Artificial Intelligence Initiative Act (S 1558, 116th Congress)
The Artificial Intelligence Initiative Act (S 1558, 116th Congress) calls for the formation of an “Artificial Intelligence Research and Development Initiative” that would require the Executive Branch in general and three agencies in particular—the National Institute on Standards and Technology (NIST), the National Science Foundation (NSF), and the Department of Energy (DOE)—to facilitate research and development in artificial intelligence (AI) and promote collaboration across sectors. The term “artificial intelligence” is defined by the bill as a system that performs tasks under varying and unpredictable circumstances without human oversight, or learns from experience and improves performance when exposed to data sets. The bill provides the framework for this Initiative and delineates specific requirements for various federal agencies, as detailed below.
The bill tasks two organizations within the Executive Office of the President, the Networking and Information Technology Research and Development Program (NITRD) and the Office of Science and Technology Policy (OSTP), with carrying out the provisions of the bill. NITRD is responsible for establishing objectives, priorities, and metrics for AI research and development, investing in research and development including workforce development, and helping to coordinate AI research and development efforts across governmental and non-governmental entities. OSTP is responsible for establishing the following entities:
- National Artificial Intelligence Coordination Office to serve as a point of contact for all entities involved with the Initiative and conduct outreach.
- This office, through collaboration with the Department of Commerce, Bureau of Labor Statistics, Census Bureau, educational institutions, minority-serving institutions, nonprofit organizations, and industry, is itself responsible for contracting with a federally funded research and development center to study the AI workforce. This study would propose actions for increasing the size and quality of the AI workforce, help with the formation of best practices for developing and sustaining the AI workforce, and contribute to an annual Department of Labor report on job creation in the AI field.
- Interagency Committee on Artificial Intelligence, with members from over ten federal agencies, to make recommendations for federal agencies regarding AI research and education, establish objectives and priorities for the Initiative, evaluate opportunities for international collaboration with regards to AI, and provide a strategic plan, updated every five years, to guide the activities of the Initiative.
- National Artificial Intelligence Advisory Committee, with members from industry, research, and government, as appointed by NSF and OSTP, to discuss and report trends in AI; potential societal, ethical, legal, environmental, or workplace concerns of AI; the international presence of the US in this field; and the success of the Initiative as a whole.
In support of the Initiative, NIST is responsible for collaborating with other entities to develop measures and standards for advancing commercial development of AI. They will also conduct outreach through stakeholder meetings to gain insight on topics including but not limited to cybersecurity, algorithm accountability, algorithm explainability, algorithm trustworthiness, and establishing a common lexicon for describing AI. The bill authorizes the appropriation of $40 million in funding for each fiscal year from 2020 to 2024 to implement these provisions.
The NSF is tasked with creating a program to increase participation in artificial intelligence and support research and education in the field. Topics of focus are like those required of NIST but also include minimization of bias in training data and output, consumer data privacy, and societal and ethical implications of AI use. NSF is also responsible for encouraging collaboration across fields, providing graduate traineeships, and providing grants to create “Multidisciplinary Centers for Artificial Intelligence Research and Education.” Institutes of higher education, relevant nonprofit organizations, and state and local governments are eligible to apply for these grants, and may do so by 1) submitting a plan to work collaboratively across institutions and across disciplines, integrate ethics of development, and support workforce development, and 2) providing an anticipated description of the long-term impact of the center. Each center is granted $20 million for each fiscal year from 2020 to 2024.
Lastly, the DOE is tasked with creating a program to coordinate its research related to AI; make computing hardware and software accessible to researchers in industry, government, nonprofits, and academia; and maintain a website with information on training programs and other resources related to AI. The DOE is also responsible for providing access to computing resources and for awarding grants to eligible entities to create research centers to advance research and education. The bill authorizes the appropriation of $300 million in funding for each fiscal year from 2020 to 2024 to implement all of these provisions.
This bill is one of many relating to artificial intelligence that were introduced in the 116th Congress by members of the bipartisan Senate Artificial Intelligence Caucus. Other AI-focused bills introduced in the 116th Congress are listed below in the “Related Policies” section.
Previously in the 115th Congress, the House contained the bipartisan Congressional Artificial Intelligence Caucus which also introduced a large number of AI-related bills. Of these, only the National Security Commission on Artificial Intelligence Act has passed. This bill established the Independent National Security Commission on Artificial Intelligence to address the economic and international implications of artificial intelligence and to consider methods of sharing training data while protecting privacy and security. It was signed into law as part of the National Defense Authorization Act for Fiscal Year 2019 (Public Law 115-232) on August 13, 2018.
