Frankly, I can agree that this argument, when applied to natural languages (in which one can indeed describe algorithms with sufficient detail for execution, in a way that naturally makes sense to speakers of the language)
Sufficent detail for execution... I do hope this is the case but as of now linguists have not flushed out the complete English grammar. To wit, there are several competing structural representations in each linguistic domain which make understanding grammar wholly an incomplete problem. Also, cross-linguistic uniformity is something that some linguists value, and work on less popular languages is not as extensive as English. Linguists certainly have not addressed the production errors that people make in daily conversation because they, and cognitive scientists, do not full understand the language production process.

So, well that's my addition to the discussion.