Details of the Abstract
| Title of paper | Formulating a GPT-4 assisted exploration framework |
| List of authors | Daruso, R., Gamble, D., Pattanaik, S., Russell, B., Taylor, J., Heinson, G. |
| Affiliation(s) | The University of Adelaide |
| Summary | The exponential accumulation of mineral exploration data presents an opportunity for the integration of large language models (LLMs) into the mineral exploration framework, enhancing cost and time efficiency in all stages of data assembly, analysis, and distribution. Focusing on gold prospectivity within the underexplored regions of the Woomera Prohibited Area in central South Australia, a comprehensive pre-competitive dataset—including the Gawler Phase 2 aeromagnetic and magnetotellurics dataset from the Geological Survey of South Australia—was compiled into a 250 x 250 km data cube and analyzed using GPT-4. Clustering models showed great similarity to current inferred solid geology maps, providing more detailed lithological boundaries. While extrapolated data maps might have low accuracy, they indicate areas where additional data is needed. One of the most beneficial functions of GPT-4 is format conversion, both through its server and through generation of Python code to convert formats it cannot directly work with. For instance, it successfully converted geophysical data into DICOM, a medical imaging format. The continuous development of GPT-4 and related OpenAI products suggests promising integration into future mineral exploration efforts. |
| Session Keyword | 9.0 EM induction education and outreach (poster session only) |
| File upload |
9.0_formulating_a_gpt-4_assis_daruso.pdf
|