Obama, Trump, and the Regulation of Artificial Intelligence

Near the end of his second term, President Obama announced a series of workshops and government working groups tasked with “Preparing for the Future of Artificial Intelligence.” Then, just weeks before the 2016 presidential general election, the Obama administration published two reports including one titled “The National Artificial Intelligence Research and Development Plan.” In it, Obama laid out seven strategies for AI-related R&D, including making long-term investments in AI research to enable the United States to remain a world leader in AI, developing effective methods for human-AI interaction, and ensuring the safety, security, and trustworthiness of AI systems. The Obama AI plan also included strategies for developing shared and high-quality public datasets and environments for AI training and testing, creating standards and benchmarks for evaluating AI technologies, and understanding the national AI research workforce needs. His plan also recognized the need for collaboration among researchers to address the ethical, legal, and societal implications of AI, topics that still resonate today.

Two years after Obama’s AI announcement, the Trump administration in May 2018 convened an Artificial Intelligence Summit at the White House and then published an “Artificial Intelligence for the American People” fact sheet highlighting President Trump’s AI priorities. The fact sheet highlights the President’s goal of funding fundamental AI R&D, including in the areas of computing infrastructure, machine learning, and autonomous systems. Trump’s AI priorities also include a focus on developing workforce training in AI, seeking a strategic military advantage in AI, and leveraging AI technology to improve efficiency in delivering government services. The Trump fact sheet makes no mention of Obama’s AI plan.

Despite some general overlap and commonality between Obama’s and Trump’s AI goals and strategies, such as funding for AI, workforce training, and maintaining the United States’ global leadership in AI, one difference stands out in stark contrast: regulating AI technology. While Obama’s AI strategy did not expressly call for regulating AI, it nonetheless recognized a need for setting regulatory policy for AI-enabled products. To that end, Obama recommended drawing on appropriate technical expertise at the senior level of government and recruiting the necessary AI technical talent as necessary to ensure that there are sufficient technical seats at the table in regulatory policy discussions.

Trump, on the other hand, has rolled back regulations across the board in a number of different governmental areas and, in the case of AI, has stated that he would seek to “remove regulatory barriers” to AI innovation to foster new American industries and deployment of AI-powered technologies. With the Trump administration’s express concerns about China’s plan to dominate high tech, including AI, by 2025, as well as Congressional efforts at targeted AI legislation slowed in various committees, any substantive federal action toward regulating AI appears to be a long way off. That should be good news to many in the US tech industry who have long resisted efforts to regulate AI technologies and the AI industry.

Congress Takes Aim at the FUTURE of Artificial Intelligence

As the calendar turns over to 2018, artificial intelligence system developers will need to keep an eye on first of its kind legislation being considered in Congress. The “Fundamentally Understanding The Usability and Realistic Evolution of Artificial Intelligence Act of 2017,” or FUTURE of AI Act, is Congress’s first major step toward comprehensive regulation of the AI tech sector.

Introduced on December 22, 2017, companion bills S.2217 and H.R.4625 touch on a host of AI issues, their stated purposes mirroring concerns raised by many about possible problems facing society as AI technologies becomes ubiquitous. The bills propose to establish a federal advisory committee charged with reporting to the Secretary of Commerce on many of today’s hot button, industry-disrupting AI issues.


Leaving the definition of “artificial intelligence” open for later modification, both bills take a broad brush at defining, inclusively, what an AI system is, what artificial general intelligence (AGI) means, and what are “narrow” AI systems, which presumably would each be treated differently under future laws and implementing regulations.

Under both measures, AI is generally defined as “artificial systems that perform tasks under varying and unpredictable circumstances, without significant human oversight, or that can learn from their experience and improve their performance,” and encompass systems that “solve tasks requiring human-like perception, cognition, planning, learning, communication, or physical action.” According to the bills’ sponsors, the more “human-like the system within the context of its tasks, the more it can be said to use artificial intelligence.”

