2. No, it seems too direct and not creative. I wonder if the original non-instruct models (curie etc) would be better for this use case than ChatGPT. Because they make more divergent and creative results.
2. No, it seems too direct and not creative. I wonder if the original non-instruct models (curie etc) would be better for this use case than ChatGPT. Because they make more divergent and creative results.
2. Thanks, that's good feedback to hear. I hadn't thought about trying the different models. Currently it's using gpt-3.5-turbo with the temperature set to 1.3. While testing, when I bumped the temperature up higher I started getting a lot more non-sensical words / letter combinations. I do have a lot of extra guidance added to the user's prompt before it gets submitted to the API (including to be creative), but it sounds like it's still not quite there. I'll do some more experimenting with prompt pre-processing, and try out some of the other models too!