... rather a particularly interesting venue to discuss. The idea that Linux kernel development can be used to argue emergent behaviour one way or another strikes me as odd, given the sample base of one project used in the keynote. My freedom is freer than yours is not a discussion for auditoriums, it's one for pubs. ;)
Apple actually doesn't hate the GPL, they ship quite a bit of the code in Mac OS X under it, and provide the source code for it, just not in the kernel. So the remainder of talk from there seems to be based on a mistaken premise.
Factors like who might benefit from inclusion of some code certainly matter, otherwise some 'stable kernel driver API' would have emerged by through economical interest among proprietary kernel module suppliers in that particular area of the kernel, etc.