LLM for geoscience - benefits and concerns

Free

You need to be signed in to reserve your place for this event.
» Sign In
» Create Free Account

FEATURED SPEAKERS

Paul Cleverley
» founder
» Infoscience Technologies Ltd

Full Agenda

Friday, March 28, 2025
Webinar
Online

There is a lot of excitement about the use of large language models (LLMs) in geoscience, with the hope that they could answer any geoscience based on all of the geoscience papers ever written, just as Chat GPT can give an answer to just about any question today.

Can they deliver that, or if not what can they deliver? And what are the concerns?

If a LLM is able to 'ingest' every geoscience paper ever written (bypassing copyright or limits to access), it would have enormous knowledge available. But it would not have any model of how geology works, which a geologist would have. The insights it shared would be based on only what other geologists have written, and what geologists have written more about.

If a geoscientist is seeking knowledge about a particular formation in the world, or lithography, that could be found just as well by a simple word search.

When the LLM is creating its own language (rather than indicating where further material on a topic can be found), then it can make errors, as we have seen.

In this webinar, Paul Cleverley presents his perspective on geoscience LLMs. Paul is managing director of Infoscience Technologies based in Oxford, England, a company specialising in Subsurface and Geoscience Natural Language Processing, Artificial Intelligence.

He is Visiting Professor of Information Science & Technology at Robert Gordon University, Advisor Geoscience Natural Language Processing at British Geological Survey, chair of the rtificial Intelligence Ethics Task Group at International Union of Geological Sciences and a former Head of Knowledge Management, Global Geoscience Information Advisor with Shell.

Agenda