Can't they train these LLMs to produce human-readable English?

And human-readable code, for that matter.