Many people would like to use AF to solve alignment. AF may not require empirical grounding to be sound or valid but it must describe the contemporary examples of the problem in order to form a solution to said problem.
I agree that “describing contemporary examples of the problem” is the most satisfying scenario, but it’s not inconceivable that some more powerful yet structurally simpler agents could be understood well enough to help us avoid dangers posed by agents that are more complex in some sense but less powerful.
We might not be able to chemically characterize arbitrary molecular compounds, but we know enough chemistry to synthesize simpler materials that contain them.
Many people would like to use AF to solve alignment. AF may not require empirical grounding to be sound or valid but it must describe the contemporary examples of the problem in order to form a solution to said problem.
I agree that “describing contemporary examples of the problem” is the most satisfying scenario, but it’s not inconceivable that some more powerful yet structurally simpler agents could be understood well enough to help us avoid dangers posed by agents that are more complex in some sense but less powerful.
We might not be able to chemically characterize arbitrary molecular compounds, but we know enough chemistry to synthesize simpler materials that contain them.