Reflecting on the conversations at URTeC a fortnight ago (June 10, 2025), it's clear that the integration of AI into the energy sector presents a complex set of challenges. At the heart of the matter: improving the quality of AI-generated outputs and enhancing user trust. Key hurdles are:

1) The need for deep domain-specific knowledge.

2) Restricted access to critical data (often paywalled or proprietary).

3) A strong preference in the field for physics-based models (no black-box AI systems).

Take Geothermal energy as an example.

These challenges reveal an urgent opportunity: applying AI to solve tangible geoscience and engineering problems—like designing tougher, more economical drill bits that can operate at extreme temperatures exceeding 375 °C.

A New Computational Paradigm: Getting Physical (Quantitative)

Move beyond LLM [1] from probabilities to precision [2]. In this light, the case for Large Quantitative Models (LQMs) [3] is not just timely—it’s essential.

Further reading & background:

[1] Geocentric LLM for Energy Transition, https://www.aapg.org/care...aft/684/groupid/979, June 3, 2025

[2] LLM are not Large Numeric Models, https://www.aapg.org/care...aft/683/groupid/979, May 9, 2025

[3] SandboxAQ on LQM: https://www.sandboxaq.com...-quantitative-models