Here are some hard problems to crack then, in particular Value Learning: https://intelligence.org/research-guide/