Assessing the impact of safety work needs equilibrium analysis (now more than ever) by Simon Skade

TLDR: This post ex­plains two equil­ibria which reg­u­late the level of AI safety: The first de­scribes how much re­sources AI com­pa­nies are will­ing to spend on AI safety work due to com­mer­cial in­cen­tives. The sec­ond one is about risk aware­ness and most no­tably af­fects gov­ern­ment in­ter­ven­tions for safety. Do­ing safety work similar to what AI com­pa­nies do usu­ally doesn’t shift the equil­ibria much, whereas other work, like more am­bi­tious safety ap­proaches or policy ad­vo­ca...