Effective altruism, AI safety, and learning human preferences from the state of the world
Briefly