Teachers and school administrators leverage a number of tools on a daily basis. Keeping up to date with the details of the tools and how and what can be done with them is a daunting learning task. The goal of our Help Desk prototype is to test the hypothesis that we can alleviate this problem with a chat interface using retrieval-augmented large-language models.
This tool ingests a corpus of process, product support, troubleshooting, and standard operating procedures to provide support to teachers.
Behind the scenes, our Large Language Model (LLM) uses retrieval augmentation to get the most recent information from the school-specific knowledge repository and presents it in a concise, actionable manner. It also uses the dialog history to continuously refine the content to provide the most relevant information in the next turn of the conversation.
In our prototype demonstration video above, we ask the Help Desk a few questions about how to use the Merlyn AI assistant. Almost instantly, Merlyn retrieves the correct information and presents it in a conversational voice. We learn that Merlyn responds to voice commands and can perform several other functions. We’re also provided with a link to learn more.
Marrying a search function with a domain-specific LLM is an important advancement when it comes to helping teachers with their everyday tools.
In this case, we used the Merlyn AI assistant as an example. But Help Desk can be used for any number of tools that educators leverage.
Stay tuned as this prototype develops and for more last-mile solutions for education from Merlyn Mind’s AI Labs team.