This legislative push for an increased focus on artificial intelligence research is paralleled by steps taken by the Executive Branch:
- In May 2018, the White House held a summit on AI for American Industry.
- On February 11, 2019, the President signed Executive Order 13859, which established the American AI Initiative.
- On September 19, 2019, the White House Held another summit, this one on Artificial Intelligence in Government.
- The White House has also created a public website that has details of other steps taken to further investigation into artificial intelligence, as well as examples of artificial intelligence work in various agencies.
The international community has gotten involved as well. For instance, the Organization for Economic Co-operation and Development created a set of principles in their May 2019 Recommendation on Artificial Intelligence that was incorporated into the G20’s AI Principles adopted in June 2019.
Artificial Intelligence, as defined by S 1558, is an artificial system that:
- Performs tasks under varying and unpredictable circumstances without human oversight, or learns from experience and improves performance when exposed to data sets;
- Solves tasks requiring human-like perception, cognition, planning, learning, communication, or action;
- Is designed to think or act like a human, including cognitive architectures and neural networks;
- Is designed to approximate a cognitive task; or
- Is designed to act rationally, including an intelligent software agent or embodied robot that achieves goals using any of the above skills.
Machine learning is defined in the bill as “a subfield of artificial intelligence that is characterized by giving computers the autonomous ability to progressively optimize performance of a specific task based on data without being explicitly programmed.”
Despite these definitions provided by the bill, artificial intelligence has been defined in a variety of ways over time and across fields.
- There is some bias implicit in machine learning and this bias is inappropriate (Section 101, but also throughout): There have been multiple, well-publicized examples of the detrimental effects that bias in machine learning can cause. Bias can be introduced in many aspects of the machine learning process, and in many cases cannot be easily addressed.
- Machine learning will have a non-negligible effect on the workforce (throughout the bill, but especially Section 101): Most researchers and industry leaders agree about this assumption, although it is still unclear what form this effect will take. There is a large body of research dedicated to speculation on this topic.
- There are valid concerns about safety, security, and reliability of artificial intelligence for science and technology (Section 101 and throughout): Experts have given much consideration to potential technological and ethical implications surrounding the use of AI.
- Hardware development and the development of new types of computing resources, including cloud-based computing, will be necessary to ensure adequate access to research and education infrastructure (Section 301): Artificial intelligence and machine learning require large amounts of resources for computation and data storage, which could be prohibitively expensive. Nonetheless, there are various kinds of technology under development to solve this problem.
- Creation of an initiative to coordinate research efforts will provide better or faster outcomes than allowing all departments to pursue objectives independently (throughout): The federal government has a history of creating science and technology initiatives (such as the National Quantum Initiative, BRAIN Initiative, and National Nanotechnology Initiative) to unify priorities and coordinate research efforts. However, many of these initiatives are still relatively nascent and there have been few public assessments about their effectiveness.
Scientific Controversies / Uncertainties
Although the bill does not go into detail about the scientific principles underlying AI, it does touch on many of the broader concerns surrounding AI. Proponents of artificial intelligence argue that it will change the way we interact with our world through technologies like self-driving cars or immersive home devices like smart speakers. They also point out that AI will increase our ability to make new discoveries and insights into everything from business to medicine to law enforcement. However, others argue that artificial intelligence will replace human jobs and could have a negative effect on the workplace. Other critics point out the privacy, security, and bias risks to artificial intelligence, including the harms that could come from using artificial intelligence in sensitive situations.
Another topic of contention is explainability—while certain types of machine learning are easily interpretable, others are a “black box,” meaning it can be difficult if not impossible to determine what factors a machine considers when making its decision. Furthermore, interpretability often comes with a cost to performance. While some experts say that this loss of interpretability is acceptable for some applications, especially since these “black box” models are often more accurate, others argue that interpretability is an incredibly important factor for any algorithm.
The bill does not attempt to take a side on any of these controversies, but many of its proposals touch on these issues: minimization of bias, explainability, reliability, privacy, and security are among the many focus areas proposed by the bill.
Endorsements & Opposition
- Adrian Bowles (analyst, Aragon Research), article, June 21, 2019: “ It would make much more sense for the Interagency Committee on Artificial Intelligence (to be established by the Director of the Office of Science and Technology Policy under this Bill) to make allocation recommendations periodically based on demonstrated needs and progress of entities bidding for funds.”
