What they’re generally talking-about, Gentle Reader, is the fact that a “texture” in the CG world is really a sort of mathematical filtering mechanism. It enables inputs, such as the “normal” of an object (which way an arrow emerging from the front faces, in relation to the camera), to affect one-or-more outputs, which are characteristics of the image such as “alpha” (transparency). The effect is calculated in three dimensions, X Y Z, and you can “map” those three channels in interesting and useful ways, mostly by experimentation.
When you consider any object, such as the lamp, being X-rayed, what the X-ray beam would actually be measuring is its density: how much “air” vs. how much “object” the beam is actually passing through. Passing through a solid object attenuates the beam. But as far as I know there is no way to do that, at least not without some fancy programming.
But we can “cheat.” In the case of the thin, curved, lamp we know that the “sides” will appear to be a little denser, and we know that “sides” are where the normal-vector points away from the camera; it points to one side. It’s not accurate but it’s convincing and looks good. The trick won’t work for a cube because the entire face points to one side and the eye sees it as “strange, funky lighting,” like black-light on a 1960’s velvet poster, not “an X-ray.”
A more complicated solution would be to figure out each “layer” of the object from back to front and then merge, or “composite” them together into a single image. Also a “cheat.”