By John Mark Shaver, The Preston County News & Journal
MORGANTOWN, W.Va. — WVU Medicine has begun incorporating artificial intelligence into its day-to-day operations, utilizing it to make data retrieval more efficient while still protecting patient information.
West Virginia University Health System Vice President and Associate CIO Ilo Romero said Large Language Models (LLMs), such as those used by ChatGPT and Meta AI, have a massive amount of data to pull information from. But since they don’t have access to private information or internal company documents, it’s difficult to leverage the power of LLMs to make the workplace more efficient.
However, using a Retrieval-Augmented Generation (RAG), Romero said organizations can plug their documents into LLMs, giving users within the organization the ability to search for things more efficiently with the exact context needed.
“(These documents) could be tip sheets and how-to guides or clinical protocols or any private knowledge you have in your company,” Romero said.
Through a three-step process, RAG converts large swaths of data into numerical vectors that can be more easily retrieved by LLMs. By using this technology in a hospital setting, for example, a hospital employee can search the LLM for a specific protocol or guideline and, since the LLM is pulling from that hospital’s data, it will return a more accurate and specific answer to the employee in a matter of seconds.
Romero said using a RAG limits “AI hallucinations,” instances in which AI produces wrong or misleading information, by supplying the AI with as much relevant information as possible. Additionally, he said, a RAG allows users to employ any LLM they choose as long as their own information is accurate and up-to-date.