- Senator Martin Heinrich (D-NM; bill sponsor), press release, May 21, 2019: “If we defer AI development to other nations, important ethical, safety, and privacy principles will be at risk, which not only harms the United States, but the international community as a whole. When a new AI advancement is made – the [Artificial Intelligence Initiative Act] will have already put into place policy, people, and potential for this new market to truly take transportation, health care, manufacturing, and national security to the next stage of opportunity.”
- Information Technology and Innovation Foundation, press release, May 21, 2019: “The future of US competitiveness in AI will depend on how effectively policymakers support key inputs for innovation, particularly R&D [research and development]. Thus far, efforts to support AI R&D have been lackluster. By providing clear, concrete steps for the federal government to take to accelerate and strengthen AI R&D, [this bill] would secure the United States’s status as the global leader in R&D.”
- Internet Association, statement, May 21, 2019: “The internet industry supports the Artificial Intelligence Initiative Act.… [The bill] will foster additional AI development through a coordinated strategy that aims to increase data sharing between academia, the public, and private sector, as well as expand the education pipeline to prepare America’s workforce for the jobs of the future.”
- Senator Rob Portman (R-OH; bill co-sponsor), press release, May 21, 2019: “By coordinating and synchronizing our country’s research and development efforts, this bill ensures not just that the United States remains an AI leader, but that it does so by developing AI technology that prioritizes American values.”
- Senator Brian Schatz (D-HI; bill co-sponsor), press release, May 21, 2019: “Our bill will give researchers and innovators the resources to study and further develop AI technology so that we can use it in smart and effective ways.”
- Swami Sivasubramanian (Vice President of Amazon Machine Learning), press release, May 21, 2019: “Greater investment in AI research and development is essential to ensuring the technology benefits all of society and advances US leadership in AI innovation. The Artificial Intelligence Initiative Act is an encouraging step in realizing this vision.”
- Garnett S. Stokes (President, The University of New Mexico), press release, May 21, 2019: “I am particularly enthusiastic about the attention paid in the bill to workforce development and the need to reach out to minorities and underrepresented groups to build a diverse workforce for the future.”
- US Chamber of Commerce, open letter to Congress, July 29, 2019: “[This bill] would be a critical first step to cement American leadership in artificial intelligence and ensure that AI benefits our society as a whole.”
The proposed legislation would provide funding and support for artificial intelligence research, signaling an increased political and governmental interest in AI research, development, and oversight. Senator Heinrich (D-NM), the bill’s sponsor, states that the bill will complement previous executive branch actions on AI by adding resources and requiring further interagency coordination.
The US Chamber of Commerce, among others, believes that this bill demonstrates American interest in pursuing a global advantage in artificial intelligence. The Center for Data Innovation expresses a similar concern that the US may lose the advantage in AI development unless they pass this bill, or something similar, to create funding for research. Furthermore, according to the bill’s sponsors, by becoming and remaining a leader in AI development, the United States can ensure that all new technologies will have “American values”, opinions echoed by the Center for Data Innovation, Amazon, New Mexico State University, and BSA.
The bill could have military applications as well. The New Mexico Institute of Mining and Technology suggests that new discoveries in artificial intelligence could “radically accelerate the cycle of military operations and deeply influence military practices.”
Furthermore, the bill demonstrates a federal interest in considering the effects of artificial intelligence on the workforce. The Internet Association believes that the bill would help to “expand the education pipeline to prepare America’s workforce for the jobs of the future,” and the University of New Mexico claims that the bill will build a diverse workforce because if its focus on minority-serving institutions and underrepresented groups.
Lastly, the bill’s focus on ethical and inclusive development could suggest a commitment to principled development. Concerns about privacy, security, and bias in data are mentioned multiple times in the bill. As Senator Heinrich argues: “If we defer AI development to other nations, important ethical, safety, and privacy principles will be at risk, which not only harms the United States, but the international community as a whole.” Furthermore, Senator Heinrich claims that this focus on ethics and innovation will lead to outcompeting China because the work environment in the United States will be more appealing and will draw more people that otherwise would have worked in China, thus furthering the US’s global advantage. Industry and academia alike laud the bill’s consideration of ethics: both Microsoft and New Mexico State University mention the focus on societal implications.