During meetings, we often encounter detailed and specific questions, such as whether a particular well had shows in the Jurassic, why a licence was relinquished, or if there are any downthrown traps near a prospect. Or, what was the porosity in the Sleipner Formation in this or that well? Unfortunately, finding the necessary information is a time-consuming process that involves searching through large in-house data stores or government web pages.
Once the relevant document is located, it must again be searched for the required information. By the time the answer is provided to the meeting attendees, they may have moved on mentally and physically. Wouldn’t it be great if we could find a way make this process more efficient?
Late 2022, a new large language model (LLM) called ChatGPT emerged and forever changed the way we approach text knowledge retrieval and processing. With this amazing technology, we can now ask natural language questions and receive concise and accurate answers for a wide range of topics.
However, there are also some drawbacks to using LLM’s. For instance, if they lack sufficient information about a topic, they may provide incorrect answers – hallucination. Additionally, they are not trained and therefore cannot retrieve information and documents in your in-house or government data stores. Furthermore, it is prohibitively expensive to train these models from scratch. That’s where the geo-bot project comes in.
An AI assistant
The geo-bot project aimed to develop an assistant that overcomes the challenges of text knowledge retrieval and processing by combining the strength of a semantic search engine with the power of a large language model such as ChatGPT.
The Fabriq platform has created the first such geo-bot for the oil gas industry and will soon be available for public testing. This platform has ingested licence relinquishment reports, wikis, final well reports and more into a semantic search engine. This allows terms and concepts to be related to each other across the entire document corpus. ChatGPT then uses the semantically retrieved information to provide the most relevant answer to the user’s query.
Attend the upcoming NCS Exploration Technology Conference in Oslo – 12-14 September, and hear how Jesse Lord is Building Trust with ChatGPT: Navigating Conversations in Exploration Documents & Wikis
With this AI assistant, geologists can easily access knowledge about documents that are in-house or were not part of the training of the large language model. For example, they can ask the bot to summarise the findings of a particular production license. The bot can even create a list of prospects that share similar characteristics, such as being a four-way or being AVO-supported.
A significant time-saver
We recommend that you test out the new assistant and other upcoming AI-driven language tools. The technology is not yet perfect, it may occasionally miss key information, retrieve the wrong document, or even hallucinate. But it is undoubtedly a significant time-saver when it comes to obtaining a quick overview of a topic or document that has been ingested.
