Practical Governance for AI Alignment

Author: Zafar Shaikhli

Summary: As artificial intelligence (AI) advances rapidly, the critical challenge of AI alignment - ensuring AI systems behave in accordance with human goals, preferences, and ethics - has come to the forefront. However, amidst the technical and moral debates surrounding AI alignment, the potential corrupting influence of corporate interests is often overlooked. This paper explores the risks associated with the convergence of AI alignment research and corporate objectives, highlighting the potential for "AI malalignment," where alignment is skewed by malicious or self-serving corporate interests that may contrast with humanity's best interests. The paper examines the phenomenon of mission drift in AI companies, using OpenAI as a case study, and discusses the existential danger of corporate control over AI alignment, particularly in the context of artificial general intelligence (AGI). The paper emphasizes the urgent need for mechanisms and safeguards to ensure that AGI's potential is harnessed in alignment with the best interests of humanity, rather than corporate ambitions.

Link to PDF version.