I would think that either FB-level facial recognition, or Disney-level copyright/ContentID frame recognition would be helpful for all this. I don't suppose they're thinking of open-sourcing it.