I think it’s a great idea. In fact, there was a thread in the forums where I said that this may happen in the future.
I’m no AI expert, but I wonder what’s stopping them from doing this today. It could require a massive amount of manpower to train the AI system to fix this, with all of the possible photogrammetry scenarios there are.
Nonetheless, I definitely support the idea.