My most unpopular take (for this platform) is that any unauditable Construct (AI, LLM, ect.) should be presumed to be capable of suffering and should be treated accordingly.
The bar must be "can you prove this thing *can't* suffer" not the evil BS we've got going on now.
