One user's definition of AI 'alignment': doing what the user wants

One user clarifies that by “alignment,” they mean “getting your AI to do what you the user want it to do,” not guardrails against misuse. The replying user says that differs from what they assume people usually mean by alignment. Under that definition, they expect significant transfer from closed to open AI. But they argue that no frontier lab wants that outcome, so people shouldn't assume those labs will develop all the necessary techniques.