NSF Comes to SC24 With Money Map, AI Blueprint

The 2023 Supercomputing Conference (SC24) saw a significant shift towards artificial intelligence (AI) as a primary focus, with much of the discussion and technical program centered on how AI will transform high-performance computing (HPC). A key highlight was a presentation by Tess DeBlanc-Knowles, Special Assistant to the Director for Artificial Intelligence at the National Science Foundation (NSF) and former advisor to the White House Office of Science and Technology Policy (OSTP). DeBlanc-Knowles provided an overview of the NSF's approach to funding new AI initiatives, underscoring the organization’s commitment to fostering a secure, inclusive, and innovative AI ecosystem in the United States. The NSF currently allocates around $750 million annually for AI research and development (R&D), which is distributed across various technical priorities. DeBlanc-Knowles mentioned that over 500 new projects are set to be funded each year, covering a broad spectrum of AI research areas such as combining statistical and logical AI systems, enhancing the cost efficiency and reasoning capabilities of large language models (LLMs), and addressing privacy and security concerns. This funding reflects the NSF's emphasis on foundational research, infrastructure development, workforce training, and the translation of AI technologies into practical applications. One of the flagship programs highlighted by DeBlanc-Knowles is the National AI Research Institutes, which has received over $500 million in funding since its inception in 2019. These 27 institutes are tasked with integrating AI into various sectors, including agriculture, weather forecasting, public health, and education. The institutes collaborate with different federal agencies to align their research with national goals. For instance, the Department of Agriculture and Department of Education are working together on projects related to sustainability and agricultural resilience, while the latter focuses on developing AI-driven educational tools. Another notable initiative is the Regional Innovation Engines program, designed to decentralize technology hubs and promote regionally appropriate AI research. In 2023, NSF funded 10 such engines across 18 states, each eligible for up to $150 million over a decade. Seven of these regions are specifically concentrating on AI-driven techniques. Examples include the Great Lakes Water Engine, which is developing intelligent water resource management systems, and North Dakota’s agricultural engine, which is using AI for crop genetics and climate modeling. Infrastructure development is also a critical component of the NSF’s AI strategy. DeBlanc-Knowles emphasized the growing divide between private-sector researchers, who have access to large-scale computational resources and data, and academic researchers, who often lack such resources. To bridge this gap, the NSF is supporting initiatives like the National AI Research Resource (NAIR) pilot, a consortium of 12 federal agencies and 26 non-governmental partners. NAIR provides academic researchers with access to computational resources, datasets, and pre-trained models, thereby addressing disparities in access and enabling more researchers to engage in cutting-edge AI work. The NAIR pilot serves 20,000 researchers annually and includes the creation of AI-ready testbeds for real-world applications in sectors like urban transportation and power grid management. International collaboration is another growing dimension of the NSF’s AI strategy. DeBlanc-Knowles highlighted partnerships with countries like Australia, Israel, and India, focusing on joint research in areas such as drought resilience and digital twins for urban infrastructure. Supplementary funding has expanded these collaborations to include countries like South Korea, Belgium, and New Zealand, with shared interests in severe weather modeling and eldercare technologies. The recently released Global AI Research Agenda, developed in coordination with the State Department, outlines U.S. priorities for international partnerships in fundamental AI research and risk management. The NSF’s efforts are closely aligned with broader federal strategies, including the Biden administration’s Executive Order on AI. This order, the longest in U.S. history, details over 100 actions to enhance AI capabilities while managing associated risks. Complementary guidance from the Office of Management and Budget and the National Security Memorandum on AI further underscores the commitment to responsible AI adoption in both civilian and defense sectors. In her address, DeBlanc-Knowles stressed the importance of building the infrastructure, talent, and trustworthiness necessary to ensure U.S. leadership in AI. While traditional HPC was not a primary focus of her presentation, the shift towards AI at SC24 and the NSF’s substantial investments in AI R&D indicate a new direction in the field. This shift is likely to have significant implications for labs and universities, which may face challenges in justifying the costs of large GPU clusters and high power consumption for AI research. The presentation at SC24 serves as a clear signal that the NSF and the broader U.S. government are prioritizing AI as a strategic technology, with a strong emphasis on ethical considerations, international collaboration, and infrastructure support to ensure that the benefits of AI are widely accessible and responsibly managed. As the field of HPC continues to evolve, the integration of AI is poised to play a central role in shaping future research and development efforts.