ABSTRACT
This article makes a simple argument. First, the potential benefits of autonomous artificial intelligence (AI) are massive but cannot be tapped without solving governance issues related to authority, liability and control. Second, corporate theory has grappled for centuries with the core governance issues facing artificial intelligence, so it has much to offer to AI theorists and practitioners.
To the first point, autonomous AI will soon be capable of managing assets more efficiently than humans. This raises a variety of questions. The largest is known as the alignment problem, which asks how we can align the actions of an AI system with the goals of its users or society more broadly. A classic example of misalignment is a request for an AI system to manufacture paperclips that ends in a system that ‘convert[s] first the Earth and then increasingly large chunks of the observable universe into paperclips’.
To the second point, the alignment problem has been the central focus of corporate theory for the last century, though it goes by different names. When discussing end-user alignment, corporate theorists refer to agency costs. When discussing societal alignment, we call it corporate purpose. Corporate governance, at its core, is a collection of methods for aligning the interests of various parties toward a common goal even when the parties have different preferences, capabilities, resources and access to information. AI governance and corporate governance are addressing the same problem, but AI governance won’t have a century to debate governance, so this article recommends a field that did.
The article shows that corporations and AI systems are similar in purpose, nature and the challenges they face. Next, the article tests this proposition by exploring corporate law’s theoretical framework to identify new solutions in AI governance. To do this, I resituate AI governance solutions back into the corporate theory framework, then consider adjacent, untapped solutions in the corporate theory. I then apply one of these solutions – multitiered governance structures – to AI systems and shows that it improves control and performance over single-agent AI systems. This shows that corporate governance theory can be useful to find solutions in AI governance. While this does not mean the two fields are identical – adaptations will be needed. But it does support the broader proposition that AI governance may be able to copy some of corporate theory’s homework and adapt its centuries worth of experience.
Weitzel, Paul D, AI Governance as Corporate Governance (February 29, 2024), Tennessee Law Review, Forthcoming.
Leave a Reply