I have a photo of an elephant seal that Seek informs me is actually a clouded monitor lizard.
I also have two photos of identical plants growing inches apart from each other that Seek informs me are unrelated.
I'd love to see the little spiders or maybe silkworms that lay tiny lines in the grass of my neighborhood that shimmer in the low morning light and dew.
Not. You still would need an human to sample this moths, kill them appropriately, and manipulate it carefully to move it in the correct position under your binocular in the correct light and place. They are fragile and you can't do it without an human.
Would be worse than what you have currently. No capability to sample selectively, or to jump in advance to all the interesting parts (instead to do all the walk for every single species).
In the end you will have a worst performance, much more work to do, and none of the trained experts available to put any control in your system. Mimetic moths can deceive even a well trained expert.
We have videos from last years conference that highlight the tech that tries to rapidly barcode species, that can give you a flavour of the complexity involved. This is perhaps the world's leading taxonomists exploring how, practically, to do such: https://www.youtube.com/watch?v=VX_1wdmROhU.
In any case the main idea is not destroy the samples throwing it into a blender