AI systems for scientific advancements and national security will be developed by the Department of Energy by utilizing its current supercomputers and labs.
The FASST effort, a $12 billion, five-year endeavor to develop integrated scientific AI systems, has a roadmap released by the Department of Energy (DOE). FASST, an acronym for Frontiers in Artificial Intelligence for Science, Security, and Technology, seeks to leverage DOE’s computational infrastructure to establish a network of AI research clusters. The project’s main goals will be to advance national security applications, energy solutions, and scientific research.
“The world’s most powerful integrated scientific AI systems” would be established under the project, which would concentrate on AI initiatives across four key pillars:
- Data: Creating techniques, platform protocols, and additional instruments to distribute and aggregate AI training datasets effectively and securely.
- Computing Infrastructure and Platforms: Creating and implementing next-generation AI computing infrastructure, including testbeds and high-performance computing systems, as well as computing platforms.
- AI Models and Systems: creating, honing, evaluating, and verifying base models for use in research and national security.
- Applications: Building models to address problems in science, energy, and security using AI foundation models and other technologies.
FASST also addresses the creation of AI tools to “dramatically reduce the time to discovery and extend the nation’s competitive edge in technological innovation,” as well as a large talent procurement project to attract the best AI talent to the United States.
Additionally, the initiative will examine clean energy, enhancing grid resilience and optimizing energy use.
Energy Secretary Jennifer M. Granholm declared, “AI is an innovative technology that can help unleash breakthroughs in energy technologies and enhance our national security. FASST expands on DOE’s historic responsibility to manage the country’s cutting-edge supercomputing and research infrastructure throughout our 17 national labs, thereby establishing a national AI capability and paving the way for future technological innovations.”
The DOE already possesses some of the most potent AI supercomputers in existence, such as Frontier, the most potent supercomputer in the world, located in the Oak Ridge Leadership Computing Facility, and the Perlmutter unit in the US National Energy Research Scientific Computing Center.
The agency said that it would be able to create a “national AI capability” that researchers could use to find new inventions, from battery materials to possible advances in fusion, by utilizing its experience and current infrastructure.
According to the organization, “These capabilities will also be leveraged to promote safety, security, trustworthiness, and privacy and provide insight into properties of AI systems at scale.”
Using AI, the DOE has already made several scientific advances, including the development of a novel battery material in partnership with Microsoft and the discovery of a medication that is currently undergoing clinical testing as a possible cancer treatment.
Under the recently sponsored bipartisan DOE’s AI Act, which aims to leverage existing DOE infrastructure to advance the nation’s AI efforts, the DOE would be given $2.4 billion a year for five years to spend on the FASST project.
Additionally, the bill would mandate that the DOE report to Congress once a year on the status and costs of the FASST initiative.
Along with Sen. Joe Manchin, Sen. Lisa Murkowski introduced the bill. “AI has the potential to help the Department of Energy and our National Labs accomplish enormous challenges in the fields of science and technology, as well as offer capabilities to streamline permitting for large-scale energy and critical mineral projects in Alaska,” Murkowski said.
