The horror you witness is YouTube.
And yet, I am not surprised.
Visual Free Will in action. I just noticed at least 5+ videos are blended with every other clip translated. It's like a visual dance mix of Youtube.
The (air)plane turned into a plane (land). The black kid (with stethoscope) inexplicably turned into a white version of himself kid on a webcam. An old woman turned into a young man. The mind is strange; and makes you wonder about the subject(s) tested.
We gotta play a game of "Spot the difference" with the video. This can get interesting.
It seems to be related to image association and the way the software works. At one point, the clip is of Steve Martin standing in a particular pose, wearing a medal around his neck. Subject 3's primary "posterior clip" is of Adam Savage standing in a very similar pose, with a convention badge around his neck. This image dominates the "reconstructed" image, down to the text on Adam's t-shirt being in the reconstruction. What I'd be curious to know is, does that mean that the subject's subconscious is relating the two images, or is it simply that similar attributes in the images cause similar "memory locations" in the brain to fire, so that what the software picks up is "Hey, this Adam Savage image is a 98.5% match in neural activity to the current state", so that forms the primary template for the image, and then all the other similar patterns are overlaid on the image with decreasing strength?
That's a neat idea if that's how it works, but it doesn't say much of anything about the test subjects. Though it does maybe say something about how our brains process details and kinesthetics.