Conference: a workshop to identify translational opportunities in AI efficiency and scaling

Grant Summary

The National Science Foundation (NSF) awarded $375,213 to the University Industry Demonstration Partnership, Inc. (UIDP) under the NSF Technology, Innovation, and Partnerships (TIP) program (CFDA 47.084) on August 15, 2025, for a facilitated workshop addressing translational opportunities in artificial intelligence (AI) and machine learning (ML) efficiency and scaling. The primary deliverable is a structured workshop that convenes AI/ML researchers from academic institutions with model builders from industry to identify collaboration opportunities and develop concrete collaborative project formulations. The workshop aims to generate at least five collaborative project concepts focused on incorporating academic research and research results into proprietary frontier models, with explicit emphasis on improving AI/ML performance and energy efficiency.

The grant activity addresses a critical gap in U.S. technological leadership by bridging researchers with innovative ideas and enterprises developing large-scale proprietary AI models. By facilitating cross-sector collaboration between academia and industry, the workshop is intended to accelerate AI/ML progress, enhance model effectiveness, improve energy efficiency, and ultimately advance U.S. economic vitality and national defense capabilities. The performance period extends through July 31, 2026, with the workshop to be conducted in Columbia, South Carolina.

Abstract

Leading ?frontier models? In artificial intelligence and machine learning (AI/ML) have progressed primarily by exponentially scaling costly, energy-intensive datacenter infrastructure. Most research institutions, in particular universities, do not have the resources to engage effectively in frontier model research because of the cost of those models. Meanwhile, the most advanced models tend to be proprietary rather than open sourced, further hobbling the ability of most researchers to influence and improve the most expensive models. As continued scaling becomes prohibitively costly, and because energy is soon expected to become a limiter of progress, U.S. Technological leadership may depend on the degree to which researchers can effectively experiment and incorporate new techniques into proprietary frontier models. To bridge the GAP between researchers with ideas and enterprises that are building large-scale proprietary models, this facilitated workshop will bring together AI/ML researchers and model builders to identify collaboration opportunities and to ideate concrete collaborative project concepts for incorporating research and research results into practice (?research translation?). If these projects are adopted and executed by the model builder participant organizations, the impact will arguably be more rapid AI/ML progress, more effective AI/ML, and more energy-efficient models, extending progress to benefit economic vitality and U.S. National defense. The goals of this facilitated workshop activity are to generate at least five collaborative project formulations for achieving significant improvements in AI/ML performance and energy efficiency. These project formulations will be developed by teams comprising representatives of both AI/ML model producers and leading AI/ML researchers. The responsibility of the researchers is to come armed with promising research ideas, especially those improving energy-efficient scaling, while the responsibility of model producers is to help align those ideas with their organizations? Needs. Facilitators will help guide ideation, collaboration and project formulation. A program committee will help recruit the workshop participants (approximately 35-40) such that there is a balanced representation of model-building companies, researchers and other stakeholders. The program committee will also provide impartial advice to project teams in support of the workshop facilitators. Through team formation, model builders will gain a greater appreciation of relevant research results and opportunities. Researchers will correspondingly gain a greater appreciation of deployment challenges and opportunities. Ideally, the results will be compelling to the model-building organizations and will lead to implementation. Overall, new collaborations resulting from the workshop are expected to accelerate impactful research and research translation that is not feasible in today?s costly, proprietary AI/ML deployments. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are not planned for this award.