The only way to test whether Hobbits were a different species would be to have sex with one and see if viable offspring could be produced. Ditto for Neanderthal. Everything else is just empty academic autoerotica.

or extract some ancient nuclear DNA and inject into a human ovum - whose nucleus had been previously extracted - and see what happens...

A different way of interpreting innate grammar capabilities would be to first ignore language. Brain architecture evolved to support many distinct functions. The interaction of those functions provides a structure that constrained later language development. After humans acquired rudimentary language skills, language and brain structure co-evolved. The present innate grammar capabilities could reflect brain structures that evolved before language rather than a general grammar capability that evolved later.

It sounds like he's proposing a variant on how our immune system functions: generate tons of diversity, and the unused stuff gets overtaken (outreproduced) by the adaptive stuff. Clever, but not right -- why bother trying lots of things out to see which is right, when you can just listen in and use that data to point you toward one template or other? No one's nailed them all down, but there are strikingly fewer parameters of variation than you'd think, in terms of syntactic or phonological rules; and there's no variation in composition semantics (how you interpret a phrase or sentence).

The immune system thingy would only be worthwhile in an utterly unpredictable world (like which pathogens you'll encounter). Linguistic patterns are much more predictable. Take the one part that actually is unpredictable -- vocabulary (sound-meaning pairings). Are we born with every possible combination of sound & meaning, which we then whittle away based on whether or not the words are present in our ambient language or not? Nope. I haven't looked at the state of the art in a few years, but last I checked, the "fixing parameters and acquiring vocab" model was more or less correct.

Interesting linguistic sidenote for the non-British:
My extended family speak perfect English.

Not a tongue-in-cheek joke! They treat collective nouns as plural, not "my family speaks"... "The army are delivering the mail," etc.

Yang apparently argues that babies are born with an extensive set of innate grammatical capacities - sufficient to cover all the world's languages - most of which are then discarded or suppressed as the baby learns the specific grammar of its own 'mother tongue'.

In a related story, they also discovered that babies are born with an extensive set of innate athletic capacities - sufficient to cover all the world's sports - most of which are then discarded or suppressed as the baby learns the specific sports of its own 'homeland'.

I've heard that babies will make various sounds not associated with the language that they will eventually acquire, such as clicks and velar fricatives and that these are lost after exposure to the sounds that are present in the language of the community in which they are raised. Not the same thing as grammar, but still, it's something.

It's not that they begin producing all those sounds -- if you've ever heard newborns or infants talk, they can't produce lots of sounds even in their native language -- but that they are capable of mentally distinguishing all the sounds of the world's languages if they hear them. For example, they can tell the difference between the aspirated and unaspirated consonants in Indic & Dravidian languages, even if the sole ambient language is English. They can also distinguish stress-timed languages like English or Dutch from syllable-timed or mora-timed languages like French or Japanese (respectively) based on the prosody.

So again, you can view this more of a linguistic "attendance list" that the baby conducts, and only produces the checked-off sounds by linguistic maturity, rather that producing everything right from the start and whittling that away over time.