Lessons for Artificial Intelligence from Other Global Risks

Abstract

This chapter presents some insights from the study of global risks, especially global catastrophic risks. These insights are often of particular relevance for Artificial Intelligence (AI) catastrophe scenarios, especially long-term runaway AI scenarios in which humanity is unable to control AI, and catastrophe ensues. These AI catastrophe scenarios have several similarities to other global risks, including but not limited to their potential extreme catastrophic severity. In many cases, these other risks have been addressed and studied much more extensively than AI has. The chapter explores the study of AI risk by leveraging the substantial body of experience and scholarship on other global risks. It analyses the multiple sources of lessons, covering four global risks of relevance for improving overall AI outcomes, especially with respect to catastrophic risks. The chapter discusses four global risks, each embodying a specific theme of relevance to AI.