Prompt to ask ChatGPT NOT to do something?

0xse0

Junior Member
Joined
May 26, 2023
Messages
121
Reaction score
57
Hi guys,

I have to be honest, despite all of the trends about AI being the "future", I still see it as CRAP.

For me AI is just a trend, where site news are using to attract visitors and where OpenAI is making money. Well, I'm not debating, I think it could have great benefits for some industries but I still can't ask ChatGPT not to do something!!

When I ask ChatGPT to NOT include some words in the content, it still does.

For example:
Write a sales letter for web design but don't include words like "we" or "our".

Output:
WE utilize the latest technologies ... With OUR responsive designs, your website will...

Any help?
 
Emphasis
Something like: please refrain from using .....
Also at the end of the prompt, repeat it again, like: Once again, please refrain from using...

You can also take the prompt and give it to chatgpt to rewrite and make it more understandable, it worked for me many times, sometimes, it doesn't.
 
Actually chatgpt is revolutionary. There is no limit to what can be done with it. I use it as an assistant in my software works and it saves me a lot of time. Gpt 4 is always better in understanding but gpt 3.5 is not bad .
 
After a few prompts, you can tell GPT to “avoid using, bullet points, adding descriptions to etc. If ChatGPT still adds it, you can repeat the same thing again and it’ll understand you the next time.
 
Thanks for all your answers but it stills does an AWFUL job. Have you tried? It sucks!!
 
AI does not follow "Not to" commands very well. They follow "Do" commands better.
 
Hi guys,

I have to be honest, despite all of the trends about AI being the "future", I still see it as CRAP.

For me AI is just a trend, where site news are using to attract visitors and where OpenAI is making money. Well, I'm not debating, I think it could have great benefits for some industries but I still can't ask ChatGPT not to do something!!

When I ask ChatGPT to NOT include some words in the content, it still does.

For example:
Write a sales letter for web design but don't include words like "we" or "our".

Output:
WE utilize the latest technologies ... With OUR responsive designs, your website will...

Any help?
Yes.., I once had to generate a text of 2 thousand characters and it always made it to 700, I could not get the desired result, I think it's just a temporary bug that will soon fix
 
It's an inherent problem of the LLMs. It's in the official openAI guidance somewhere. They don't handle negative constructions as well. Create a few-shot prompt with multiple examples of your desired output in order to get the behavior you're looking for.

Re: AlphaStore that's not a bug that's the maxtokens parameter in chatgpt. You can adjust it in API calls.
 
Back
Top