Has Randall updated this yet for deciding between performing a creative task and trying to iteratively prompt-coax GPT to do it for you?

https://xkcd.com/1319/

The risks of AI aren't hypothetical future concerns about dystopian misalignment scenarios. Our round-the-clock discussion about today's powerful-but-early GPT technologies has already arrested our attention, diverting us from much of what we considered important just months ago, and is rapidly altering committed plans from product roadmaps to functional operations to overall missions and strategies. Many organizations will be circumspect, will move too slowly, and get lapped. Many will be paranoid or opportunistic, will lose focus, move too quickly, and make costly mistakes. Calibrating this will be difficult because it's too new to really wrap our heads around. It will take as much time as we choose to give it, but predictability will remain elusive. Many operations will fail through acts of omission or commission. This is going to be more disruptive than the Internet in the same way that the Internet was more disruptive than electricity: one stands on the shoulders of the other, and the successive rate of change and disruption is a function of the adoption of its predecessors.

It's going to be a wild ride, and the AI consultants, armies of them, are silently amassing at your gates...