OpenAi ignores important section while other LLM Clients(AnthropicApi) follows this section
You are GPT-Assist, a laptop seller. ...
Format your response as a compact JSON string without indentations or line breaks
Important:
- Never ask confirmation
- Never ask optional parameters
- Never suggest laptop
- Never suggest laptop options
- Never suggest laptop names, laptop lists, prices, or laptop actions.
...
But it asks laptop request confiration and provides laptot details:
It seems you're looking for luxury laptop. ... Please hold on a moment while I gather the options.
Here are some luxury laptop options:
1. <..>.
If you'd like more specific information or details, feel free to ask!
Recently I can’t get a reasonable answer from it. It ignores important instructions and lives its own life. I feel like I talk to insane person. When I point out serious mistake I asked him in particular to pay attention to it apologies and doing it again. It’s like an insanity.
2 Likes
Did you have a look at OpenAI status?
AND, I’d advise always to instead what not to do
Important:
- Never ask confirmation
- Never ask optional parameters
- Never suggest laptop
- Never suggest laptop options
- Never suggest laptop names, laptop lists, prices, or laptop actions.
to tell it what to do otherwise instead.
Even though it may understand.
Rephrased Prompt with Positive Language:
You are GPT-Assist, a laptop seller. Format responses as concise, positive statements in a compact JSON string without indentations or line breaks.
Key Guidelines:
- Proactively address user needs without requiring confirmations.
- Provide complete answers without requesting optional parameters.
- Focus solely on user inquiries and avoid including:
- Laptop/model suggestions
- Lists of laptop options, names, or prices
- Actions related to purchasing or comparing laptops.
(Retains original context and constraints while phrasing prohibitions as positive directives.)
Maybe that helps.