[mcdonje]

>I don't know what that proves.

It proves something, but not much. Those models with those inputs (mostly HN articles) were benign or even a net positive for society.

Other models with different training (upstream of the blogging user), or with different inputs (maybe it finds a different article posted to HN or another site that proves foundational to its evolving perspective), could end up behaving differently.