Is the specification problem basically solved? Not the alignment problem as a whole, but specifying human values in particular. Like, I think Claude could quite adequately predict what would be considered ethical or not for any arbitrarily chosen human