This is a great point, one that I think is reinforced by the fact that these models (at least so far) arn't being trained and benchmarked on "good writing". Which in and of itself is going to be, to your point, a moving target.
And to be clear I don't doubt Fable couldn't pump out 13 thousand alternative fan fictions to Game of Thrones and they would follow some distribution curve that you'd see in meat-space, some garbage, some good, and a lot in the middle.
And to be clear I don't doubt Fable couldn't pump out 13 thousand alternative fan fictions to Game of Thrones and they would follow some distribution curve that you'd see in meat-space, some garbage, some good, and a lot in the middle.