While those definitions and descriptions include plenty of ambiguity, characteristic of early legislative efforts, the bills also provide several clarifying examples: AI involves technologies that think like humans, such as cognitive architectures and neural networks; those that act like humans, such as systems that can pass the Turing test or other comparable test via natural language processing, knowledge representation, automated reasoning, and learning; those using sets of techniques, including machine learning, that seek to approximate some cognitive task; and AI technologies that act rationally, such as intelligent software agents and embodied robots that achieve goals via perception, planning, reasoning, learning, communicating, decision making, and acting.

The bills describe AGI as “a notional future AI system exhibiting apparently intelligent behavior at least as advanced as a person across the range of cognitive, emotional, and social behaviors,” which is generally consistent with how many others view the concept of an AGI system.

So-called narrow AI is viewed as an AI system that addresses specific application areas such as playing strategic games, language translation, self-driving vehicles, and image recognition. Plenty of other AI technologies today employ what the sponsors define as narrow AI.

The FUTURE of AI Committee

Both the House and Senate versions would establish a FUTURE of AI advisory committee made up of government and private-sector members tasked with evaluating and reporting on AI issues.

The bills emphasize that the committee should consider accountability and legal rights issues, including identifying where responsibility lies for violations of laws by an AI system, and assessing the compatibility of international regulations involving privacy rights of individuals who are or will be affected by technological innovation relating to AI. The committee will evaluate whether advancements in AI technologies have or will outpace the legal and regulatory regimes implemented to protect consumers, and how existing laws, including those concerning data access and privacy (as discussed here), should be modernized to enable the potential of AI.

The committee will study workforce impacts, including whether and how networked, automated, AI applications and robotic devices will displace or create jobs and how any job-related gains from AI can be maximized. The committee will also evaluate the role ethical issues should take in AI development, including whether and how to incorporate ethical standards in the development and implementation of AI, as suggested by groups such as IEEE’s Global Initiative on Ethics of Autonomous and Intelligent Systems.

The committee will consider issues of machine learning bias through core cultural and societal norms, including how bias can be identified and eliminated in the development of AI and in the algorithms that support AI technologies. The committee will focus on evaluating the selection and processing of data used to train AI, diversity in the development of AI, the ways and places the systems are deployed and the potential harmful outcomes, and how ongoing dialogues and consultations with multi-stakeholder groups can maximize the potential of AI and further development of AI technologies that can benefit everyone inclusively.

The FUTURE of AI committee will also consider issues of competitiveness of the United States, such as how to create a climate for public and private sector investment and innovation in AI, and the possible benefits and effects that the development of AI may have on the economy, workforce, and competitiveness of the United States. The committee will be charged with reviewing AI-related education; open sharing of data and the open sharing of research on AI; international cooperation and competitiveness; opportunities for AI in rural communities (that is, how the Federal Government can encourage technological progress in implementation of AI that benefits the full spectrum of social and economic classes); and government efficiency (that is, how the Federal Government utilizes AI to handle large or complex data sets, how the development of AI can affect cost savings and streamline operations in various areas of government operations, including health care, cybersecurity, infrastructure, and disaster recovery).

Non-profits like AI Now and Future of Life, among others, are also considering many of the same issues. And while those groups primarily rely on private funding, the FUTURE of AI advisory committee will be funded through Congressional appropriations or through contributions “otherwise made available to the Secretary of Commerce,” which may include donation from private persons and non-federal entities that have a stake in AI technology development. The bills limit private donations to less than or equal to 50% of the committee’s total funding from all sources.

The bills’ sponsors says that AI’s evolution can greatly benefit society by powering the information economy, fostering better informed decisions, and helping unlock answers to questions that are presently unanswerable. Their sentiment that fostering the development of AI should be done in a way that maximizes AI’s benefit to society provides a worthy goal for the FUTURE of AI advisory committee’s work. But it also suggests how AI companies may wish to approach AI technology development efforts, especially in the interim period before future legislation becomes law.