I agree with your claims, and while it is still in its "infancy stage", good proper use with proper data (not some weird "woke" input) might adjust its "skewed" logic about us humans.
It is a good tool to use, when we know how to use it. There is "no human" behind basic models just as yet.

I heard specialised language models actually still hire humans to "ping in" the respomse, but i also heard they are underpaid