Mathematical Foundations of Artificial Intelligence | 24 569

Opportunity Information: Apply for 24 569

The National Science Foundation (NSF) grant opportunity titled Mathematical Foundations of Artificial Intelligence (Funding Opportunity Number 24-569) supports research that digs into the core mathematical and theoretical questions behind modern machine learning and AI. NSF is framing this program around a basic reality: recent AI successes in areas like protein folding, natural language processing, drug discovery, recommender systems, and materials engineering have moved incredibly fast, but the field still lacks a solid, unified explanation for why these systems work as well as they do, when they fail, and what their real limits are. The solicitation argues that simply scaling up compute and running more experiments is not likely to close the most important foundational gaps. Instead, NSF is explicitly calling for deeper mathematical understanding to clarify both the capabilities and limitations of AI, reduce the risk of unintended consequences, and help ensure AI progress remains sustainable and beneficial for society.

The program is jointly sponsored across multiple NSF directorates: Mathematical and Physical Sciences (MPS), Computer and Information Science and Engineering (CISE), Engineering (ENG), and Social, Behavioral and Economic Sciences (SBE). That structure signals that NSF wants projects that genuinely connect disciplines rather than staying inside a single research silo. Proposed collaborations are expected to bring together teams that may include mathematicians, statisticians, computer scientists, engineers, and social and behavioral scientists, with the shared aim of building rigorous foundations for AI. In practice, this means research that is not only technically deep, but also attentive to how AI systems behave in real settings and how they can be evaluated and trusted.

NSF emphasizes research directions that tackle the hardest underlying questions about AI methods and their emerging properties. The solicitation highlights a range of contemporary paradigms, including foundation models, generative models, deep learning, statistical learning, and federated learning, while also leaving room for other evolving approaches. The main research goals include: (1) developing a fundamental mathematical understanding of what drives performance, generalization, robustness, and failure modes in current and next-generation AI systems; (2) creating mathematically grounded principles for designing and analyzing new AI methods, rather than relying on ad hoc heuristics; (3) developing rigorous ways to characterize and validate machine learning algorithms and the predictions they produce; and (4) enabling AI systems that are provably reliable and general-purpose in a way that can translate across domains. Another explicit goal is to broaden and strengthen interdisciplinary collaboration, including partnerships that span different types of institutions and research communities.

A unifying theme across the opportunity is the push toward principled AI: approaches that are explainable and interpretable where appropriate, and that support sustainable, socially responsible, and trustworthy deployment. While the solicitation is fundamentally about mathematics and theory, NSF is clearly positioning this work as essential infrastructure for the next phase of AI, where society will increasingly demand reliability, accountability, and clearer guarantees around behavior and risk. The expectation is that successful projects will deliver creative theoretical frameworks rooted in mathematics and statistics that can explain observed AI phenomena, predict when methods will succeed or fail, and guide the construction of better algorithms.

Eligibility is limited to certain U.S.-based organizations. Proposals may be submitted by accredited U.S. Institutions of Higher Education (two-year or four-year, including community colleges) acting on behalf of faculty, and by non-profit, non-academic organizations such as independent museums, observatories, research labs, and professional societies, as long as they are located in the U.S. and directly associated with education or research activities. For Principal Investigators and other senior/key personnel, the solicitation requires that, by the proposal submission date, individuals hold either a tenured or tenure-track position, or a primary full-time paid research or teaching appointment at a U.S.-based campus of an eligible submitting organization (with limited exceptions such as family or medical leave as determined by the submitting organization). The text also clarifies that individuals whose primary appointments are at for-profit non-academic organizations or at overseas branch campuses of U.S. institutions are not eligible under these PI appointment rules.

Key administrative details provided include an award ceiling of $1,500,000 and an original closing date of 2024-10-10. The opportunity is categorized as a discretionary grant under Science and Technology and other Research and Development, with CFDA numbers 47.041, 47.049, 47.070, and 47.075. Overall, the program is designed to push beyond incremental performance gains and instead build the theoretical backbone needed to understand, validate, and responsibly advance the AI systems that are increasingly shaping science, engineering, and everyday life.

• The National Science Foundation in the science and technology and other research and development sector is offering a public funding opportunity titled "Mathematical Foundations of Artificial Intelligence" and is now available to receive applicants.
• Interested and eligible applicants and submit their applications by referencing the CFDA number(s): 47.041, 47.049, 47.070, 47.075.
• This funding opportunity was created on 2024-05-02.
• Applicants must submit their applications by 2024-10-10. (Agency may still review applications by suitable applicants for the remaining/unused allocated funding in 2026.)
• Each selected applicant is eligible to receive up to $1,500,000.00 in funding.
• Eligible applicants include: Others.

Apply for 24 569

[Watch] Creating a grant proposal using the step-by-step wizard inside the applicant portal: