While I claim that all intelligence that is capable to understand “I don’t know what I don’t know” can only seek power (alignment is impossible).
the ability of an AGI to have arbitrary utility functions is orthogonal (pun intended) to what behaviors are likely to result from those utility functions.
As I understand you say that there are Goals on one axis and Behaviors on other axis. I don’t think Orthogonality Thesis is about that.
Your last figure should have behaviours on the horizontal axis, as this is what you are implying—you are effectively saying, any intelligence capable of understanding “I don’t know what I don’t know” will on.y have power seeking behaviours, regardless of what its ultimate goals are. With that correction, your third figure is not incompatible with the first.
Orthogonality Thesis
It basically says that intelligence and goals are independent
Images from A caveat to the Orthogonality Thesis.
While I claim that all intelligence that is capable to understand “I don’t know what I don’t know” can only seek power (alignment is impossible).
As I understand you say that there are Goals on one axis and Behaviors on other axis. I don’t think Orthogonality Thesis is about that.
Your last figure should have behaviours on the horizontal axis, as this is what you are implying—you are effectively saying, any intelligence capable of understanding “I don’t know what I don’t know” will on.y have power seeking behaviours, regardless of what its ultimate goals are. With that correction, your third figure is not incompatible with the first.
I agree. But I want to highlight that goal is irrelevant for the behavior. Even if the goal is “don’t seek the power” AGI still would seek the power.