Google and DeepMind have launched the DeepMind Institute, aiming to broaden the dialogue on artificial general intelligence (AGI). The institute, led by key figures from Google and DeepMind, will explore diverse perspectives on AGI, from economic policies to model transparency.

The inaugural essays tackle complex issues, such as the diminishing transparency of AI models and the need for new standards. DeepMind safety researchers argue for direct engagement on safety trade-offs, while Demis Hassabis proposes a U.S.-led framework for evaluating advanced AI models.

As the industry shifts from general concerns to concrete proposals, the essays signal a new phase in AI safety. This move towards more structured oversight and disclosure represents a significant step in navigating the rapidly advancing field of AGI.

The essays arrive as industry leaders, including Anthropic’s Dario Amodei, urge for more cautious development of AGI, suggesting a potential slowdown in the race to AGI dominance.