GPT-4-Vision and Undecyphered Scripts!

So I just got gpt-4v at ChatGPT and I attempted to get some responses from undecyphered scripts!

Here’s the image I used:

and after some prompt engeneering, here are the results with a disclaimer at the end:

Proto-Elamite: “Record of [commodity/item] in the quantity of [number].”

Cretan Hieroglyphs: “Invocation or reference to [deity/concept], blessings or offerings of [item/animal].”

Linear A: “Transaction or event relating to [location/item], during the time or reign of [individual/period].”

These are highly speculative translations, and they’re more about capturing the potential essence or nature of each script than providing an accurate reading. The actual content of the scripts would require expertise and potentially more contextual clues.

pretty cool stuff right? What do you guys think? What are some interesting results and interesting stuff you guys are doing with it? Please feel free to share any results, I’m finding all of this very fascinating! Can’t wait to get gpt-4-vision on the API and loop it for experimentation!

Try it on Cherokee or Inuktitut. That will tell you if it’s just making up stuff.

I cant find documentation for submitting the image as a part of open ai api
Can you share how you did it? Are you able to select model=“GPT-4V” or you still use GPT4

when computer vision is deployed to you (right now only in North America and UK, you will see an upload picture icon in the normal gpt-4 model. It is still in limited release.