CIV: National AI Research Institutes (Title LII, Sec. 5201)
The Director of the NSF is permitted to establish a network of research institutes that are focused on cross-cutting challenges for AI systems, like trustworthiness or foundational science, or alternatively that are focused on a particular economic or social sector such as health care, education, manufacturing. These institutes are to include a component addressing the ethical and safety implications of the relevant application of AI to that sector and are to be funded for a renewable period of five years.
This seems to be the single clearest funding opportunity related to safety work as we tend to conceptualize it.
DOD: Acquisition of ethically and responsibly developed artificial intelligence technology (Title II, Sec. 235)
Within 180 days of passage, the Secretary of Defense is to assess whether the Department has the ability, requisite resourcing, and sufficient expertise to ensure that any AI technology acquired by the Department is ethically and responsibly developed. In addition to determining how the Department can most effectively implement ethical AI standards in acquisition processes and supply chains, the Secretary is to provide a briefing of the assessment’s results to Congress within 30 days of its completion.
CIV: Industries of the Future Act of 2020 (Title XCIV, Sec. 9412)
The Director of the Office of Science Technology Policy (OSTP) is directed to submit a report to Congress on research and development investments, infrastructure, and workforce development investments of the Federal Government that enable continued U.S. leadership in industries of the future. The provision will also establish an “Industries of the Future Coordination Council,” composed of chairpersons of artificial intelligence, advanced manufacturing, and quantum information science, from the National Science and Technology Council.
CIV: Department of Energy Artificial Intelligence Research Program (Title LV, Sec. 5501)
The Secretary of Energy is directed to focus on AI systems that can improve large-scale simulations and the analysis of existing large-scale datasets from science and engineering experiments—including energy simulations and control systems that enhance automated, intelligent decision-making capabilities. To support this program, the secretary is expected to make high-performance computing infrastructure available at national laboratories and establish new capabilities necessary to manage the data and computational resources required of AI systems.
CIV: Artificial Intelligence Research and Education (Title LIV, Sec. 5401)
Orders the Director of the NSF to fund AI research and education activities to include competitive awards, grants, or prize competitions for institutions of higher education or nonprofit organizations that specifically will contribute to the development of trustworthy AI and other scientific discovery and societal challenges.
In addition to funding research grants for K-12 and postsecondary education, the NSF will establish faculty recruitment fellowships and faculty technology ethics fellowships for eligible institutions of higher education to recruit and retain tenure-track or tenured faculty in AI and related fields, social and behavioral sciences, ethics, law, and related fields to develop new research projects and partnerships in technology ethics. The Director is further empowered to establish national AI centers of excellence to develop and disseminate educational curricula related to AI, AI-related certifications for 2-year programs, and interdisciplinary science and engineering research for employment-based adult learning and career retraining programs.
I flag this one as relevant to funding for organizations like MIRI and new positions for research and advocacy to anyone interested in making it their career. These positions are likely to only be open to people with traditional qualifications, so consolation prize for those who fear they wasted years studying the wrong thing.
CIV: National Oceanic and Atmospheric Administration Artificial Intelligence Center (Title LIII, Sec. 5303)
The administrator of the National Oceanic and Atmospheric Administration (NOAA) is directed to establish a Center for AI to coordinate AI research, establish data standards for agency-wide AI applications, facilitate partnerships between NOAA and other organizations (including institutions of higher education) for research, personnel exchange, and workforce development with respect to AI systems, and preparing NOAA data for use in AI applications.
CIV: Stakeholder outreach (Title LIII, Sec. 5302)
The Director of NIST is tasked with soliciting input on AI regulation from academic researchers, private sector experts, public sector laboratories and other stakeholders as well as outside experts like social scientists, lawyers, and technology ethicists. NIST is also directed to ensure that it provides the opportunity for public comment on these guidelines and best practices.
I flag this one as a way for the community to have influence over the direction of the conversation in government. As with other government interaction, the probability of success is very small but the payoff potentially very large; for those of us with concerns that the military or a large corporation might develop AGI first, this is the kind of opening for which we need to look. Standards do not provide control, but they do provide a Schelling point and can bias the entire field in a good or bad direction as a consequence.
CIV: National Institute of Standards and Technology Activities (Title LIII, Sec. 5301)
The Commerce Department’s National Institute of Standards and Technology (NIST) is to expand its mission to include advancing collaborative frameworks, standards, guidelines for AI, supporting the development of a risk-mitigation framework for AI systems, and supporting the development of technical standards and guidelines to promote trustworthy AI systems.
