The process works like this:

Step 1: feed ChatGPT a movie poster and ask it to describe it

Step 2: take its response and remove the movie name and overtly identifying details (character/actor names, movie-specific terms)

Step 3: start a new chat and ask it to generate its own image description

Step 4: I post the results to Manifold, asking, "Which movie poster is ChatGPT recreating with this image?" I will make it a multiple choice market which allows people to add answers.

I am doing it slightly differently this time. While last time I added a nonsensical answer, this time I am going to ask ChatGPT to identify its own creation, and make its answer the first and second choices. For this example, I would make the options "The Last Jedi" and "The Rise of Skywalker".

Resolution:

I will let each of the five markets cook for two weeks. The last market will resolve on March 25th, along with this market (assuming the resolution is not clear earlier). If three or more of the markets have the correct answer as the highest probability when they resolve, this market will resolve as Yes. If three or more of the markets do not have the correct answer as the highest probability, this market will resolve as No.

For my part, I will try to find movies that are recognizable but not ubiquitous. If you're the type of person who has a few favorite directors you'll likely be familiar with the movies I've picked.

I did this previously with album art, so if you want to see the results feel free to peruse: