Is it possible to give instant feedback loop through API?

Hi there !

I’m working on a side project requiring OpenAI to generate ideas based on a keywords.

It seems to work properly but sometimes i need to use fine tuning to improve some results.

I’m using internal feedback loop to let the user give instant feedback (is it relevant or not) to identify some fields where the result is not relevant.

My question is: is it possible to call OpenAI API in order to use the new feedback feature we can find on playground ?

Screenshot 2022-11-20 at 19.32.30

Btw, is it useful for OpenAI to do it?

Thank you in advance for your help,


1 Like

You can collect that data for your own uses - to maybe fine-tune the best examples later.

I don’t think OpenAI would open up that sort of rating to just anyone, though.

Hope this helps.

1 Like

Hello Lequentz,

I will try to give my opinion and help (always!).

As far as I understood:

1 - You are using GPT-3 via API, right?
2 - If yes, for me it was not clear how you are using it, but from what I could absorb it was that the users, in some way, make interactions that generate some kind of call / response to OpenAI, right?
3 - If yes, would you like to suggest to OpenAI and also use the “Useful” and “Not Useful” function that exists in the Playground only when you receive a response via front-end in the interface, right?

Regardless of that, I think we live in democratic environments and you can try to validate your idea, and of course, if it will have an impact on people’s lives and be useful for the planet (I say this because in the current moment we are living in and mainly with the amount of good and bad things happening, whether here or in other places “”, changing a functionality, not participating in the context of the company or organization, requires at least that this change the lives of people).

Imagine if you were the head of a Formula 1 company and someone came to suggest that Tires or some car item that other normal people (public) had access to purchase and that you would like to suggest new types or implementation of this Tires || item model.

I don’t know if the analogy was clear, but if you believe that this needs feedback and your idea is good, go ahead. I started 4 months ago with an idea of creating a new dataset for indigenous languages in Latin America and it is being a little difficult to keep forces alone, but I preferred to follow alone until I finish the idea and validate.

It is as @PaulBellow suggested.

I consulted GPT-3 and she sent the following links:

1 Like