I think there's potential for things like Apple's U1 chip and Google's Soli chip to add cheap and accurate gesture control to consoles without the need to hold controllers at all. Add an array of mics and the next gen of Guitar Hero can be based on a group's singing and dancing but graded on a per-voice / per-body basis.

Because of a console's placement in front of a group instead of attached to a single person's head, I think it's better positioned for a wide variety of entertainment experiences.

Large TVs are now cheap enough that they can fill an entire group's field of vision with a shared experience as opposed to having to buy 1 headset per person. Obviously turning your head ruins the immersion of a TV, but I would argue that's not always a bad thing. Sometimes it's fun to be "trapped" in VR, but sometimes it just makes the experience more isolating and difficult to operate (eg how do I watch my kids playing on the other side of the room while I have a headset on?).

That same argument could have been made for 3D tech in TVs a few years ago, and we all know how well that technology took off. It seems most people seem to want an incrementally “better” experience, not a fundamentally different one.

I'd prefer physically accurate skeletal animation because even with all the mocap, animations still look sooo fake. Maybe neural network post processing step would do some good? Maybe some life-like flexibility in interactions instead of everything being little finite state machines? Maybe some neural network game optimisation to be more fun for given player?