Northwestern University Libraries have begun exploring new ways to interact with our collections using the natural language capabilities of OpenAI's GPT family of large language models (LLM). We're building a chat interface that answers queries by pairing an LLM's conversational ability with the digital collections data under our control in a vector database. This allows users to interact with digital collections using natural language while mitigating AI challenges, often referred to as "hallucinations" in the current discourse. Throughout our prototype, we have dedicated effort to exploring ways to increase transparency and safety for our users. We are eager to share our experiences with the Samvera Community, and look forward to hearing about other initiatives currently underway!
