Readout of the Tenth Meeting of the National Artificial Intelligence Research Resource (NAIRR) Task Force

The NAIRR Task Force deliberated on recommendations related to the standup timeline, security requirements, data resources, and administrative home for the envisioned NAIRR

On October 21, members of the National Artificial Intelligence Research Resource (NAIRR) Task Force met virtually for their tenth public meeting. As the penultimate convening of the Task Force before the submission of its final report to the President and Congress in December 2022, the meeting focused on refining details of the implementation plan related to the standup timeline, security requirements, data resources, and administrative home for the envisioned NAIRR.

Co-chaired by the Office of Science and Technology Policy and the National Science Foundation, the Task Force has been working since June 2021 to develop a vision and implementation plan for the NAIRR – a national cyberinfrastructure that would democratize access to resources and tools that fuel Artificial Intelligence (AI) research and development (R&D). Expansion of access would aim to broaden the range of researchers involved in AI, grow and diversify approaches to and applications of AI, and open opportunities to advance AI R&D across scientific fields and disciplines, including in critical areas such as AI auditing, testing and evaluation, bias mitigation, and security.

The first topic taken up by the Task Force related to the timeline for a phased establishment of the NAIRR. The Task Force discussed integrating an option of a pilot approach into the recommended timeline, with the aim to leverage existing resources, software stacks, and service providers to launch a pilot that would expedite making resources (albeit on a limited scale) available to the AI R&D community while the full NAIRR is coming online. Task Force members agreed on the merits of pursuing such a pilot option and highlighted the priority of supporting user-facing staff through the pilot, to ensure it is effectively serving the AI research community.

The Task Force also revisited the concepts it developed around setting a security framework for sensitive data that could be made available through the NAIRR, considering whether to adopt a tiered access model or a configuration that would include a fully open NAIRR coupled with a single and distinct secure layer. Task Force members talked through the importance of embracing an approach that would include identity management and compliance auditing, and enable the transfer of secure data from external sources and across various NAIRR resources.

The approach to connecting data repositories through the NAIRR was raised to clarify the envisioned role for the NAIRR in developing incentives for data sharing and mechanisms for search and discovery. The Task Force discussed issues of data integrity and how to measure both the value and impact of data included in the NAIRR.

The Task Force’s final discussion focused on the design of the NAIRR governance structure to strike a balance between the practicalities of administering the operations of the NAIRR and its goal to be a truly national resource. The Task Force discussed the concept of an agency-level administrative home for the NAIRR paired with an interagency steering committee that would provide strategic direction and oversight. Task Force members highlighted the critical characteristics of an ideal NAIRR administrative home, such as mission alignment, capacity and capability for effective program management, and a focus on diversity and equity.

The meeting closed with questions and comments from public attendees, which focused on details of the possible pilot approach discussed in the meeting, the role for small businesses and startups in development of the NAIRR user portal, and prioritization of a vendor-agnostic platform.

The Task Force will hold its eleventh public meeting on December 7, 2022. Details on how to participate in the meeting will soon be available at AI.gov/nairrtf, along with materials from this and prior meetings.

###