Well, I watched the video.. hair was done nearly as nicely in 3DMark03, which was what, about 8 years before Alice with PhysX that slowed down even 2x GTX 560 Ti to a ~20fps stutter.. being so poorly optimized. I read on in that link, and the optimizations show much better performance with nearly as good physics.
Moral of the story: We didn't need proprietary NV GPU PhysX for swaying hair about 10 years ago.
Superseding moral of the story: If AMD and NV collaborated on an open-source physics API, it would have boosted both AMD and NV's success a good deal, by making games more attractive and complex, encouraging gamers to spend a little less on the less important CPU and a bit more on the more important GPU doing more physics. It would have also boosted PC gaming further. AMD and NV could have had greater sales thanks to this. Intel did the smart thing by snatching Havok, and trying to keep the CPU as important as possible. There was not as much constructiveness until it becomes standardized within DX, the standard API for PC games.
Same goes for Tessellation. It could have been done with this flying head monster:
Gosh, why did that flying head have to look like an octagonal stop sign that I see everyday on the road?!?
A few years after UT used Truform, Doom 3 could have had a way better use of tessellation to get rid of the stupid polygons. Even mild tessellation costing only 20% of GPU performance back then would have taken care of it real nicely for its time. So nicely that in fact, ID Software would've made some more revenue from Doom3 copies (selling more, plus people buying them earlier, at higher prices). The PC gaming industry would have more of an uproar with its advantage over console versions. Carmack stabbed PC users in the back, giving consoles the priority when it came to how many polygons would be used for monsters, etc.. $.
We had to suffer polygonal boobs for a decade because of selfish corporate reasons.
Sometimes, collaborating together brings greater $ for both sides, than otherwise.