A federally funded AI research cloud is moving forward, and startups should be able to join the party.
A task force set up to design The National AI Research Resource, or NAIRR, a repository of data, tools and computing power needed to develop machine learning and other AI systems, published a preliminary report today outlining plans and expectations for the service.
Following months of public meetings, the task force, which is overseen by the White House Office of Science and Technology Policy and the National Science Foundation, said the resource should be operated by an independent, non-governmental entity.
And, despite expectations by some that the NAIRR would be available solely for academic research, task force leaders reaffirmed interest in opening it to startups.
The NAIRR is intended primarily for academia, said Lynne Parker, director of the National AI Initiative Office within the White House Office of Science and Technology Policy and co-chair of the task force. However, during a press conference today, she said, “Certainly the task force is open to enabling startups that have, for instance, received federal grants.” She specifically mentioned Small Business Innovation Research and Small Business Technology Transfer grant programs.
Exactly who will be able to access the NAIRR has been in question throughout the initial development phase of the resource. Some supporters of the NAIRR, including the Stanford Institute for Human-Centered Artificial Intelligence (HAI), have pushed against the idea of opening it to private corporations, noting that it should focus on the needs of academic and nonprofit researchers.
In their report, the task force also supported creation of “an independent, non-governmental entity with dedicated, expert staff” to manage the NAIRR’s infrastructure, resource allocation, user support and security. They called for federal agencies to make new or existing infrastructure resources — including some from private sector providers — available to the NAIRR for AI research and development, including data, compute and testbeds.
Several corporations including the big three cloud providers — Amazon's AWS, Google Cloud and Microsoft's Azure — have all submitted proposals for the project.
“Importantly, NAIRR computational resources should span the full range of possible offerings, including commercial cloud, high-performance and high-throughput computing, on-premise (at academic and/or government sites) resources, ‘edge’ computing resources and devices, and novel computing approaches and platforms,” stated the task force report.
Task force co-chair Manish Parashar, office director for the Office of Advanced Cyberinfrastructure at the National Science Foundation, pointed to existing NSF facilities that could serve as models for managing NAIRR’s shared infrastructure. “We can look at those over the next few months to see how can we learn from those, and see how effective they will be for a resource such as what's envisioned as the NAIRR," he said.
NAIRR planners have emphasized the need for the resource to be accessible to a diverse and inclusive group of people, and to incorporate responsible and trustworthy AI principles in data resources and AI developed using them. “The task force recommends that the NAIRR establish an ethics review process to vet all resources included in the system, and the research performed with it. NAIRR users will be required to complete regular updated ethics training modules before being granted access to the network,” said Parashar.
The task force is seeking public comments on the report and a public listening session will be held on June 23.