Scholarlink

<p>It is important to know whether and how artificial intelligence (AI) will improve future exploration for undiscovered economic mineral deposits. Early neural network applications showed promise but typically were trained with local data and did not provide generalized predictions. There are simply too few mineral deposits in local regions necessary to build robust networks. Over the past 40+ years, three-part quantitative mineral resource assessments have been developed and demonstrated how using deposit types documented with consistent world models of geological settings and grade and tonnage models led to success at generalized predictions of amounts of undiscovered resources. The advent of powerful deep learning in the 2010s and 2020s led to many studies in mineral exploration based on prospectivity mapping, but they did not demonstrate the ability for generalization because they were not trained with consistent world data. New mineral deposit discoveries have not been made with prospectivity mapping. Experiences in unbiased three-part mineral assessments and successful mineral exploration demonstrate that models of mineral deposit types including standardized definitions of deposits and associated spatial rules are necessary for training AI if the results are to be generalized and useful in predicting undiscovered economic deposits. The development of tools proposed here including modern deposit type classifications, fault classification guides, extrapolated geology under cover, and spatial distance estimators should aid experts by reducing the time required and improving the accuracy of mineral deposit discovery. It is not clear whether AI will replace experienced economic geologists, but training deep learning with global deposit information is necessary, and the development of tools like those proposed here could significantly aid experts in improving efficiency and developing deep networks for discovery of economic deposits including under cover. Eventually, new AI technology will have a significant impact on these tasks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep Learning Requirements to Find Undiscovered Mineral Deposits

  • 期刊论文

  • 发表日期：

    2026年3月31日

查看全文
  • Donald A. Singer

摘要

It is important to know whether and how artificial intelligence (AI) will improve future exploration for undiscovered economic mineral deposits. Early neural network applications showed promise but typically were trained with local data and did not provide generalized predictions. There are simply too few mineral deposits in local regions necessary to build robust networks. Over the past 40+ years, three-part quantitative mineral resource assessments have been developed and demonstrated how using deposit types documented with consistent world models of geological settings and grade and tonnage models led to success at generalized predictions of amounts of undiscovered resources. The advent of powerful deep learning in the 2010s and 2020s led to many studies in mineral exploration based on prospectivity mapping, but they did not demonstrate the ability for generalization because they were not trained with consistent world data. New mineral deposit discoveries have not been made with prospectivity mapping. Experiences in unbiased three-part mineral assessments and successful mineral exploration demonstrate that models of mineral deposit types including standardized definitions of deposits and associated spatial rules are necessary for training AI if the results are to be generalized and useful in predicting undiscovered economic deposits. The development of tools proposed here including modern deposit type classifications, fault classification guides, extrapolated geology under cover, and spatial distance estimators should aid experts by reducing the time required and improving the accuracy of mineral deposit discovery. It is not clear whether AI will replace experienced economic geologists, but training deep learning with global deposit information is necessary, and the development of tools like those proposed here could significantly aid experts in improving efficiency and developing deep networks for discovery of economic deposits including under cover. Eventually, new AI technology will have a significant impact on these tasks.