Designing responsible AI for the future

Written by NSF Director Sethuraman Panchanathan

In many ways, artificial intelligence is a wonder of the modern world, impacting everything from how we learn to how we do business to how we build resilience against natural disasters. It is the incredible engine behind everyday innovations such as virtual assistants, navigation apps and personalized streaming services with the potential to spur new capabilities and discoveries that can power practical solutions to critical global challenges from food production to poverty, climate change to cancer.

Many of the AI advancements making news today — both the innovative products and the talent behind them — are made possible by NSF’s long history of investments in foundational and use-inspired AI research and the translation of that research to use-inspired technologies.

However, as AI technologies are transforming the world around us, NSF understands the importance of taking bold, decisive steps to ensure AI’s safe and responsible use.

On Monday, the White House unveiled its sweeping and long-awaited executive order on artificial intelligence. Among other things, this order "establishes new standards for AI safety and security, protects Americans' privacy, advances equity and civil rights, stands up for consumers and workers, promotes innovation and competition (and) advances American leadership around the world.”

As a global leader in AI, NSF stands ready to fully implement the actions outlined in the executive order and its eight guiding principles and priorities, which include developing methods to help ensure secure and responsible AI systems, protecting Americans from AI-enabled fraud and deception, and requiring that developers of powerful AI systems share their safety test results with the U.S. government.

NSF’s role in this critical endeavor is not new. For decades, NSF has been one of the largest federal investors in AI across all 50 states, Washington, D.C., and every U.S. territory, accelerating responsible AI innovation, driving cutting-edge research and preparing the next-generation AI workforce — even through AI winters.

In Fiscal Year 2023, NSF invested over $800 million in AI research and innovation activities. These investments include strategic partnerships with other federal agencies and industry, to include the National Artificial Intelligence Research Institutes, a growing network of research hubs enabling academia, industry and government to accelerate discovery and innovation in AI, inspiring talent and ideas that will lead to new capabilities that improve lives and position the nation in the vanguard of competitiveness and prosperity. Together with our partners, NSF has announced 25 AI Institutes over the last three years, a $500 million investment, with plans to award additional AI institutes in FY 2024 and FY 2025.

This week, NSF announced major investments in two of its AI programs: Expanding AI Innovations through Capacity Building and Partnerships (ExpandAI) and Safe Learning-Enabled Systems (SAFE).

ExpandAI will receive $16.3 million to advance artificial intelligence innovation by strengthening and broadening participation in AI research and education at minority-serving institutions, unleashing untapped talent across our nation and ensuring a more equitable future for the science and technology enterprise.

SAFE is receiving $10.9 million to support research to develop responsible artificial intelligence technologies. This investment not only strengthens NSF's commitment to fostering the safety of AI systems but sends a clear message to the research community that NSF considers safety paramount to the responsible expansion and evolution of AI. SAFE is an NSF partnership with Open Philanthropy and Good Ventures.

Leadership in action

But NSF’s leadership in AI extends far beyond supporting research.

Under the National AI Initiative Act of 2020, NSF was called on to form a task force in conjunction with the White House Office of Science and Technology Policy to create a roadmap for a National AI Research Resource (NAIRR).

NAIRR is intended to be a national cyberinfrastructure to shape the future of AI technology and its role in society. It will expand access to resources, data, educational tools and user support to researchers and students from diverse backgrounds as they pursue foundational, use-inspired and translational AI research.

Society can only benefit from greater brainpower and more diverse perspectives and experiences. And a national cyberinfrastructure offers a unique and critical opportunity to “design in” standards for responsible AI research practices that reinforce democratic values and personal freedoms.

In January, the task force presented its report, “Strengthening and Democratizing the U.S. Artificial Intelligence Ecosystem," to Congress and the president. It recommended that NAIRR have four foundational and measurable goals:

•   Spur innovation

•   Increase diversity of talent

•   Improve capacity

•   Advance responsible AI

Investments in action

NSF’s high-risk, high-reward investments in AI have powered incredible achievements.

Recently, that includes establishing the AI Institutes — 25 of them over the past three years in partnership with other federal agencies and industry. Today, AI Institutes reach almost every state in the U.S., supporting cutting-edge research to address some of society’s greatest challenges.

This year, NSF announced a $140 million investment to establish seven new institutes to advance foundational research that promotes responsible and secure AI systems and technologies, develops novel approaches to cybersecurity, contributes to innovative solutions to climate change, expands our understanding of the brain and leverages AI capabilities to enhance education and public health.

The new institutes will also support developing the country’s diverse AI workforce by launching new opportunities, engaging talent, and inspiring the leaders of tomorrow.

One of those amazing research hubs is the NSF AI Institute for Agent-based Cyber Threat Intelligence and Operation (ACTION) at UC Santa Barbara. ACTION develops revolutionary approaches to accelerate U.S. cyberdefense, in which AI-enabled intelligent security agents cooperate with real people to jointly improve the security of complex computer systems over time.

Another is the NSF AI Institute for Trustworthy AI in Law & Society (TRAILS), which works toward future AI systems enhancing not only human capacity, but human rights and dignity. TRAILS researchers investigate what trust in AI looks like, how to create AI solutions that build that trust, and which policy models are effective in sustaining it. The institute is a partnership among the University of Maryland , The George Washington University and Morgan State University .

Since the 1960s, NSF has advanced foundational AI research breakthroughs and products we use today. And we have only scratched the surface of what is possible. NSF will continue to advance AI technology for decades to come and enable opportunities for innovation coupled with the right safeguards so that AI can reach its amazing promise of making innovation happen anywhere, and opportunities happen everywhere.

In this critical time for AI research, it is imperative that American values are engrained in the AI innovations of the future. The opportunity before us is immense, and I am confident that the U.S. will remain in the vanguard of competitiveness in AI and all its benefits.