So I know that I'm basically asking to identify the main problem to be solved in creating realistic 3d Graphics, but as someone without much technical experience in the field, I'm curious that if these problems can be identified, what the issue is with implementing them programmatically.

The picture below from Hitman is what I would call "ultra-realistic". But, it still definitely looks like a 3d model. If you were to put a photograph of the same scene with a real actor and background, there would probably be some differences obvious enough to be able to point out which is which.

Why is that? < 100% accurate subsurface scattering? Is the lighting ever-so-slightly off? etc.

PS, sorry for tag choices. I am not familiar enough with this SE to know what to choose. Please edit better ones in, if you know any.