Workshop 17

Deep dive into GenAI and Large Language Models (LLMs) for subsurface applications Friday June 14th | Room 14

Convenors

  • George Ghon (Vår Energi / Capgemini)

  • Cédric M. John (Environments Research Institute (DERI))

  • Lukas Mosser (Aker BP)

Description

This workshop will cover an introduction to Large Language Models (LLMs) and Generative AI and evaluate their respective application space for subsurface solutions. In the morning, we will have a series of introductory talks covering the basics of Natural Language Processing (NLP), the transformer models architecture, as well as highlighting examples where generative models have delivered promising results on tasks related to geosciences. The sessions are tailored for subsurface domain specialists as well as generalists with a basic understanding of machine learning and data driven applications. The second part of the workshop will introduce the toolkit available to train large language models on custom datasets and provide a hands-on lab session to show a path toward building web applications. Toward the end of the day, we will host a panel discussion to sum up the learnings, discuss potential pitfalls and ethical dilemmas, and provide an outlook that balances opportunities and risks in modern AI applications.

Sub-Topics that will be covered in the workshop:

Basics of Natural Language Processing (NLP)

Introduction to Generative AI and the architecture of Large Language Models (LLMs)

The integration of language and computer vision tasks

Toolkit to build a GenAI application for an E&P organization.

Hands on lab exercise

Ethical considerations and opportunities vs. risks panel discussion

WS17Pic_LLMs_2024

Participants Profile

Machine Learning practitioners with an interest to apply Large Language Models; Subsurface professionals with an interest in NLP applications, LLMs, and GenAI in subsurface domains; Managers in energy companies who want to understand more about this disruptive new technology.

Workshop Programme

TimeActivity
.PART I: Theory and use cases
08:30Keynote Lecture: Introduction to the theory and applications of generative AI, with a particular focus around recent advancements in NLP and language models such as ChatGPT. - Bjarte Johansen, Equinor
09:15Q&A
9:30Leveraging generative AI for geological document digitization and information retrieval – Song Hou, CGG
10:00Coffee Break
10:20GenerativeAI for Image Data. - Lukas Mosser, AkerBP
10:50Chatting with your data using a RAG framework – case studies. - Gerrit Toxopeus, Equinor and Eirik Haughom, Microsoft
11:20Q&A
11:30Lunch
.PART II: Lab, practical considerations, and introduction to infrastructure requirements
12:15Introduction to guardrails: GenAI guardrails with the NeMo framework. - Oleg Ovcharenko, NVIDIA
13:00Lab & Practical, facilitated by Microsoft
- Practical Session / Lab – Matt Hall, Equinor
.Coffee break
-Practical Session / Lab – Matt Hall, Equinor
.PART III: Discussion
15:00Panel discussion, moderated by George Ghon and Lukas Mosser. Participants: - Peder Aursand, AkerBP - Therese Rannem, Equinor - Song Hou, CGG - Tina Koziol, WDEA - Gareth O’Brien, Microsoft - Nils Nilsen, Slb -
16:00End of the Workshop