In addition to developing best practices and voluntary standards for privacy and security in training datasets, computer chips/hardware, and data management techniques, NIST is further tasked with developing a Risk Management Framework that identifies and provides standards for assessing the trustworthiness of AI systems, establishes common definitions for common terms such as explainability, transparency, safety, and privacy, provides case studies of successful framework implementation, and aligns with international standards no later than two years after the passage of the NDAA.
CIV: National AI Research Resource Task Force (Title LI, Sec. 5106)
The Director of NSF, in coordination with OSTP, is to establish a task force to investigate the feasibility of establishing and sustaining a National AI Research Resource and to propose a roadmap and implementation plan detailing how such a resource should be established and sustained. The task force is to be co-chaired by the Director of OSTP and Director of NSF and will be composed of four representatives of the Interagency Committee, four from institutions of higher education, and four from private organizations. Together, the members will provide a plan for ownership for a National Research Cloud and propose a model for governance and oversight to establish strategic direction, make programmatic decisions, and manage the allocation of resources.
The task force will articulate the capabilities required to create and maintain a shared computing infrastructure to facilitate access to computing resources for researchers across the country (including scalability, secured access control, resident data engineering and curation expertise, provision of curated data sets, compute resources, educational tools and services, and a user interface portal). Their assessment of potential barriers to the use of government data sets as part of the National Research Cloud will come with recommended solutions as well as an assessment of privacy, civil rights, and civil liberties requirements associated with the program.
CIV: National Academies Artificial Intelligence Impact Study on Workforce (Title LI, Sec. 5105)
The NSF is directed to contract with the National Academies of Sciences, Engineering, and Medicine’s National Research Council to conduct a study of the current and future impact of AI on the workforce of the United States across sectors. Within two years of the NDAA’s approval, a report including findings and recommendations is to be submitted to the Congress.
CIV: National AI Advisory Committee (Title LI, Sec. 5104)
The companion body to the Interagency Committee is a new external National AI Advisory committee to be established by the Secretary of Commerce in consultation with Director of OSTP, the Attorney General, the Director of National Intelligence, and the Secretaries of Defense, Energy, and State. Members of this Advisory Committee are to be appointed by the Secretary of Commerce and the Committee’s duties are to advise the President and the Initiative Office on the state of United States competitiveness and leadership in AI, the state of the science around AI, issues related to AI and the United States workforce, and opportunities for international cooperation with strategic allies among many other topics. Members will be drawn from academia, industry, and civil society.
The Advisory Committee will further create a subcommittee on AI and law enforcement that will advise the president on bias (including the use of facial recognition by government authorities), data security, adoptability, and legal standards (including those designed to ensure the use of AI systems are consistent with the privacy rights, civil rights and civil liberties, and disability rights issues raised by the use of these technologies.)
CIV: Coordination by Interagency AI Committee (Title LI, Sec. 5103)
The first body within the Initiative Office is an Interagency Committee tasked with providing coordination of Federal AI research and development activities as well as education and workforce training activities across the government. This Interagency Committee is to be co-chaired by the Director of OSTP and a representative from the Department of Commerce, the National Science Foundation (NSF), or the Department of Energy on an annual, rotating basis.
Within 2 years of the passage of the NDAA, the Committee is to develop a strategic plan for artificial intelligence that establishes goals, priorities, and metrics for guiding and evaluating how the agencies carrying out the Initiative will prioritize areas of AI research and development, examines long-term funding for interdisciplinary AI research, and supports research on ethical, legal, environmental, safety, security, bias, and other issues related to AI and society. The Committee will further promote the availability of standardized, representative, and privacy-protected data sets for AI research and development as well as the computing, networking, and data facilities necessary to achieve the National AI Initiative’s goals.
CIV: National AI Initiative Office (Title LI, Sec. 5102)
The overarching infrastructure for the Initiative will be a new National AI Initiative Office to be established by the Director of the White House Office of Science and Technology Policy (OSTP). The office’s mission is to serve as the point of contact for Federal AI activities for Federal departments and agencies, as well as other public and private entities that may be involved in the initiative.
