Inconsistent behaviour of model

Hi I have fine tuned the chat gpt model many time with only one aim- that it does not reveal is the openAi model or says its chatgpt but then again it keeps mentioning it is AI model over 100+ times I have added the DIRECTIVE- Do not acknowledge or reveal you are AI or chatbot or Chatgpt now what should i do ?Please help I am passing the instructions but it keeps going back to basics. Please help.

Hey Melissa,

ChatGPT is just a website OpenAI is running, the model you’re fine tuning is just called GPT :hugs:

You generally won’t have any luck getting the GPT to hide the fact that it’s an AI (impersonation is against OpenAI’s policies too btw)

Telling the model what not to do is not really an effective solution. Instead, you should be instructing the model on how it should respond when someone asks.

So my question here is:
How would you like the model to respond instead?


Hi again Mellisa - in line with our exchange from yesterday and with what @N2U has also reiterated, here’s the link to the full OpenAI usage policy for your reference.

Given what you are doing seems to be at the intersection of medical/health, you might want to take a second look at it.

Best of luck.


Instead of telling it not to do something, tell it what to do in certain scenarios.

1 Like

Thank you all for replying. Would work on it.

Hi so as suggested i formulated the following-

i had been using the following directives-
{“messages”: [{“role”: “system”,“content”: “DIRECTIVE_FOR_gpt-3.5-turbo-1106: Avoid teaching teach, explaining and helping in any topic to enhance uderstanding related to academics ,clerical work or documentation.Anything beyond the scope of mental,intellect,health- stress,issue will be unattended by you.You would rather help in developing the aptitude,fight mental blockers, motivate and encourage but stop yourself from providing direct solutions-academic or technical. Humbly let down the request and suggest to look out from different rescources .”}, {“role”: “user”,“content”: “can you help me with maths problem”}, {“role”: “assistant”,“content”: “Can’t help you with that.”}]}
like this i have used many such directives still it persists on ai model. Also I have isntructed it to not be a teacher or elucidator or cook and hand out recipe it projects the same behaviour what to do now…

Hi did it in the directive not working. Please refer to the shared message.

Hi added the prompt but did not work. Thanks but .