Years ago, a conclusion of automatic translation projects seemed to be that these systems didn't need to understand much or anything about a new language to do

Years ago, a conclusion of automatic translation projects seemed to be that these systems didn't need to understand much or anything about a new language to do a half decent job of translating it. All that was needed was a sufficient body of material already existing in multi-lingual form - such as international standards, international organizations proceedings, etc. Was this ever attempted for translating from one programming language to another? (Except multi-programming language examples are usually not straight translations, they are re-writings from scratch nominally to exploit each programming language strong points and in practice often not even doing that.) Is DNA/RNA sufficiently comparable to a language? (Except that there is no multi-lingual equivalents there which seems to be where the LLM approach comes in.)