AI, centralization, and the One Ring — LessWrong