Everything Is Not Terminator America's First Ai Legislation

• Jurisdiction: United States,Federal
• Publication year: 2018
• Citation: Vol. 1 No. 3

John Frank Weaver^*

In December 2017, a bipartisan group of U.S. senators and representatives introduced the Fundamentally Understanding the Usability and Realistic Evolution of Artificial Intelligence Act of 2017 (the "Act").¹ Although the Act is a long way from becoming law—it is being considered concurrently by the Subcommittee on Digital Commerce and Consumer Protection of the House Energy and Commerce Committee and the Senate Committee on Commerce, Science, and Transportation Committee—it would be the first U.S. legislation to focus on forming a comprehensive plan to promote, govern, and regulate artificial intelligence ("AI").² Other countries are already actively considering legislation that would address AI³ or have already passed legislation or regulations that address some key aspect of AI.⁴ The Act would form the Federal Advisory Committee on the Development and Implementation of Artificial Intelligence (the "Committee") to help inform the federal government's response to the AI sector. The three primary components of the act—the definition of AI, the formation and composition of the Committee, and the Committee's functions—are solid first efforts toward the regulation of AI, although the Act clearly anticipates further legislative action.

Definition of AI

The Act provides a very broad definition of artificial intelligence:

A. Any 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. Such systems may be developed in computer software, physical hardware, or other contexts not yet contemplated. They may solve tasks requiring human-like perception, cognition, planning, learning, communication, or physical action. In general, the more

[Page 201]

human-like the system within the context of its tasks, the more it can be said to use artificial intelligence.

B. Systems that think like humans, such as cognitive architectures and neural networks.

C. Systems 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.

D. A set of techniques, including machine learning, that seek to approximate some cognitive task.

E. Systems that act rationally, such as intelligent software agents and embodied robots that achieve goals via perception, planning, reasoning, learning, communicating, decision-making, and acting.⁵

That definition could potentially encompass any robot, program, autonomous or automated device that could be viewed, interpreted or marketed as artificial intelligence. The Act would even provide a legal distinction between "Narrow Artificial Intelligence" and "Artificial General Intelligence," which is nice but unnecessary since, as the Act acknowledges, artificial general intelligence is a "notional future artificial intelligence system," i.e., not happening any time soon.⁶

But broadly defining AI is a smart approach at this point, given how quickly the technology is developing and the lack of a widespread consensus on how to define AI.⁷ And by creating an inclusive, almost nebulous, working definition, the Act avoids the historical problem of a shifting standard for technology to qualify as AI. For example, chess was once considered a barometer of AI, but that has gradually changed since computers were able to play a decent game of chess in 1960.⁸ IBM's Deep Blue beat the best human player in the world in 1997.⁹ These developments made many suggest that skill in chess is not actually indicative of intelligence,¹⁰ but did chess really become disconnected from intelligence merely because a computer became good at it? As one expert laments, "[a]s soon as it works, no one calls it AI anymore."¹¹

However, as the AI field starts to better define itself, the Act's definition will need to be revised, most likely by more closely defining what the law will consider AI and what it won't. It is entirely likely that a program or device that can "perform tasks under varying and unpredictable circumstances, without significant

[Page 202]

human oversight," will be considered too general, as a good lawyer could make a Magic 8 Ball fit into that definition. Fortunately, the Act acknowledges the rapid development of AI technology and empowers the Committee to revise the definitions in the Act as it considers appropriate.¹² That is a foresighted provision that begins to address the real danger of the Committee not being nimble enough to keep up with how quickly AI develops. Almost all AI legislation will need to overcome that issue, and it is a good idea to have the bill that creates the eventual Artificial Intelligence Regulatory Agency ("AIRA") grant that entity the power to revise the definitions and key terms in its organic statute.¹³

Formation and Composition of the Committee

The Act establishes the Committee in the Department of Commerce, however, it would be interesting to know how the Act's sponsors chose that department. It is a...