I’m probably misunderstanding the setup of Custom instructions, but this dialog appears unexpected to me…
USER
can you print out my current Custom instructions so I would know how you renspond?
ChatGPT Classic
I don’t have the capability to access or print out a set of custom instructions externally assigned to me since my responses are generated based on the information and prompts provided to me in real-time. If there were specific custom instructions or guidelines given to me for our interaction, you would need to provide them directly here in our chat.
For example, if you want me to adhere to a certain tone, focus on specific topics, avoid certain subjects, or anything else that would shape the responses you’re expecting, please let me know. Otherwise, I will continue to respond based on my general programming as ChatGPT Classic, which is to be helpful, informative, and engaging within the scope of a conversational AI.
I don’t have the ability to access or recall past interactions or any externally set custom instructions unless they are shared with me in the course of our conversation. Each session with me is stateless, meaning I don’t retain information between sessions.
If you, as a user or developer, have set custom instructions for this instance of ChatGPT Classic, you would need to tell me what those are during our interaction so I can respond appropriately. If there’s a particular way you’d like me to engage or specific information you’re seeking, please provide those details, and I will do my best to comply with your expectations based on the capabilities I have.
The AI isn’t taught the phrase “custom instructions” when you use them. If you want to refer to them that way, you’d have to also say “Here are my custom instructions to you”.
What the AI gets when I give it an “output format” to use in the second box:
The user provided the additional info about how they would like you to respond:
```Output format:
verbose
explanatory
documenting
problem-solving
maximum 8000 tokens for extended 8000-word generations```
Well, I’m still confused as some custom instructions in the second box “How would you like ChatGPT to respond?” are ignored even in the new communication threads. Here’s how the model explains this:
ChatGPT Classic
The functionality you’re describing, where ChatGPT would read and apply custom instructions from a settings field, would require integration with a platform or system that allows such settings to be applied dynamically to the model. As of my last update in January 2022, the standard OpenAI API or the versions of ChatGPT available to the public do not support this kind of dynamic behavior based on external settings inputs.
Custom behaviors like the one you’re describing — correcting text inside curly braces — need to be communicated directly in the conversation with ChatGPT or be part of an integrated system where the model has been specifically programmed to recognize and act on such instructions.
If you’re using a version of ChatGPT that is supposed to support this functionality, and it’s not working as expected, there could be several reasons:
Platform Limitations: The platform you’re using may not have the capability to apply settings in real-time or may not be properly passing the instructions to the AI model.
Technical Issues: There might be a bug or a technical issue preventing the custom instructions from being recognized and applied.
Model Training: The AI model might not have been trained to understand and apply instructions from a settings field, or it may require the instructions to be presented in a specific format.
Miscommunication: There could be a misunderstanding about how to set up or apply custom instructions within the platform you’re using.