Recently introduced legislation could help the United States maintain its leadership in artificial intelligence (AI). Earlier this month, Reps. Eddie Bernice Johnson (D-TX) and Frank Lucas (R-OK), the chairwoman and ranking member of the U.S. House Committee on Science, Space, and Technology, respectively, co-sponsored H.R. 6216, “The National Artificial Intelligence Initiative Act,” legislation which would provide nearly $6.5 billion over the next five years to increase funding for AI research and development (R&D), education, and standards development. To date, this bill is the most ambitious attempt by Congress to advance the development of AI in the United States.
Despite the positive efforts of the White House, the United States’ continued AI leadership is not certain. China has rapidly increased its ability to innovate, and its government may be outspending the United States on AI R&D. For example, estimates of China’s government non-defense AI R&D funding for 2018 range from $1.7 to $5.7 billion, whereas the United States planned to spend $1 billion on non-defense AI R&D for fiscal year (FY) 2020. In addition, the United States is heavily reliant on importing AI talent. For example, the majority of students in AI-related graduate programs at U.S. universities are international students, and the number of U.S. graduate students in AI-related degree programs has not increased since 1990.
As such, the U.S. government must continue to take action. The National Artificial Intelligence Initiative Act includes several positive measures. First, the bill provides significant funding for AI research and education. For example, it provides the U.S. Department of Energy $1.2 billion between FY 2021-2025 to carry out an AI R&D program. Furthermore, the bill would allocate $4.8 billion to the National Science Foundation (NSF) over the next five years for NSF to fund AI research and education activities. The increased funding will help NSF fund all the projects it considers highly competitive, which it has struggled to do in the past.
Second, the bill requires NSF to create and coordinate a network of AI research institutes that focus on a specific economic sector, social sector, or cross-cutting AI challenge. This requirement could expand upon NSF’s current National Artificial Intelligence Research Institutes (NAIRI) program and help ensure that AI research advancements are widespread.
Third, the bill charges NSF with providing grants for education programs, including for AI fellowships, traineeships, and K-12, undergraduate, and community college curriculum development. Importantly, higher education institutions can use the traineeship grants to pay the tuition and fees of students pursuing AI-related graduate degrees. These efforts will enhance the supply of AI talent in the United States. Increasing the number of postdoctoral researchers at AI universities could also increase the supply of AI lecturers.
Fourth, the bill provides the National Institute of Standards and Technology $391 million between FY 2021-2025 for several measures, including creating performance benchmarks for AI systems, a framework to assess the trustworthiness of AI systems, and data sharing best practices. These programs would provide AI developers useful guidelines for designing AI systems, allow for the comparison of AI systems, and inform future legislation and regulatory actions. Moreover, the standardization and creation of datasets will spur innovation—data scientists and firms often cite issues related to data quality as a significant obstacle to successful AI projects.
Fifth, the bill would help the U.S. government better adapt to a changing AI landscape in several ways. For example, the bill would establish an advisory committee made up of members from academia, industry, and national labs that would evaluate the initiative and U.S. competitiveness. The bill would also require NSF to fund a study that addresses how AI will impact the workforce and current workforce needs. Finally, the bill would require the Government Accountability Office to evaluate civilian computational resources the government supports and project the nation’s future needs.
These provisions expand the United States’ capacity to lead in AI and should be supported. In addition, Congress should evaluate additional ways to advance the development of AI in the United States, including AI education. For example, Congress should consider establishing an AI equivalent to the “CyberCorps: Scholarship for Service” program, which provides scholarship money to cybersecurity students that agree to work for the government for a period equal to the length of their scholarship. Creating this type of program would provide a steady pipeline of AI experts to work in government. In addition, Congress should direct NSF to provide grants for part-time “academic tours of service” fellowships for both government and private workers to teach university-level AI courses and competitive awards to professors teaching AI-related courses, conditional on their remaining in academia for five years.
AI is already providing the United States significant benefits, ranging from rapid detection of wildfires to aiding in the fight against COVID-19 to helping find victims of human trafficking. However, the United States will not receive all of the benefits AI can provide unless it continues to make advancements, and the National Artificial Intelligence Initiative Act would help the United States make the necessary improvements.
Image: David Maiolo