Limitations to Use Cases of the Edit Endpoint

Based on some of the terrible results I’m getting in the Edit endpoint playground, it’s not clear to me the kinds of instructions that are interpretable by the respective models. I requested “add descriptive comments to this code”, and it simply pasted a comment “-- add descriptive comments to this code” at the beginning of the script. Other times I got back nothing. I gave it a function written in R and told it to “refactor this code as javascript”, and it commented the original R code.

Given these results, what is this endpoint supposed to be doing? I feel like these are valid requests, but maybe I’m not interpreting the edit endpoint correctly?

Hey champ and welcome to the forum!

What you’re experiencing is the edit endpoint working as intended, it uses the models text-davinci-edit-001 and code-davinci-edit-001 to edit the text you give into other text. Try using chatGPT or the “chat” mode in the playground instead.

1 Like