The final report from the government’s National AI Research Resource recommends a new multibillion-dollar research organization to improve the field’s capabilities and accessibility for U.S. scientists. The paper presents “a roadmap and implementation plan for a national cyber infrastructure aimed at overcoming the access gap, reaping the benefits of increased brainpower and more diverse perspectives and experiences.”
The NAIRR Report (PDF) has been a long time coming: since the creation of the task force in 2020 under the leadership of the White House Office of Science and Technology Policy. They have been busy, producing numerous smaller reports and a comprehensive “blueprint for an AI Letter of Rights” in that time. which you can read here.
The full thing is many pages long, but the summary gets to the point:
To realize the positive and transformative potential of AI, it is imperative to harness all of America’s ingenuity to advance the field in a way that addresses societal challenges, works for all Americans, and upholds our democratic values.
Yet progress on the current frontiers of AI is often linked to access to large amounts of computing power and data. Such access today is too often limited to those in well-equipped organizations. This large and growing resource gap can limit and adversely affect our AI research ecosystem.
A widely accessible cyber infrastructure for AI research that brings together computational resources, data, testbeds, algorithms, software, services, networks, and expertise, as described in this report, would help transform the AI research and development (R&D) landscape in the United States. democratize for the benefit of all.
To that end, they propose a new independent research organization (under the control of the appropriate agencies and departments) that would “make available through an integrated portal a federated mix of computing and data resources, testbeds, software and testing tools, and user support services.” .”
They recommend this not include setting up its own data centers, at least initially, which would be costly and potentially difficult to scale, but rather working with partners who can allocate existing resources to the project. (These could be private companies or National Labs, one assumes.)
The “executing entity” – that is, the research organization – should also be “proactive in addressing privacy, civil rights and civil liberties issues by integrating appropriate technical controls, policies and governance mechanisms from the outset.”
Congress would need to fund the new organization at a cost (NAIRR suggests) of approximately $750 million every two years over a six-year period to build its capabilities, totaling $2.25 billion. It would then need about $60 million to $70 million a year for its ongoing operations. This does not include associated grants or such programs, which would likely go through the National Science Foundation or other existing programs.
Much more detail on how it would all be run is in the full report, but of course the details will ultimately have to wait for money and other resources to be allocated.
“We view the NAIRR as a foundational investment that would bolster federal government efforts to cultivate AI innovation and promote trusted AI,” the team writes in the introduction to the final report. “Research, experimentation and innovation are integral to our progress as a nation, and it is imperative that we engage people from every zip code and every background to deliver on America’s unique promise of opportunity and ensure our leadership on the global stage.”