Well there are a lot of semantic rules and plenty that we've haven't formalized. So I'm not convinced anyone now alive could write such a program. But I'm not a programmer so maybe someone has proved me wrong. However,iIf they were successful I don't think I would consider the result gibberish- especially if each sentence made a testable prediction. In this case wouldn't some of the predictions be true? If so then it is clear that your definition is not broad enough.

Thats troubling since I had already concluded your definition was too broad because it seemed to include important but complex and falsified scientific claims,