My idea is to buy to a unpublished novel or screenplay with a detailed, internally consistent world built in to it and a cast of characters that have well crafted motivations and then ask it to continue writing from an arbitrary post-mid-point by creating a new plot line that combines two characters that haven't yet met in the story. If it understands the context it should be able to write a new part of the story and will be able to use a reader's intuitive sense of the character's motivations to move through their arc.
This whole thing would have to be kept under lock-and-key in order to be useful, so it would only serve as a kind of personal benchmark. Or it could possibly be a prestige award that is valued for its conclusions and not for its ability to use the methodology to create improvements in the field.
I have been thinking about this for use in evaluating locally run models, so I didn't make that connection in this case. I guess it would have limited utility.
This whole thing would have to be kept under lock-and-key in order to be useful, so it would only serve as a kind of personal benchmark. Or it could possibly be a prestige award that is valued for its conclusions and not for its ability to use the methodology to create improvements in the field.