Why responsible AI development needs cooperation on safety

OpenAI · 7 years ago · 38

Researchers identified four strategies—risk communication, technical collaboration, transparency, and standards incentives—to encourage AI companies to cooperate on safety norms. Competitive pressure creates a collective action problem where firms may under-invest in safety relative to what would be optimal for the industry overall. The paper proposes these approaches as a way to align individual company incentives with broader safety goals.