global market. If we want more people to work on our data, they need to know about it. We might think $1 million is a good enough incentive, but that alone doesn’t dictate that a story is shared far and wide.
Mining does not have a positive public brand, which leads to the broader investment community often sharing the same sentiment. However, this may be changing.
Recently, the Bill Gates backed Breakthrough Energy Ventures, based in Silicon Valley, invested in KoBold Metals. This is the first time we’ve seen a significant, publicly discussed investment from the valley in exploration tech and the language is all about the search for ethically-sourced materials. The dialogue is similar to that which we saw around the boom in supply chain blockchain startups focused on tracking ethically-sourced materials from source to product.
However, as we know, investment is driven by economic decisions as well as ethical ones. Another potential change from the use of ML in exploration, is the ability to more quantitatively explain to investors the risk and potential return on their investment.
This could also see a shift in the type of investors attracted to the exploration market. Again, as Steve points out, this requires communication of success stories to build the case. We need to publicly share examples of where ML has increased our success rate, as well as decreased costs and time, to boost investor confidence. Hopefully, with the likes of companies such as Goldspot Discoveries publicly listing, we will see a significant increase in the success stories shared.
Do we have the skills and capabilities to do this?
Any discussion around future business models for exploration and ML needs to consider where these skill sets are going to come from. The companies on the market have a good balance of geoscientists and data scientists on the whole, but are we focused enough on upskilling our future geoscientists to use this tech and be able to think and deal with large multilayer datasets?
Tim, through his role on the Frank Arnott Award Committee, is driving this change by upskilling students. The competition is designed to teach students real world applications of effective data integration and visualisation, arming them with the skills they need to succeed in industry 4.0.
A final word
Thanks again to all of the panelists and PDAC attendees for being part of the discussion. This was certainly a conversation worth having and a necessary one to consider how exploration may change in the future. The panelists all had different views in the level of change we will see in the next 5–10 years.
My position? If we’re seeing some of the world’s richest and most impactful investors focus their gaze on finding necessary minerals, change cannot be far behind.
Who are the players right now?
As always PDAC is a great opportunity to get a broad snapshot of who is doing what in the industry, and you may find there is a surprisingly higher number of startups and junior companies focused on ML in exploration than you may have thought.
Here’s a quick list to check out. Please comment if you see some that I have missed:
OreFox, Earth AI, Solve Geosolutions, Goldspot Discoveries, Koan Designs, Complete Target, Minerva Intelligence, Albert Mining, Azimut Exploration, and Quantum Discoveries.
