Whitepaper

Abstract

Recent progress in AI has demonstrated its tremendous potential to solve problems at scale. However, the current AI paradigm is highly centralized and limited in domains where data and knowledge are distributed across organizations and geographies. This paper advocates for a self-organized approach to building and deploying AI in decentralized contexts, such as healthcare, climate science, and supply chain management. We identify five key technical challenges — (1) privacy, (2) verifiability, (3) incentives, (4) orchestration, and 5) crowdUX that need to be addressed to unlock the potential of AI in these domains. By overcoming these challenges, we can enable distrusted, disincentivized, and disinterested entities to collectively solve global problems while pursuing local objectives. This perspective aims to stimulate discussion and research towards a more decentralized, participatory, and resilient future of AI.