Suggestion on a new option to help with evaluations and improve the model's approach to queries.

[JamesBethison]

I just had a short chat with GPT4 via ChatGPT+ and discovered a "flaw" in the approach used by the model to try and answer a semi-vague query of its knowledge base. I then discussed that flaw with the AI and as part of this latter discussion came up with a rather interesting idea for how to get more and better feedback from users. Imagine if the AI could automatically summarize the item of feedback and its potential value for review, but only for chats where the user first provided useful feedback which was acknowledged as such by the AI.

I am not quite sure how to correctly share this chat session but here is the link to it. If there is a better way to share it please advise.

Try reading from the bottom up, one prompt/reply pair at a time.
https://chat.openai.com/c/9b42eb3f-3bfe-4a2e-bef8-ff0efb51c9f6

Thank you.

[mmtmn]

perhaps screenshots would be better?