Your analogy reminds of messed up fingers and hands in image generation models just a year ago. Now that is pretty much solved. These days they are generating videos you can't tell apart from reality. This makes me believe these nuances will keep reducing and eventually become very hard to notice and find in may be every task.