Richard Taylor (MBA GAICD) is a mining executive with a love of new products and new markets. Richard is CEO and director of SensOre Ltd., applying big data Artificial Intelligence (AI) and Machine Learning (ML) to resource exploration. SensOre aims to be at the front of a new wave of discovery generated from AI innovation and technology. Richard has held senior roles with Mineral Deposits, PanAust, Oxiana, and the World Bank, including stints in Asia and North America.
Thanks for joining The Assay, Richard. Can you start off by giving a short intro to SensOre?
SensOre is part of a new breed of mining technology companies that are translating developments in adjacent industries into mining. AI and ML have held promise for geoscience for some time, but it has taken advances in combining geology and data science with computing capacity and new algorithms to realise the opportunity. SensOre does what geologists could if they had the time and ability to visualise and consider all of the available data. SensOre’s data cube has 1500 data layers and more than 14 billion data points, and this is growing all the time. That’s a lot of data – too much data for a geologist to ever make sense of conventionally. Many of the targets SensOre has developed make use of this problem. The technology filters out the noise to identify patterns that, when reviewed, stand out from the data conventionally.
We started with a vision to organise all of the data for the Eastern Goldfields of Western Australia and make use of this to define areas of high prospectivity. We quickly found that organising all of the available geochemical, geological, and geophysical data, derivatives, and imputation layers gave us fantastic visibility into other companies’ areas. Now, we can see discoveries before they’re announced and see projects that are being drilled out even when no announcements have been made. These insights have turned into client relationships as we develop the technology as a service to others.
“AI and ML have transformed retail, transportation, mining and processing. Why shouldn’t they have the same impact on exploration?”
That said, we are mineral explorers at heart and some opportunities on open ground have been too good to pass up. We have acquired or joint-ventured around nine of the most prospective targets generated through the technology and validated conventionally, and will be drilling many of these over 2020 and 2021. Ultimately, we want to be in a position to organise all of the world’s geological data, from which the mining sector and host governments would benefit the most.
Can you explain the technology and why it is so unique?
The data cube and Discriminant Predictive Technology (DPT®) have been developed by
explorers for explorers. The brains behind this has been our CTO, Alf Eggo, who has been working on machine learning in exploration since the early days of the discipline in the 1980s. It is not the product of data science magic, but the result of hard work and dedication. DPT® respects all of the knowledge that has been created since geology was first established as a discipline. We’re not out to replace geology, or geologists for that matter, but to improve performance with the best that developments in AI, ML, and computing capacity can deliver.
Most of the competition in this area have adopted quick and cheap alternatives to generate insights from the data they can access from a company that they deem to be of good quality. Many people baulk at making use of all of the data available and building it for an entire terrane, state, or continent. SensOre has tackled this challenge head-on because the big prize is being able to move from data to drill target with confidence. This can only be done with a big data approach.
DPT® also provides targets on a scale that is of use to mineral explorers by providing economic markers to these targets. SensOre’s technology predicts location, grade, and depth at a cell level sufficient to be able to site a drilling program. This supplants the expensive and time-consuming process of target generation and validation. The validation is all there in more detail than most people can make practical use of without advanced computing practices.
How does the data cube work?
To begin with, think of conventional geology as about organising the relationships between several important data sources in order to generate a target. This might be gravity, magnetics, geochemistry, etc.. If you think of each data set as a layer spread out over the area of interest, then a geologist is looking for patterns between those layers. The data cube mimics this very human approach to conceiving relationships between and within data. However, the data cube has many more layers of data, including the derivatives of those data sets.
Each area of the terrane is broken up into a cell and each cell is populated with data from each of those layers. The data cube ingests data from many sources and grows in predictive capability. It also gets better at recognising and cleaning errors as anomalous data is flagged and stands out from the other data layers. This is increasingly an autonomous process and is developed from completely separate ML approaches to the predictive analytics we use in generating targets.
Does the technology have a lower carbon footprint than traditional methods?
Currently, there are a lot of inefficient practices in generating each mineral discovery success story. Based on industry CO2 emission data, each discovery comes at the cost of testing around 200 targets which generates between 14,000 to 29,000 tonnes of CO2 per discovery. Adjacent industries, such as oil and gas, rely more heavily on data science before drilling. Less targets are tested for each new discovery. An uptick in exploration discovery rates, particularly in reducing inefficiency in greenfield exploration by going from data to drill target, will deliver order of magnitude improvements in CO2 emissions from mineral exploration.
How does the technology increase ROI?
As above, DPT® promises order of magnitude improvement in discovery rates, given the relatively low success rate of greenfield exploration i.e. less than 1 percent.
Can you expand on some of the projects that SensOre is working on in Western Australia?
We have some pretty exciting targets in the Western Australia goldfields. While they were identified by AI and ML, they also stand out in their own right conventionally. They are all in good neighbourhoods and stood out to us from the data, as being promising drill targets. In early 2020 we completed an infill gravity program at the key sites and have been on-site preparing for our upcoming drill program. Two of our top targets are joint-ventures: the North Desdemona earn-in with Kin Mining (ASX:KIN), and the North Darlot earn-in with a private group, which is near ongoing exploration by Red 5 (ASX:RED) and proximate to the Darlot Gold Mine, which has produced 2.8 million ounces over 30 years.
We aim to test up to seven targets this year in order to validate the technology and input feedback into the system. Of course, not every target is expected to generate mineralised intercepts. The system predictions are probability-based and we are preparing to test a number to generate the validity required.
Are there other opportunities for DPT to be used outside Australia?
SensOre’s technology will expand internationally in the next 12 months. There are logical areas, where large public and proprietary data sets exist, in which the approach will work best. These are mostly in Europe and North America. However, SensOre’s approach is simply to organise and use all of the data available. The system ingests data in multiple formats quickly and efficiently and can be applied to areas where data can be scarce, such as in Africa. It can also grow over time in circumstances where data resides in private companies, as is the case in some parts of South America. The system will continue to improve as it grows, and learning from one terrane to another is not only possible but probable.
Is this technology proprietary? Could other mining companies license the technology?
The technology is proprietary and at present we are working with clients on target generation initiatives directly. The value uplift of DPT® is significant, so our model is leveraged towards benefiting from the value uplift generated from new discoveries. We are looking at developing the technology further towards software as a service that will allow more rapid data integration and provide clients more interoperability. The initial results are promising and we are looking at developments with major tech companies that we could leverage from to deliver an effective customer interface.
What are some of the near-term milestones we can look forward to?
Certainly the testing of our portfolio of highly-prospective gold targets over 2020 and 2021 will be one of the most keenly awaited developments from our investors. However, I am most excited about developing our client base and enhancing our customer interface, so that more companies can benefit from the big data sets we have put together and the predictive software we have developed. AI and ML have transformed retail, transportation, mining and processing. Why shouldn’t they have the same impact on exploration?