SoerenMind comments on [AN #78] Formalizing power and instrumental convergence, and the end-of-year AI safety charity comparison