A Submission by the AI Now Institute and Data & Society Research Institute

The White House Office of Science and Technology Policy (OSTP) and the National Science Foundation (NSF) recently launched the National Artificial Intelligence Research Resource (NAIRR) with the stated goal of “democratizing” access to AI research. This “National Research Cloud”, which will draw hundreds of millions in federal funding, purports to create a shared research and data infrastructure that provides AI researchers with access to computational power, high-quality government datasets, and other resources.

In response to OSTP and NSF’s request for information, the AI Now Institute and Data & Society Research Institute submitted detailed comments on their implementation plans.

While the NAIRR attempts to grapple with issues around access and the resources required to produce AI and similar technologies, we argue that the project falls far short in its efforts to ‘democratize’ AI. We encourage the Task Force to fundamentally reconsider whether the investment in shared computing and data infrastructure is consistent with the Biden Administration’s explicit commitment to challenging the concentrated power of the tech industry.

We emphasize that the only plausible short- to mid-term path to implementing the infrastructure required for the NAIRR would be to license it from the same large tech companies responsible for concentrating tech power. While the NAIRR claims to ‘democratize’ access to AI resources as a way to contend with tech companies’ concentrated power, in reality it would work to expand and entrench the power and control these companies.

As the Task Force pushes forward on its implementation plans and required reports to Congress, we encourage them to consider alternative policy strategies to reduce the power and control of the handful of companies currently dominating AI and AI research. The Task Force should ensure that the determinations about whether AI is developed and deployed are subject to more democratic deliberation.

Below, we summarize our five key arguments from our submitted comments:

  1. The NAIRR will entrench, rather than challenge, corporate control over the AI field, contrary to the Biden Administration’s bold stance against the power of large tech companies in society.
  2. The NAIRR Task Force must reckon with mounting evidence of the harmful impacts of large-scale AI systems, including discriminatory consequences for marginalized groups and long-term climate impact.
  3. The NAIRR Task Force must expand their understanding of what disciplines constitute “AI research” and redirect NSF resources and programs toward constructing mechanisms for meaningful democratic control of AI and related technologies.
  4. The NAIRR and related proposals raise serious ethical and data privacy challenges, particularly with the use of government data. Given the lack of demonstrated best practices and global policy precedents for data privacy, the NAIRR Task Force should recommend pausing the NAIRR until these challenges are resolved.
  5. The NAIRR, as presently conceived, bolsters misleading and dangerous “tech cold war” narratives, which reflect the self-interest of Big Tech and the defense contracting industry, without being backed by robust evidence.

In our expanded comment, we express concerns about the stated aims for the NAIRR, and offer alternative policy strategies to support research into AI and to expand access to data, resources, and educational opportunities. Our comments urge the NAIRR Task Force to recommend that Congress consider alternative spending priorities and deeper engagement on the harms unleashed by uncritical investment in AI.