One year after the proposal, the National AI Research Resource is are coming online — at least in pilot form — as a coalition of US agencies and private partners begin implementing billions in federal funding for publicly accessible tools for aspiring artificial intelligence scientists and engineers.
The NAIRR is the Biden administration’s response to the sudden rise of artificial intelligence on the global tech scene and the concentration of its resources and expertise in a relatively small group of tech giants and privately funded startups. In an effort to democratize technology a bit and keep the US competitive with its rivals abroad, the feds decided to set aside some money to make a variety of resources available to each qualified researcher.
The National Science Foundation, Department of Energy, NASA, NOAA, DARPA and others are all partners in the effort, providing resources (such as datasets and consulting services) and working with applicants in their areas of expertise. And more than two dozen major tech companies are also contributing in some way. The whole thing is budgeted at $800 million a year for the next three years, subject of course to congressional approval.
In a plethora of statements, executives from OpenAI, Anthropic, Nvidia, Meta, Amazon, Microsoft are devoting various resources, expertise, free access, and so on to the NAIRR effort.
The resources to be allocated have not been listed anywhere. Instead, the overall organization will accept applications and proposals, which will be evaluated and resources allocated. Think of it more like a grant process than a free supercomputer.
As NSF’s Katie Antypas said, NAIRR “will provide the research community with access to the computing data, models, software, and training resources necessary to advance the AI ecosystem. The NAIRR pilot is really necessary because the resources needed to even start and participate in the ecosystem have become increasingly concentrated and out of reach for the many, many communities that are really needed to develop a healthy and responsible AI ecosystem. And so the pilot is the first step in bridging that gap.”
He gave three examples: a researcher looking at large AI models who needs large-scale computing resources and has no way to access them; a teacher who wants to let kids do AI-related tasks (like training custom models), but needs resources like virtual notebooks and timing. and someone looking at climate and weather forecasting, who can access NASA and NOAA datasets and combine them with hosted models.
For the two-year pilot period, there will be four areas of focus:
- NAIRR Open, the most general category, including “access to various AI resources”, possibly for research and projects that do not fit into the narrow categories that follow.
- NAIRR Secure focused on AI applications that need privacy and security, such as medicine and critical infrastructure. This section is led by NIH and Energy, unsurprisingly.
- NAIRR software focuses more on tools, platforms, services and interoperability.
- NAIRR Classroom is about outreach, education and training.
It may surprise some that there is no external military research category (considering the presence of DARPA and DOD in the partner agencies), but remember that this is a civilian research effort led by the Executive Services. Presumably, any military research will be done in silos, and the military agencies are there to coordinate and assign or provide the appropriate resources in this case.
The idea is that if someone has a worthwhile idea on how to apply or advance AI in any field, there should be a field-specific and pending review. But it won’t be like a public library, where you walk into your local AI center and someone builds you an H100 in half an hour. (That said, I wouldn’t be surprised if there was some sort of library-centric viewer.)
You’ll be able to peruse the list of resources on the NAIRR Pilot page starting today, and while there aren’t any hard numbers yet, project leaders said it’s likely only 25-50 proposals will be accepted in this initial pilot period, with hundreds more on the way to open in the spring when more systems come online.
