VRChat's AI Labeling Fight Points at a Split in How VR Uses AI

VRChat just walked into a backlash over refusing to label AI-generated content. The survey asked users if the platform should tag AI slop. The answer from the community was basically: yes, please. VRChat said no.

That argument is worth sitting with, because it points at a real split in how AI is used in VR.

One camp uses generative AI to flood a space with cheap volume — unlabeled, unaccountable, indistinguishable from human work. The other camp uses AI as a deterministic instrument inside a task: not to produce content, but to shape perception while you engage with content you already chose.

I think about this a lot. My nephew has amblyopia — lazy eye. His left eye does almost nothing. For years it was patches, and he hated every minute. Then we found Amblyotube, a Meta Quest app from Seven Sports, and the whole thing flipped.

It doesn't generate anything. It takes a YouTube-style video he actually wants to watch and renders a different experience to each eye. A Dominant Eye Shader applies gentle, adjustable digital blur to his stronger eye. An AI Sharpener (MFBF) detects human figures in the lazy eye's view and outlines them in bold red with yellow-green highlights and a steady pulsing flicker. Optional Wedge Occlusion blacks out portions of each eye's view so the two eyes stop fighting over the same pixels. And a magenta focus cue in the lazy eye plus a neutral grey dot in the dominant eye train binocular fusion — the brain learning to merge both signals into one 3D image.

No adhesive patches. No red-blue glasses. Thirty to forty minutes a day, watching content he'd watch anyway. He doesn't even notice he's training.

The VRChat story is really about transparency and intent. If you're going to put AI between a person and what they see, be explicit about what it's doing and why. In this case the "why" is written on the box: supplementary vision training and an assistive tool for binocular coordination and depth perception. Not a diagnosis, not a treatment, not a cure.

Meta Quest app store: https://www.meta.com/en-gb/experiences/amblyotube/25906906972338493/

Ages 13+. Recommended 30–40 minutes daily, max 1 hour.

The labeling debate is healthy. What matters is whether the AI is there to replace the human or to help them see.