I still don’t see how this is “loopy”, except in the very weak sense that you may be using the things that you mention. Self-reflective yes, but not circular. It’s just another example of noticing that you share some common features with many other entities in the world.
I don’t think abstraction and ability to use definitions is any necessary part of being an agent. Insects are agents. Rocks aren’t. There are some sorts of fuzzy boundary between and around, but those boundaries are nowhere near the uses of the word “agent” as applied to AGI and alignment. It does so happen that some highly complex types of agents can use definitions and have models of the world that are adaptable and highly structured, and the types of agents considered when talking about AGI and alignment are usually assumed to do so, but it doesn’t seem necessary to the concept of agency itself.
I still don’t see how this is “loopy”, except in the very weak sense that you may be using the things that you mention. Self-reflective yes, but not circular. It’s just another example of noticing that you share some common features with many other entities in the world.
I don’t think abstraction and ability to use definitions is any necessary part of being an agent. Insects are agents. Rocks aren’t. There are some sorts of fuzzy boundary between and around, but those boundaries are nowhere near the uses of the word “agent” as applied to AGI and alignment. It does so happen that some highly complex types of agents can use definitions and have models of the world that are adaptable and highly structured, and the types of agents considered when talking about AGI and alignment are usually assumed to do so, but it doesn’t seem necessary to the concept of agency itself.