I honestly think it's far more likely that game developers of the future will make use of virtual reality to create realistic games, rather than photorealistic graphics rendering. The latter would be hugely expensive both in computational power and manpower (although procedural generation could eliminate much of the latter, it still depends on your goals).
Also, the great thing about games is that they're not necessarily meant to conform to reality as a standard. If you want photorealistic graphics, go outside. A game that relies primarily on its graphical fidelity is doomed to obsolescence, but a game with good mechanics and an aesthetic is potentially timeless.
I read a blog post recently that outlined why we may never actually see photo-realistic games even as the technology improves.
Basically, it's a MASSIVE amount of work/cost to create individual blades of grass that independently react to a person walking on them, vs just pasting an static grass texture on the ground. And unless the movement of the grass is really important to the story, it will never be cost-effective to create all the art work, physics, CPU processing time, and disk space to include it.
Getting to the point where that's in the eye of the beholder. Those intimately experienced with nuances of rendering may see glaring indications of unrealism, but many are starting to just not see the difference. At one extreme we have movie CGI special effects which nobody notices isn't real; at the other are viewers who don't notice a 4:3 frame stretched across a 16:9 screen.
I haven't read that post. But I think you could have said the same thing about things like global illumination - well, in terms of processor time - yet look at how much more advanced realtime lighting is today than a decade ago. As for artwork and disk space, grass for one can easily be procedurally generated.
Remember when 640kb was ought to be enough for anybody? Never say never in tech world. Someone might and probably will come up with tech to realistically render individual grass blades, just like they did with TreesFX, hair strand simulation technology used in most recent Tomb Raider game.
I'm betting some combination of procedural generation and scanning tech a la Kinect will start taking a lot of the workload off the artists. This has already begun to some extent, with many if not most AAA titles using procedural generation of terrain and vegetation.
The algorithms to control grass reacting may be difficult to write, but that only has to be done once. You could as well have said ten years ago that it would never be cost-effective to create photo-realistic CG films, but that has been pretty much accomplished at this point, and game developers are already borrowing techniques from films.
