The Department of Energy (DOE) wants feedback on its plan to launch a public-private consortium to curate scientific data from its national laboratories to create advanced “self-improving” artificial intelligence (AI) models.
In a request for information (RFI) posted to SAM.gov, the DOE requested feedback by Jan. 14 on how it can best mobilize national labs, establish governance for shared data and infrastructure, prepare AI-ready datasets at scale, and integrate general-purpose and domain-specific AI models.
The department also requested guidance on priority scientific disciplines, necessary data modalities, use-case evaluations, cloud-based delivery of models to researchers, and the balance between centralized and federated data repositories.
All of that information will eventually lead to what the DOE called “self-improving AI models for science and engineering.” Those models will be trained on both data from the national labs and from partners.
Models will also be made available to “a system of United States government, academic, and private-sector programs and infrastructure” using cloud technologies, according to the RFI.
The consortium is a direct result of the White House’s AI Action Plan, released this July, which directed DOE, among other federal agencies, to “invest in automated cloud-enabled labs for a range of scientific fields.”
Policy recommendations outlined under the plan include using AI to make fundamental scientific advancements while incentivizing the “highest quality AI-ready scientific datasets.”
“This historic mobilization of DOE, the National Laboratories, and private partners will serve as a force multiplier in executing America’s AI Action Plan to achieve global dominance in AI, and to advance scientific discovery, energy, and national security,” the DOE said in its RFI.