The AI extreme risk mitigation philanthropic sector

Abstract

This chapter focuses on the AI extreme risk mitigation philanthropic sector (AIERMPS), a small and unique sector at the forefront of the interaction between AI and philanthropy. After providing an overview of AI’s implications for philanthropy broadly, we introduce the sector, which emerged in the early 2000s to tackle issues related to artificial general intelligence and existential risk. We provide a historical review and landscape analysis, including a description of the sector’s ideologies and culture, and its relations with other communities. From its origins to the present, the core approaches, stakeholders, and culture of the AIERMPS have evolved significantly. We further describe the many challenges that the AIERMPS has faced. This includes the lack of feedback loops that could foster community learning, disagreements about the most promising approaches to risk mitigation, concerns about rapid technological change and uncertain impacts, and disagreements within the field related to research and grantmaking. Finally, we compare the AIERMPS and its goals to other philanthropic and cause-based movements, noting that the AIERMPS has pursued a strategy very different from that of the animal welfare or climate change movements. We show that, while it has succeeded in building coalitions with unusual actors, it has also failed to do so with natural allies. To more effectively advance its goals, the AIERMPS may need to work pluralistically, deepen its coordination with natural allies in AI ethics, develop frameworks to make sense of AI’s harms across multiple timescales and risk dimensions, and adopt social movement-building techniques common to philanthropic efforts.

Ähnliche Arbeiten