NSF launches AI resource pilot to spur US innovation

Ryzhi/Getty Images

Nine other federal agencies and several private sector entities have signed on to support the program.

The National Science Foundation launched the National Artificial Intelligence Research Resource pilot on Wednesday, marking the federal government’s first step in working to democratize widespread access to key components of artificial intelligence technologies. 

Announced on Wednesday, the pilot program aims to promote the Biden administration’s goal of establishing the U.S. as a leader in AI innovation by making federal resources — including advanced computing, datasets, training models, software assistants and user support — open and publicly accessible. 

Immediate goals for the NAIRR pilot are to cultivate an AI-ready workforce and bridge socioeconomic gaps to provide quality AI training and education materials to all corners of the U.S. 

“The NAIRR pilot is really needed because the resources needed to even begin participating in the ecosystem have become increasingly concentrated and inaccessible to many, many communities that are really essential for developing a healthy and responsible AI ecosystem,” Katie Antypas, director of the NSF’s Office of Advanced Cyberinfrastructure, said during a press call ahead of the launch. “And so the pilot is the first step to bridging this gap and will provide access to the research and education community across our country.”

The NAIRR pilot is intended to provide historical datasets to train AI models and computing resources to test the validity of a larger model. Researchers across the country will be able to access these tools to ensure more disadvantaged populations can still learn how to use AI and machine learning systems. 

Four categories define the NAIRR pilot’s focus areas: NAIRR Open, which specializes in general AI resource access; NAIRR Secure, which focuses on AI research for privacy-preserving technologies; NAIRR Software, which helps investigate interoperable uses of AI tools for pilot resources; and NAIRR Classroom, which provides educational initiatives and outreach resources. 

“The pilot is really the first step in unlocking the potential of our research community to advance AI for the public good,” Antypas said. 

Industry partners, including Anthropic, Amazon Web Services, IBM, Meta, Intel, NVIDIA, OpenAI and Microsoft and others, will provide model access, educational resources for experimentation, researcher collaboration, technical training for proprietary software and workshop opportunities.

Antypas confirmed that these companies are not receiving payment for their participation. Some, namely NVIDIA and Microsoft, have pledged $30 million and $20 million respectively to support the pilot program. Stakeholders including government officials, academics and private sector firms collaborated on the pilot's design.

“I think the variety of entities that have come to the table — nonprofits, the private sector, philanthropy — really speaks to this shared urgency to develop this national platform and accelerate AI innovation for our country,” she said. 

The pilot’s format will feature a “diverse variety of architectures” to house these resources. Antypas said that the pilot is meant to grow into a platform united by common software stacks that can support diverse engagement.

“There is not going to be one single entity that is going to be building the NAIRR,” she said. “We're going to need the best ideas from the community in order to really go through this community design process.”

NAIRR’s Community engagement is also meant to foster greater trustworthiness in both mature and newer AI systems. 

Tess deBlanc-Knowles, special assistant to the director for artificial intelligence at NSF, said that researchers can play a “critical role” in developing NAIRR past the pilot.

“I think also in the context of broader federal efforts, the work that is going to be supported through the NAIRR pilot is going to help inform some of these other efforts, such as those being run through [the National Institute of Standards and Technology] or the AI Safety Institute as they move forward to kind of formalize some of these benchmarks around how do we test, how do we verify that these models are trustworthy,” deBlanc-Knowles said. 

Nine federal agencies will join NSF as partnering entities: the Department of Energy, the Department of Veterans Affairs, NASA, the National Institutes of Health, NIST, the National Oceanic and Atmospheric Administration, the Defense Advanced Research Projects Agency, the U.S. Patent and Trademark Office and the Department of Defense. 

These agencies will work together in close coordination alongside other federal efforts that could benefit or inform NAIRR’s work, deBlanc-Knowles said. 

Researchers will be able to apply for access to the NAIRR portal on Wednesday. The pilot program is slated to run for two years. Antypas said that in the pilot’s first launch, officials anticipate supporting 25 to 50 research projects. More projects will come online as additional resources from partnering entities are made available. 

In terms of the application process, researchers will need to first request access to NAIRR tools. They will be vetted based on their responsiveness to the open opportunity call, and a matching process will determine the outcome of each request. 

The NAIRR pilot’s launch is a result of President Joe Biden’s October 2023 executive order on AI. Sethuraman Panchanathan, the NSF director, said that NAIRR is meant to inspire and motivate innovation and talent across the U.S. with quality resources. 

“We need resources to advance AI that is open to all so that every community across our nation may reap the benefits of AI,” Panchanathan said. “Therefore, a National AI Research Resource simply put, has the potential to change the trajectory of our country's approach to AI. It will lead the way for a healthy, trustworthy U.S. AI ecosystem.”