What's the state of Ai rigging and mocap?

I’m curious - Is anyone using AI rigging or motion capture as a regular part of their workflow?

I’ve had a shower thought about how to use openpose, or similar, to auto rig a model and I wanted to hear thoughts because I don’t actually know how to code it but I’m sure it could work.

Openpose takes a 2d image and outputs x,y coordinates of various joints, with lines connecting them. Shouldn’t it be possible to generate a series of images of a 3d model from different angles, get openpose’s joint coordinates from those images, project lines from those cameras and calculate the nearest intersection for each joint, thereby rigging a model?

I need the pivot of the bone I’ve highlighted to be exactly in the middle of that edge loop. Not “somewhere around there” give or take 40 pixels based on a depth map, exactly that position. All the other finger joints - same thing. And on, with the rest of the rig.

I doubt that AI works with this level of accuracy.