“Metacrisis” huh. As a name for the thing I like “the Long Emergency” better. AI in its currently-existing forms is important only in that it’s yet another refinement that can have some use in enhancing the efficiency of the systems we’ve been building up for the past few hundreds of years, but we are at the point where the diminishing returns from such novelties will no longer be effective in staving off disaster for any appreciable length of time. Large language models are already passé.
I find ‘metacrisis’ more descriptive and satisfying for the reasons Daniel talked about in the video - that it’s not just the many crises we face, but the underlying systems that are creating the crises (ie, Moloch). Also, it doesn’t matter if AI is not effective at staving off disaster, as long as it creates value for the market it will be deployed with mind-boggling scale and resource use even as the world burns.