In class we saw the capabilities of the Unreal Engine 4 which got me interested in the challenges real time rendering faces.
I learned from an Nvida webpage (http://www.nvidia.com/object/tessellation.html) that one of the challenges with rendering complex environments is that in order to keep the gpu workload manageable, game engines must switch between different levels of detail. Some engines accomplish this by forgoing rendering distant objects in the environment such that distant objects can suddenly appear on screen (which doesn't make for a very realistic gaming experience).
Dynamic tessellation solves the balancing act between realism and workload by "varying the level of detail on the fly." This means that far away objects are rendered with only a few triangles. As those objects move closer and the level of detail demanded increases, they are rendered with an increasing number of triangles.
This approach also reminds me of the point Professor Crane brought up in class, which is that if any object is far enough from the viewer, it can be sufficiently rendered as a point (or in this case, just a few triangles).