@SecurityWriter @chipotle I don't think it can really improve, at least the limit is likely how humans themselves also understand it.

And we often fail to interpret properly, only reason it somewhat works is due to redundancy in both the languages and asking/looking-at others.
Programs that try to look clever will just fail, just like people who are too sure of themselves to assert that their interpretation was indeed correct.