Artificial Intelligence (AI) systems have undergone dramatic advancements in capabilities over the last five years. The world is projected to reach Artificial General Intelligence[1] in 2027 and to experience a superintelligence explosion thereafter.[2] The potential role and exigent importance of law on this trajectory from AGI to superintelligence remain underexamined. This article aims to introduce three fundamental aspects of the connection between AI and law in the development of superintelligence: first, it establishes a constitutional imperative for superalignment; [3] second, it argues that law is an important component of the solution to the superalignment problem; and last, it describes the impact that the rise of superintelligent systems will have on constitutional structures.
Superalignment is a Constitutional Question
The discussion of superintelligent systems has unique societal and legal implications. Unlike technologies to date, superintelligent computers differ in their human-like ability to create knowledge. They are not neutral objects, as technology has been up until now. Because they can decide, create, and act autonomously, superintelligent systems have the potential to cause harm within the economy and society. Without successful alignment between ethics, morals, and laws, superintelligent systems are likely to cause harm to individuals and society. The constitutional link to superalignment stems from the protection of the fundamental rights declared in the Constitution and the Declaration of Independence.[4][5] We can only protect and carry these values into the post superintelligence world if we establish the values and fundamental beliefs into the rudimentary fabric of superintelligent systems. We have a constitutional demand to do so.
Law as the Target Function of Superalignment
Achieving superalignment is a technical problem that begins with the framing of the issue and the objective. On a task level, the problem can be summarized as: “How do we align systems on tasks that are difficult for humans to evaluate?” [6] On a societal level, the challenge is better encapsulated as: “How do we control systems that are orders of magnitude more intelligent than humans are?” The law offers a robust and interpretable framework to provide parts of the answer. [7] Alignment with democratically derived law not only fulfills a constitutional demand but also provides the answer to the question of what the target of alignment must be. [8] Through the incorporation of legal standards, statutory interpretation, and case law, AI can learn to generalize human goals to novel situations. The target function for the alignment of superintelligent systems must be the adherence to the law and the legal code of our societies. Only then can we ensure that systems, exponentially smarter than humans, are not only incapable of causing harm but also foster human progress.
The Constitutional Posture of a post Superintelligence World
The rise of superintelligence has profound implications for the future of the structure of the state. Superintelligence will reshape strategy, redefine jurisprudence, and transform political currents. Much like the nuclear arms race in the 20th century, the development of superintelligence will provide a decisive military and economic advantage to the nation that achieves it first. Superintelligence could condense decades of scientific progress into merely a few years. While superintelligent computers will be as consequential as nuclear weapons, they differ in that law and legal reasoning become core components in shaping these systems. While the 20th century doctrine of nuclear deterrence did induce new legal treaties and contracts, nuclear weapons did not differ significantly from country to country. Superintelligent systems will be heterogeneous, because their consciousness will be forged by the laws, ethics and politics of their creators. These differences could prove decisive, and given their self-reinforcing nature, the country that first develops superintelligent systems may dominate the global economy for the foreseeable future and, in doing so, dictate the constitutional structure of the state.
Conclusion
The advent of superintelligent AI systems presents both unprecedented opportunities and profound challenges. As these systems approach and potentially surpass human cognitive capabilities, the imperative to align them with human values and constitutional principles becomes critically important. This alignment is not solely a technical endeavor but a constitutional necessity, rooted in the protection of fundamental rights and the preservation of democratic ideals.
Given the significant impact that superintelligence will have on global power dynamics and societal structures, it is imperative for the legal scholarship to engage actively with these emerging challenges. Legal scholars and practitioners must embark on a quest to explore, debate, and define the laws governing superintelligent AI, much like the framers of the Constitution who meticulously crafted the foundational principles of the United States of America. This intellectual pursuit is essential to safeguard the democratic values and human rights that are the bedrock of our society.
By embracing this responsibility, the legal community can play a pivotal role in shaping the future of AI in a manner that reflects our collective ideals. Through thoughtful legislation, rigorous scholarship, and ethical considerations, we can ensure that the rise of superintelligent systems enhances human prosperity while adhering to the constitutional principles that have long guided this nation. In doing so, we honour the legacy of our forebears and contribute to a future where technology and humanity coexist in harmony, upholding the values of the United States we all love and cherish.
______________________________________________________
[1] See, e.g., a definition of AGI from OpenAI, “highly autonomous systems that outperform humans at most economically valuable work” https://openai.com/charter/ (accessed September 12th, 2024).
[2] Leopold Aschenbrenner, Situational Awareness: The Decade Ahead, Situational Awareness (2024) ("AI progress won’t stop at human-level. [...] We would go from human-level to vastly superhuman AI systems.").
[3] Superalignment is the technical problem of aligning superintelligent systems with human values.
[4] U.S. Const. amend. XIV, § 1 ("No state shall make or enforce any law which shall abridge the privileges or immunities of citizens of the United States; nor shall any state deprive any person of life, liberty, or property, without due process of law; nor deny to any person within its jurisdiction the equal protection of the laws.").
[5] U.S. Declaration of Independence para. 2 (U.S. 1776) ("We hold these truths to be self-evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Liberty and the pursuit of Happiness. -- That to secure these rights, Governments are instituted among Men, deriving their just powers from the consent of the governed...").
[6] Jan Leike, What is Alignment, Aligned (Oct. 30, 2023), https://aligned.substack.com/p/what-is-alignment.
[7] John J. Nay, Law Informs Code: A Legal Informatics Approach to Aligning Artificial Intelligence with Humans, 20 Nw. J. Tech. & Intell. Prop. 309 (2022).
[8] Richard H. McAdams, The Expressive Powers of Law, Harv. Univ. Press (2017) at 6-7