NAIRR Task Force Recommends $2.6B for the New AI Resource

Published: February 08, 2023

Federal Market AnalysisArtificial Intelligence/Machine LearningNSFResearch and Development

The innovative infrastructure will provide computational and data resources to an AI user base of researchers, educators, and students.

Formed by the National AI Initiative Act of 2020, the National Artificial Intelligence Research Resource (NAIRR) Task Force released a final report proposing a detailed research infrastructure to bolster the U.S. AI research ecosystem. The task force, led by the White House Office of Science and Technology Policy (OSTP) and National Science Foundation (NSF), is made up of 12 experts across government, academia, and the private sector. The report is the result of an 18-month effort and includes a roadmap and implementation plan to stand up the platform.

In its opening letter to Congress, the task force details the challenges faced by AI researchers in the U.S., including lack of accessibility to computational power and federated data resources, and limited breadth of ideas in AI innovation due to systemic inequalities among researchers leading to bias in AI systems. As a result, the proposed NAIRR platform seeks to accomplish four goals: spur innovation; increase diversity of talent; improve capacity; and advance trustworthy AI.

The shared infrastructure is set to bring together computational resources, data, testbeds, algorithms, software, services, and networks to help “democratize the AI research and development (R&D) landscape in the United States for the benefit of all,” according to the report. Specifically, computational resources include servers, computing clusters, high performance computers (including at least one large-scale machine-learning supercomputer capable of training 1 trillion-parameter models), cloud computing, and access to edge computing resources and testbeds to the AI R&D user base. Moreover, the task force recommends software resources such as that of open source frameworks like TensorFlow, Pytorch and their derivatives. Datasets in the NAIRR should extend beyond federal data, however, the report recommends that NAIRR include and engage in existing federal government data and upcoming resources such as data.gov, SAP, and the National Secure Data Service (NSDS) established by the CHIPS and Science Act of 2022. Overall, the computing, data, and training resources of the NAIRR are expected to be delivered by universities, federally funded research and development centers (FFRDCs), and the private sector.

Additional elements of the NAIRR include accessibility to a range of users (researchers, educators, and students) and a platform used for educational and community building activities. Furthermore, the NAIRR must have system safeguards in compliance with established cybersecurity guidelines.

The report details a three-part governance over the NAIRR: a single federal agency (NSF) to house operations; a steering committee (comprised of several federal agencies) to drive strategic direction, and an independent non-government operation entity to oversee daily operations. Sample responsibilities under each group are as follows:

The task force estimates a budget of $2.6B over a six-year period. Breakdown of the proposed $2.6B budget includes: $2.25B in appropriations to multiple federal agencies to fund the resources made accessible through the NAIRR, including $750M every two years to resource providers brought online; $60-70M per year to the operating entity; and $5M per year for external evaluation of the operating entity and NAIRR performance.

The report establishes a four-phase roadmap to implement the NAIRR, beginning immediately after report publication. The first phase (year 0-1) is when Congress should authorize and appropriate funds to establish the NAIRR, and NSF and NAIIO should coordinate the formation of the Steering Committee and program management office. Phase one should also see the RFP preparation and selection of the OE. The second phase (within year 1) is when the OE should establish activities and oversee creation of the NAIRR portal and interface and build in policy controls. Resource providers should be selected during phase two as well. In the third phase, (year 1-2) the NAIRR will go into operational capacity and the OE should formalize policies and processes. Phase four (year 2 and onwards) activities should transition from build out to establishing steady operations.  

Contractor Implications

Upon an approved budget and initial governance establishment, the NAIRR will potentially require extensive contractor support. In addition to being a source provider for the NAIRR, contractors may find opportunities ranging from portal set-up and software and data integration, to providing training support and continually updating the NAIRR with the latest technologies and capabilities.