Interesting. I've skimmed the article and I can easily believe that some people can consciously observe their model of the world from a physical position different from a position of their body.

From a perspective of ML/computer vision, think this way: You take bunch of photos and then build a 3d model. After that, you can construct the image taken from any location within the observed area. And if the model is good enough, you can construct realistic images even outside the visible region.