An International Consortium for Evaluations of Societal-Scale Risks from Advanced AI

• DOI: 10.48550/arxiv.2310.14455
• Publication Date: 2023-01-01
• Authors: Ross Gruetzemacher; Alan Chan; Kevin Frazier; Christin Manning; Štěpán Los; James Fox; José Hernández‐Or...; John Burden; Matija Franklin; Clíodhna Ní Ghuidhir; Mark M. Bailey; Toby D. Pilditch; Kyle A. Kilian
• Keywords: Prioritization; Frontier

Given rapid progress toward advanced AI and risks from frontier systems (advanced pushing the boundaries of capabilities frontier), creation implementation governance regulatory schemes deserves prioritization substantial investment. However, status quo is untenable and, frankly, dangerous. A gap has permitted labs to conduct research, development, deployment activities with minimal oversight. In response, system evaluations have been proposed as a way assessing development systems. Yet, budding risk evaluation ecosystem faces significant coordination challenges, such limited diversity evaluators, suboptimal allocation effort, perverse incentives. This paper proposes solution in form an international consortium for evaluations, comprising both developers third-party evaluators. Such could play critical role efforts mitigate societal-scale AI, including managing responsible scaling policies coordinated evaluation-based response. this paper, we discuss current its shortcomings, propose issues regarding implementation, lessons that can be learnt previous institutions existing proposals institutions, finally, recommend concrete steps advance establishment consortium: (i) solicit feedback stakeholders, (ii) additional (iii) workshop(s) (iv) analyze create final proposal, (v) funding, (vi) consortium.