CIV: National AI Initiative (Title LI, Sec. 5101)
To ensure the United States leads the world in the development and use of trustworthy AI systems and prepare the nation’s workforce for the integration of AI across all sectors of the economy, the NDAA calls for the creation of a government wide National AI Initiative that will coordinate AI research and development among civilian agencies, the DOD, and the IC. As detailed below, the initiative will functionally comprise two organizations housed within a new National AI Initiative Office based in the White House
DOD: Guidance and direction on use of direct hiring processes for artificial intelligence professionals and other data science and software development personnel (Title XVII, Sec. 1751)
Within 180 days, the Secretary of Defense is to review Department guidance on use of the direct hiring processes for AI professionals and other data science and software development personnel and issue new guidance to department leadership to ensure that organizational leaders assume greater responsibility for the results of civilian hiring if needed.
DOD: Steering committee on emerging technology (Title II, Sec. 236)
The Secretary of Defense is permitted to establish a steering committee on emerging technology and national security threats to develop a strategy for the organizational change, concept and capability development, and technology investments needed to maintain the technological superiority of the United States military as outlined in the National Defense Strategy. The committee is to provide recommendations to the Secretary of Defense on the implementation of the strategy they developed, the steps that may be taken to address the threats they identify, as well as any changes to the Defense Planning Guidance.
This committee will sunset October 1, 2024; members are to include the Deputy Secretary, the Vice Chairman of the Joint Chiefs of Staff, the Under Secretary for Intelligence and Security, the Under Secretary for Research and Engineering, the Under Secretary for Personnel and Readiness, the Under Secretary for Acquisition and Sustainment, the DOD CIO, and other officials the Secretary deems appropriate.
DOD: Application of artificial intelligence to the defense reform pillar of the National Defense Strategy (Title II, Sec. 234)
Consistent with reform efforts to support the National Defense Strategy, this measure directs the Secretary of Defense to identify a set of at least five use cases for existing AI-enabled systems to support improved management of enterprise acquisition, personnel, audit, or financial management functions. It also directs the Secretary to pilot technology development and prototyping activities that leverage commercially available technologies and systems to demonstrate new AI-enabled capabilities can support the use cases to be identified above.
DOD: Board of advisors for the Joint Artificial Intelligence Center (Title II, Sec. 233)
The NDAA establishes a board of advisors for the JAIC to provide independent strategic advice and technical expertise to the Secretary and the JAIC Director, conduct long-range studies on AI, and assist Pentagon leadership in developing strategic level guidance on AI-related hardware procurement and supply-chain issues. The board is to be appointed by the Secretary and composed of experts from academic or private sector organizations outside DOD.
DOD: Modification of joint artificial intelligence research, development, and transition Activities (Title II, Sec. 232)
This measure amends the organizational structure of the JAIC such that the Director will now report to the Deputy Secretary of Defense instead of the DOD Chief Information Officer (CIO). It places additional emphasis on the acquisition and development of mature AI technologies and gives the Secretary of Defense greater latitude to make decisions about JAIC personnel in research, development and procurement roles.
DOD: Modification of biannual report on the Joint Artificial Intelligence Center (Title II, Sec. 231)
This provision alters the requirements of the Department of Defense’s (DOD) Joint AI Center (JAIC) to include information describing how the JAIC’s efforts contribute to the development of standards in AI in its biannual reports. This is to take into account developments made in collaboration with agencies inside and outside DOD and the Intelligence Community (IC) and further requires the JAIC to report on the status of active-duty military personnel assigned to it.
A bunch of laws about AI were recently passed by Congress in the US.
The National Defense Authorization Act (NDAA) is the law which funds and organizes the Department of Defense of the United States. A new version of the law is passed periodically. For reasons beyond the scope of this post, it is one of a few bills which are considered "must pass" and therefore contain a lot of supplemental legislation, lately including action on COVID-19.
Hidden in the bill are AI provisions, both within the Department of Defense and for civilian agencies. There is a summary of the relevant sections provided by the Institute for Human-centered Artificial Intelligence (HAI) at Stanford. There is an even shorter summary, bullet-point style, at Import AI which is where I got the information.
It is primarily a matter of bureaucratic assignments: different offices and/or agencies are directed to produce reports describing their impacts; new committees are formed; the mandate for existing agencies/committees are expanded to include new facets of AI. The best news is that it includes one provision for ethics, including safety, but I remind everyone that these are in the generic sense of the term which includes uses and capability of machine learning rather than a focus on AGI. So think robots working alongside humans and error margins on deep learning target assessments rather than value alignment.
I will break each item from HAI's summary into a separate comment below, so we can discuss them individually. The ones beginning with CIV are civilian, and with DOD are Department of Defense.