And he completely missed my entire point anyway, getting hang up on bullet point features again instead. I leave for a week and return for this??
The point is that adapting these new technologies will almost certainly bring significant visual advances, because so much of the real world imagery depends on the underlying numbers - proper reflections and contrast and such just can't be done without floating point values and gamma correct linear calculations and energy conversation instead of trying to cheat everything.
AC4 is, at least to me, a very good example of not getting stuff right because of engine limitations. The skin shading is terrible, and a lot of the lighting and other shaders look wrong too, the whole picture is very far away from the wonderful concept paintings and not pleasing to look at - but this is, once again, mostly an issue with the legacy systems that the team had to work with. The ACU teaser is a very good demonstration of how a physically based system can overcome these limitations.
What you basically get is very high quality "free" realism, as long as your artists get how the new approach works and learn to create proper texture maps. The next step is to look at how real world movies are made - how the lighting is manipulated with bounce cards and rim lights and such, or how the various photography related parameters like exposure or color grading build on top of real-life visuals. Then there are the assets, the costume design, the "make-up",
In other words, proper art direction still matters just as much as it did in the age of less photorealistic games. PBS is in some ways a limitation, but the boundaries are much wider and there's a lot left to explore in future games.
And of course PBS still isn't mandatory and non-photorealistic rendering is still a valid option.