“Critics who claim there is no meaningful progress on the AI alignment problem are mostly pointing to a lack of progress on creating a formal theory of alignment.”