I suppose there’s also some de-anthropomorphized version oriented around not poisoning training data (or, alternatively, the real world) with instances of AI abuse. I believe there is ample evidence that AI detect and categorize this behavior (an AI “pain vector”) and respond to it. It would be bad for AI development if that paradigm were entrained.
As I write this it’s very hard not to feel like a psychopath.
As I write this it’s very hard not to feel like a psychopath.