The Software and Systems Division is running a pilot of an on-premises generative AI language model service for use by the NIST community. Such a service would allow members of the community to do research using AI language models without worrying about leaking data outside the organization or exhausting a pre-allocated token budget.
The pilot intends to measure the interest and usefulness of having an on-premises generative AI language model service for the NIST community.
End users will have two methods for interacting with the AI models: (1) a web-based chat interface and (2) RESTful API access.
Throughout the pilot, we plan to track usage of the service. Additionally, we will invite participants to present any interesting findings from using the service and how it helped advance their research goals.
Upon conclusion of the pilot, we plan to use the data collected to determine whether to permanently deploy an on-premises generative AI language model within NIST.