Readout of the Fifth National Artificial Intelligence Research Resource (NAIRR) Task Force Meeting
The NAIRR Task Force is developing a plan for spurring innovation and expanding diversity in the artificial intelligence research community by expanding access to computational infrastructure. The Task Force deliberated on design elements related to security, privacy, and civil rights; approaches for combining a wide variety of research resources; the role of public-private partnerships in building and sustaining the resource; and indicators of success. The Task Force also engaged with a group of researchers from the prospective NAIRR user community
The National Artificial Intelligence (AI) Research Resource (NAIRR) Task Force convened its fifth virtual, public meeting on February 16, continuing its efforts launched in June 2021 to develop a vision and implementation plan for a national cyberinfrastructure that would connect American researchers from all backgrounds and regions to the computational, data, and testing resources that fuel AI research and innovation. By creating an equitable infrastructure for cutting-edge AI that builds on-ramps for participation for a wide range of researchers and communities, a NAIRR could build AI capacity across the nation and support its responsible development, thereby driving discovery and innovation in areas ranging from healthcare to city planning to education and beyond, and ensuring long-term U.S. competitiveness.
The Task Force meeting opened with three consecutive discussions that considered elements central to the design of the NAIRR, including security and access controls; privacy, civil rights, and civil liberties requirements; and technical integration of the diverse resources that would comprise the NAIRR, including high-performance and cloud computing, data, testing infrastructure, and educational tools. The Task Force members discussed the imperative to design security requirements into the NAIRR from the outset, ensuring usability is maintained as a central element of any approach and embracing a tiered access model. Building on ideas from external experts who spoke in the Task Force’s public meeting in December, the Task Force reached consensus that privacy, civil rights, and civil liberties must be built into the NAIRR’s governance through a transparent framework for vetting research and data sets, regular reporting processes, and user training requirements. In addition, the Task Force concluded that the NAIRR has the opportunity to drive transformation in cyberinfrastructure more generally, towards a more data-centric integration of data, testing, and “edge” resources that are located closer to data sources.
A panel of researchers representing different facets of the envisioned NAIRR user base provided perspectives on designing the NAIRR to best meet the needs of the anticipated broad and diverse user community, describing the potential for a NAIRR to impact their research and the competitiveness of early AI startups and small businesses. Panelists emphasized the importance of inclusivity in approaching the NAIRR user base and the return on investments in providing testing environments and accessible opportunities to experiment with novel AI capabilities, including the associated reproducibility of those experiments. The Task Force also considered options for public-private partnerships related to the specific resources and structure being envisioned for the NAIRR, identifying the role of partnerships in promoting sustainability of the infrastructure. The Task Force fielded questions from the public throughout their discussions, addressing issues related to budgeting NAIRR capacity, technical architecture, pricing models for NAIRR use, and incentives for public-private partnerships.
Co-Chairs Dr. Lynne Parker, Director of the National AI Initiative Office in the White House Office of Science and Technology Policy (OSTP), and Dr. Manish Parashar, Office Director for the Office of Advanced Cyberinfrastructure at the National Science Foundation (NSF), concluded the day’s deliberations by leading discussions related to developing indicators of success for the NAIRR and structuring the Task Force’s interim report. The Task Force agreed that the indicators of success should be grounded in progress toward the intended outcomes of the NAIRR – spurring innovation, expanding diversity, building human capacity, advancing responsible and trustworthy approaches to AI, and sustaining and enhancing U.S. competitiveness in AI. Task Force members also emphasized that a NAIRR management entity should capture and regularly evaluate information regarding NAIRR investments, usage, outputs, and impacts to inform responsive, data-driven management of NAIRR resources.
The Task Force will hold its sixth public meeting in April 2022. Details on how to participate in the Task Force meetings will be available on AI.gov/nairrtf, along with materials from this and prior meetings.
External Speakers at the February 16 Meeting:
Tom Dietterich, Oregon State University
Susanta Ghosh, Michigan Technological University
Kinnis Gosha, Morehouse College
Gail Rosen, Drexel University
Rima Seiilova-Olson, Kintsugi
Carlos Theran-Suarez, Florida A&M University
This was the NAIRR Task Force’s first public meeting of 2022. It met four times over the course of 2021, convening virtually in July, August, October, and December 2021. More information can be found here, here, and here.
###
