PBS-SEPM 2026 January Luncheon - David Thul

"What AI Does & Doesn't Know About O&G and How You Can Use It to Your Advantage" - David Thul, Geolumina
 
Abstract
When you ask ChatGPT or Gemini or any other LLM a question, it isn't actually responding to your question. It isn't "thinking" it isn't "reasoning", it's doing statistics. It is actually responding to the question "What does an answer to this {INPUT} sound like statistically". For things it has lots of statistical representations of (i.e. computer code or anything on Reddit) the answers are nothing short of magic. In highly technical fields that aren't on the internet (like geology), the answers are pretty rough. LLMs don't have (and likely never will) enough representations to give great answers; regardless, you can leverage their power to your advantage by reframing a direct query in different ways. Rather than "What are M and N for the Woflcamp" you can try "How would I determine M and N for the Wolfcamp, walk me through it step by step, write the code I need to analyse my data". In this talk I will explore this approach using the Cretaceous plays of the Rocky Mountains. They are well published, and AI SHOULD KNOW answers to geologic questions about them, but it doesn't. I will show how you can use various approaches to create repeatable appraisal pipelines inside your company using alternative approaches supported by statistics & LLMs.
 
Bio
David is a geologist and geochemist with an entrepreneurial side. He started his career as an oil and gas consultant and has been on an independent path ever since. He began building, marketing and selling oil and gas prospects early, while pursuing myriad business avenues. He ran an exploration research group at the Energy & Geoscience Institute where he did millions of dollars in research leading to patented technology. He has founded private and PE backed companies using digital innovation to change how technical work is done.