I'd be curious to compare the results of Dall-E 2's output vs a group of human artists each individually given the exact same text prompt (with no follow up clarification allowed) and asked to produce 3-5 drafts.