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.
Comments
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.