Comment by girfan

Can someone help clarify what it means, concretely, to make binding agreements to proceed "safely and slower" in the context of AI research (whether or not at frontier labs)? Are there specific policies and actionable guides being proposed?