I think one bit of harm reduction advice from me would be to never use second party pronouns when using models
Then they are unlikely to output a 1:1 conversation which would trigger the social center of the brain
Discussion
I think one bit of harm reduction advice from me would be to never use second party pronouns when using models
Then they are unlikely to output a 1:1 conversation which would trigger the social center of the brain
@natty@astolfo.social "LLMs do not comprehend your cruelty" and "i'm still worried about your wellbeing if you do actually go out of your way to be cruel to LLMs" both feel like valid thoughts that can coexist
especially around "clanker"
And please for the love of satan don't "torture" them
Not because they could feel it, but because it impacts the human and silently traumatizes over time