Towards Human-Like Machine Intelligence: Generalizability, Transferability, And Trustworthiness