The shift towards achieving a Net Zero economy also requires the extraction of various organic materials for the new infrastructure, a challenging reality. Electric vehicles are set to replace traditional machinery, with solar and wind generators, wires, transformers, pipes, and batteries becoming essential components. The construction of these elements will demand the discovery of millions of tons of fresh metals, many of which pose sourcing challenges due to local community issues or political complexities.
Earth AI has made significant strides in uncovering these valuable resources in untapped regions. In contrast to exploring near existing mines, the mining industry refers to these endeavors as greenfields. Recently, in Australia, the company identified a substantial alloy deposit with double the concentration found in the largest operating mines.
Even more noteworthy is their innovative approach, boasting a success rate of 1 in 8 for new discoveries compared to the industry average of 1 in 200. Their findings, averaging 1 in 20, surpass the outcomes of traditional exploration methods.
Considering the substantial costs associated with drilling survey holes, Earth AI has revolutionized the process by introducing a more systematic approach using compact components to keep drilling expenses low. This streamlined method not only enhances the efficiency of creating AI models but also accelerates the empirical process loop.
Roman Teslyuk, the CEO and founder of Earth AI, emphasizes the pivotal role of AI in maximizing mining profits and extending mine operations. While major mining companies have leveraged AI to their advantage, startups like Plotlogic, MineSense, GoldSpot, and StratumAI have been at the forefront of technological advancements in the field. Earth AI stands out by focusing on discovering new mineral deposits in greenfields rather than reanalyzing existing data-rich information, a strategy that presents greater challenges but also offers substantial rewards.
Embarking on a Career as an AI Scientist
Exploring opportunities in new mining territories presents unique challenges as greenfields lack the abundance of data found in well-researched brownfield sites. While acquiring new data is the conventional solution, it is costly and time-intensive.
Earth AI took a different approach by developing a more efficient method to interpret and classify existing data, obviating the need for extensive new data collection. By training their system on 400 million geographical cases from exploration records, the company overcame significant hurdles. Teslyuk explains the complexity involved in training the deep learning model to focus on relevant geographical factors, enabling the AI to analyze and interpret data akin to a geoscientist but on a larger scale, leading to consistent and reliable estimations.
The process comprises three key phases:
-
Targeting: Utilizing the vast dataset from Australia, Earth AI’s models pinpoint areas with high potential for metal system discoveries. Geologists validate these targets on-site, leveraging a complementary technology to enhance their understanding of the geological context.
-
Drilling: By drilling to a depth of 600 meters to confirm the presence of mineralization, the team tests their hypotheses. The data from each drill hole feeds back into the system, refining assumptions and generating new insights.
The Significance of High-Quality Data
Earth AI assembled a specialized team of geography and deep learning experts, emphasizing continual experimentation to distill key elements for their geographical deep learning system. Over six years of analysis and 600 rounds of testing, they honed their predictions and refined their methodology based on real-world feedback.
Quality monitoring of data across multiple continents with millions of data points posed a challenge, prompting Earth AI to develop a semi-automated expert-driven data evaluation system. This domain-specific program focuses on identifying and rectifying information errors and inconsistencies, a critical but often overlooked aspect by AI companies.
A Systematic Approach to Drilling
In a departure from industry norms, Earth AI incorporated drilling as an integral part of their data collection process rather than outsourcing it. This strategic decision streamlined operations and facilitated subsequent steps in the exploration process.
The company revamped drilling technology to be adaptable, self-sufficient, and environmentally friendly, integrating waste-treatment measures to minimize environmental impact. Moreover, they optimized transportation logistics to expedite the analysis of extracted samples, reducing the turnaround time between sites from weeks to days.
Teslyuk underscores the importance of aligning equipment with drilling requirements to enhance scalability and accelerate discoveries, emphasizing the incompatibility of legacy systems with high-performance operations.
Final Thoughts
While the allure of cutting-edge AI solutions in the mining sector is undeniable, Earth AI’s focus on enhancing the affordability and interpretability of tests offers a more pragmatic approach. By discerning where these innovative tools excel and where they fall short, the industry can expect more favorable outcomes in the long run.
This shift necessitates a fundamental reevaluation, transitioning from speed-driven processes to a more deliberate and informed approach. Earth AI’s journey underscores the importance of automating data evaluation, validating hypotheses cost-effectively, and maintaining a consistent industry presence to drive success in developing Net Zero infrastructure.
