assume we do get _alignable_ AGI in the next few years. the most plausible mechanism from today’s standpoint seems to be “faithful chain of thought” (h/t AI 2027) does this mean we must rely on a pseudoreligious elite pouring over millions of pages of text to ensure alignment?(stokes)