[makeitdouble]

You put a strong focus on how we evolved to deeply care about small facial expression differences and face features to identify and interact with an individual.

These stories are about how we also deeply care about labels and categorization. Aren't we just looking at the natural selection (making them not "last long") of these way too rough AIs that step on bounderies that are pretty important to a lot of people ?

[Cycl0ps]

Ha! I like that! Yes, I guess in a way we are. These models are always being evolved in their own version of 'natural' selection. They go through tens of thousands of mutations before finding one that guesses well enough to be pushed to production. This is just another stage of that algorithms life cycle I suppose. If you want to take an optimistic view of it this is just another part of the tuning process. The AI can train for as long as it likes, but the real thing it's being weighted against is public outcry.