NSF aims to drive democratization of AI with its funding

The National Science Foundation is taking a holistic approach to promoting and utilizing trustworthy artificial intelligence systems in its research and education programming, as generative technologies stand to shake up the scientific landscape.

Speaking during a media briefing on Thursday, five officials working in different NSF departments discussed the current programs that are prioritizing the use of trustworthy systems in industries where AI is slated to make a new impact.

A partnership with the National Institute for Standards and Technology is one standout among such programs. Michael Littman, the director for NSF’s Division of Information and Intelligence Systems, said that the program, called the Institute for Trustworthy AI in Law and Society — or TRAILS — works to inform the public on how AI systems have been analyzed by the government for implementation.

“This government partnership with academia will contribute to the national discussion of standards and benchmarks by providing the needed underpinnings in foundational research and the participation of the broader AI research community and the establishment of societally beneficial practices,” Littman said. 

Researchers in the TRAILS program will specifically help develop metrics to gauge trustworthiness in AI systems, analyze the government’s role in promoting trust in AI technologies and develop incentives for more inherent, trustworthy designs features in AI softwares.

In NSF’s educational programming, the goal is to ensure AI technologies act as supplemental tools to facilitate learning rather than replacements for teachers or tasks.

Amy Baylor, the lead program director for the Research in Emerging Technologies for Teaching and Learning program at NSF, said that her office funds research to promote AI literacy and AI to augment learning. The former enables students from all backgrounds to understand how AI systems work and how they can be leveraged.

The latter helps education professionals grasp the benefits of utilizing AI systems in the classroom and other learning environments. To facilitate this, Baylor said the NSF is funding another institute focused on this area.

“The results so far and the institutes that we've funded, you know, have given us new opportunities for students to learn,” she said. 

Funding for these institutes totaled $100 million. Baylor said that equity is a “guiding principle” for AI education, and NSF’s programming hopes to further democratize AI technologies — something Littman and other NSF officials said was a crucial step to ensuring responsible innovation and equitable access in the AI and machine learning field.

“Democratizing AI implies giving power to the people in the AI context, so [that means] ensuring that the majority of the population has access and opportunity to use AI as a tool,” said Abby Ilumoka, the program director for Engineering Education at NSF. 

Ilumoka said that democratizing AI technologies will hinge on ensuring a majority of U.S. citizens have access to AI research and infrastructure.

“Folks need to be educated to a certain level in AI as a technology and how to use it in whatever field they're in,” she said. Additionally, as part of the work the 25 NSF-funded AI research institutes conduct, Ilumoka said that the data they produce will be publicly available to further independent research for AI algorithm training. 

Littman added that the NSF’s goal of making AI technologies and research accessible to the public is to form a strong national AI research resource to support innovation in the field. 

“A lot of the work going on in AI these days is machine learning-based, and a lot of those are extremely computationally intensive, and so by providing research resources for the researchers, that really makes it possible for people over the entire country to participate in this enterprise,” he said. 

Diversity and inclusion are other major components in the NSF’s bid to democratize AI access. 

“NSF is committed to expanding opportunities in STEM to people of all racial, ethnic, geographic and socioeconomic backgrounds,” Ilumoka said. 

Another more niche demographic NSF is looking to incorporate in AI research endeavors is the high performance computing community. Littman said that high performance computing is intrinsic to AI operations to help create an expanded set of computational opportunities and resources the AI research community can use to innovate. Such powerful computing capabilities are usually only available for private sector-affiliated researchers who work within proprietary computing infrastructures.

“We want to make sure that that's available to academics throughout the country,” Littman said.