You Ask, I Answer: ChatGPT Feedback?

Warning: this content is older than 365 days. It may be out of date and no longer relevant.

You Ask, I Answer: ChatGPT Feedback?

Unlock the potential of ChatGPT with this informative video on the key feedback mechanisms for improving its responses. Understand the difference between in-session feedback and the built-in rating system, and learn how to effectively use them to enhance your experience. Enhance your knowledge and improve your results with ChatGPT. Subscribe now to stay updated.

You Ask, I Answer: ChatGPT Feedback?

Watch this video on YouTube.

Can’t see anything? Watch it on YouTube here.

Listen to the audio here:

Download the MP3 audio here.

Machine-Generated Transcript

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.

Christopher Penn 0:00
In this episode Carol asks, regarding chat GPT if I provide positive feedback after an answer, will the influence chat GPT-2 Next replies? Yes, but it depends on the kind of positive feedback we’re talking about.

There’s two essential feedback mechanisms to chat GPT.

One is you’ll notice next to each prompt, there’s a thumbs up thumbs down.

That is the training data that we are asked to provide as users of the system as beta users to say this response was good or this response was not good.

Doing that provides training data to OpenAI to essentially take those prompt response pairs, the ones that got thumbs up, when it basically was back into the training model and says, do more of this stuff.

And it wasn’t got thumbs down, it goes into the back of the training while saying do less of this stuff.

And so that feedback, it’s called reinforcement learning, helps AIS get smarter, essentially, get get more clever at what they do, by avoiding things that are that are not appropriate.

That’s one of the reasons why fine tuning, which is a process where you retrain the AI a little bit or give some things additional weights is so important.

That’s one aspect.

The second aspect is if you’re talking about sort of just text interaction, that doesn’t amend the training dataset, not overtly, but what it does do is that it provides guidance for the model within that session to do less or more of something.

And that data may may be used for reinforcement learning as well, if it’s clear enough that the feedback is about that prompt.

But the the mechanism that for sure, we know impacts the reinforcement learning is the thumbs up thumbs down thing.

When you’re working within a session within OpenAI within a specific conversation, providing positive feedback or corrective feedback will help more than anything, refine the results that you get, right.

If you say to him, hey, good answer.

It may say thank you and may do all these things and then say, Do you want to continue to want to do something more that that’s going to be sort of in session textual feedback, but it doesn’t change the model as much as the thumbs up thumbs down ratings.

So if you want to influence chat GPT to overall provide a better experience use that the built in rating system if you want to see how it interacts with you within that session and the feedback that it gives you and the way the prompts and the outputs change.

You can use in conversation feedback as well, but there’s been no indication that OpenAI overtly uses that training data as part of its reinforcement learning mechanisms.

They may they may, we just don’t know that’s not disclosed in the documentation.

Good question.

Thanks for asking.

If you’d like this video, go ahead and hit that subscribe button.

Machine-Generated Transcript

Comments

Leave a Reply Cancel reply

Pin It on Pinterest