6 Comments
Oct 18Liked by Benjamin Breen

Oops! It looks like you already are doing this. At least you say so in the instructions but it's not clear from the actual example you give . Did you generate the story from images and artifacts of Pompeii?

Expand full comment
Oct 18Liked by Benjamin Breen

Because you can make tax from images as well as you can make images from text, have you tried using frescoes or mosaics, or even other visual artifacts from Pompeii, and then asking ChatGPT to reverse engineer the prompt? For some reason doing it this way provides a more detailed description of the image than when you simply ask it to caption it. This is just another way to continue the historical fiction telephone game.

Expand full comment

Thank you for these posts Benjamin. I'm an Irish historian and have been tasked with giving a talk on tech and the teaching of the history of the Irish revolution - I plan on citing your example here. I'm in the early stages of a project on video games and Irish history and your posts have also helped me think through that topic.

Expand full comment

If only you could cross-reference these prompts with some sort of retrieval-augmented generation from a database of historical facts about Pompeii and Roman life during the period to bring in relevant primary source material on a per-turn basis in the simulation to narrow in on the relevance. If this is something of interest to you, please don't hesitate to reach out! You can find me at https://twitter.com/the_key_unlocks on Twitter.

Expand full comment