This post is about AUP-the-concept, not about specific implementations. That plan increases its ability to have paperclips maximized and so is penalized by AUPconceptual. We’ll talk specifics later.
ETA: As a more general note, this post should definitely have an “aha!” associated with it, so if it doesn’t, I encourage people to ask questions.
The subagent problem remains: How do you prevent it from getting someone else to catastrophically maximize paperclips and leave it at its power level?
This post is about AUP-the-concept, not about specific implementations. That plan increases its ability to have paperclips maximized and so is penalized by AUPconceptual. We’ll talk specifics later.
ETA: As a more general note, this post should definitely have an “aha!” associated with it, so if it doesn’t, I encourage people to ask questions.
It did have that “aha” effect for me. (The drawings and the calligraphy were also amazing.)