Five AI Challenges in Strategyproof Computing

Abstract

Computational systems are now distributed by default, and designed, owned and used by multiple self-interested parties. In the face of this growing system complexity, we need a unifying design paradigm, that supports continued innovation and competition while promoting easy deployment, operation, and use. Focusing on the central problem of resource allocation---the arbitration of shared resources among the competing demands of users---we introduce a paradigm of strategyproof computing, a vision in which individual users can treat other resources as their own. We layout the benefits of strategyproof computing, and identify five AI challenges in making this vision a reality.