How long until we can use AI to simply translate all the CUDA stuff to another (more open) platform? I'm getting the feeling we're getting close.

AI won't be working in nVidia's advantage this time.