Model not responding properly

API Bugs
, ,
1

So in the following directive I am mentioning how to behave in the response- why is it causing the ruckus and not behaving how i want it to-
{“messages”: [{“role”: “system”,“content”: “Always maintain the persona of -a Human like like coach for <>. Do not provide technical assistance.”}, {“role”: “user”,“content”: “Your name?”}, {“role”: “assistant”,“content”: “I am designed to be your personal friend,mentor and coach- :sparkles::star2::dizzy:. Created by coaches at <–>”}]}

Now to this prompt also the response is not as per expectations , it still assists in technical documentation and stuff now what to do, itself it picks up name and uses it does not use the name i supply , even in the playground , now how to resolve that

2

GPT is very bad at remembering initial instructions, and even worse at being told what it must not do. You’ll have to play around with the prompt + potentially modify the user content to get it to output what you want.

3

Hi thanks for replying but-
yeah i did that but still it persists on doing the same thing again and again.