Remember one or two AI generations ago (a few months), when image-gen AIs weren't usually able to produce coherent text, and that was touted as proof that they

Remember one or two AI generations ago (a few months), when image-gen AIs weren't usually able to produce coherent text, and that was touted as proof that they don't really "understand" what they're doing and will never be generally capable?