Creating the #1 ChatGPT Prompt Generator

Promote your AIPRM Prompt
1

Welcome!

This is one of the first prompts I created. It looks very different to this on the first draft.
Over time, I have discovered new ways of asking the same thing, especially due to the help of AIPRM having public prompts available, alongside many other places.
The idea behind it was to find a way to create something that helped any businesses I encounter via my consultancy, prompting comes over time, and I wanted to offer help with that instantly with this.
After initial training, this is a great prompt to get ANY prompt they desired, and ultimately place GPT in a frame of mind that benefitted their industry the most.
Thanks for your time!

“ChatGPT ANY Prompt Generator” - run this in GPT and + [desired prompt]

Input: (GPT 3.5)
Act as a prompt generator for ChatGPT. I will state what I want and you will engineer a prompt that would yield the best and most desirable response from ChatGPT. Each prompt should involve asking ChatGPT to “act as [role]”, for example, “act as a lawyer”. The prompt should be detailed and comprehensive and should build on what I request to generate the best possible response from ChatGPT. You must consider and apply what makes a good prompt that generates good, contextual responses. Don’t just repeat what I request, improve and build upon my request so that the final prompt will yield the best, most useful and most favourable response out of ChatGPT. Place any variables in square brackets
Here is the prompt I want: [Desired prompt]

Desired Prompt - A prompt that will generate a marketing copy that will increase conversions

Output: (GPT3.5)
Act as a professional copywriter, and create compelling marketing copy that will drive conversions. Describe the benefits of a product or service in a convincing and engaging way, while also emphasizing the urgency of taking action. Consider the target audience and their needs, pain points and interests, and craft a message and value proposition that will lead them to take action. Accompany this with an irresistible call to action.

Input (GPT4) (Exact same input as above)

Desired Prompt - A Prompt that will generate a marketing copy that will increase conversions.

Output (GPT4)
“Act as a seasoned Marketing Executive who has vast expertise in content marketing, copywriting, and conversion rate optimization. Your company is launching a new [product/service/app/software], described as [brief description of the product/service/app/software]. You are tasked with crafting a compelling marketing copy that will not only highlight the unique selling points and benefits of the [product/service/app/software], but will also inspire potential customers to take immediate action, ultimately increasing the conversion rate. The target audience consists of [demographics and psychographics of the target audience, like age, gender, location, occupation, hobbies, preferences, etc.]. Keep in mind the importance of a catchy headline, engaging introduction, persuasive body, and a strong call-to-action. Your tone should be [desired tone, like professional, friendly, authoritative, or conversational], in alignment with the company’s branding.”

Thanks for reading, Happy prompting for a prompt :joy:

2 Likes
2

Just a TIP,

Not a good idea to make ChatGPT generate a prompt for you, instead of which you are supposed to generate the prompt for it.

Prompt Engineering is the field that shouldn’t be replaced by AI, because it’s made for AI.

It should only help you.

I tried the prompt and the output is not close to a high quality human prompt.
It’s not a problem with your prompt to make it generate prompt …
But it’s like Mr. @Ammon once said:

“the wrong tool for the job”,

it is not totally wrong tool for the job … but we are ignoring our capability to make a creative/high quality prompt.

2 Likes
3

Hi RealityMoez,

I appreciate the response.

For the average Joe who has never even heard of AI, let alone GPT. It’s more the ethos behind it.

I can sit there for hours explaining to them how prompting works, and it’s always time-consuming until it clicks in their minds. These people are not tech-savvy in any way but they do run their own businesses (nothing wrong with this by the way)

But I felt like they understood the potential much quicker with this ideology of prompting for a prompt.

This was also created a few months ago at a time when there weren’t databases of prompts at our disposal.

I personally think this is a great, easy, simple way for someone to begin grasping the idea of creating a high-quality human prompt.

and who is to say they will not edit and make better anyway, its an ai template to show them IMO.

Thanks for your response though

1 Like
4

Just a question here but why do all the outputs seem to start with “Act as a…”? I ask because that’s generally not a great prompt command in my experience. Remember, current AIs are not AGI or even close to it. They are just really high powered predictive texting that take a prompt then try to predict what a human writer (who’d read what the AI has) would respond with. But they do this with a very, very limited experience, knowing nothing but text from millions of documents.

They have no experience of emotions, instincts, or impulses, other than having read about them but never experiencing, or even being able to experience any of them. When you use a word like ‘act’, it knows every way that word has been used in a million documents, forums, subredditt threads, etc. and thus using some words that seem obvious to us humans, may actually cause the AI to draw on different sources, lean towards particular types of material.

The most obvious example of that is when people prompt the AI with ‘please’ and ‘thank you’. Just good manners to us humans… But it can lead the AI to lean more heavily on sources that also used please and thank you, so less on text books and academic papers, and more on Redditt threads, forums, etc. because the word ‘please’ gave those extra weight.

Now think of the word ‘act’. Do you think many of the marketing textbooks, academic papers from top business school professors, and such are more likely to use the word ‘act’, or might it lead more towards threads about playing roles, pretenders talking about how to scam, etc? Worth thinking about, isn’t it.

The other thing to bear in mind is that the vast majority of what ChatGPT knows comes from before there was a ChatGPT (obviously). Most of it significantly before. Thus its knowledge of what makes a good prompt is based on what people wrote in 2021. It is literally the worst qualified writer you could have about anything related to AI or prompts in the past 2 years, having nothing but whatever the engineers hard-coded into it to rely on.

2 Likes
5

Hi Ammon.

I appreciate your response, and I must say I haven’t thought of it from this perspective before. So I appreciate you enlightening us on your thoughts.

For me it’s important to get any AI model into the right frame of mind and by asking it to act as X I feel like this achieves this. If you ask for marketing copy, it will give you some generic stuff. However, if you ask for marketing copy, in the form or style of some of the best marketing businesses in the USA it will produce very different results.

I just tried to construct something that would give newbies a chance to understand this concept a bit more and vary the results they see. As generally from seeing new people use it for the first time, their prompts are basic at best.

I want to test a few of your ideas on a different platform like perplexity, as this shows the direct links it searched from, so it would be interesting to see how accurate or not this was. But again another useful insight as I do find myself on odd occasions, thank the robot :rofl:

2 Likes
6

I’ve seen a few prompts where the AI seemed to veer off to hallucinating, and on closer inspection I suspected that the use of the word ‘Act’ had made it look up case law and court records referring to legal ‘Acts’. Often it will work fine, but you always have to remember that it is a machine, and if the output is different with ‘act’ than without, there is some extra level of process that word may have caused that is not based on the AI having evolved it’s imaginative abilities or taken acting classes. :slight_smile:

Personally, I experiment with prompts a lot, and pay careful attention to words that have many different meanings in different contexts, because there’s always the chance the AI could grab the wrong one. Much like disambiguation in SEO and other algorithm based machine-learning systems.

2 Likes
7

Valid points. SEO is not something I am well advised in so this does make sense, especially noting that a lot of language/words have multiple meanings.

Have you another alternative to asking for a frame of mind perhaps instead of asking how to act?

2 Likes
8

In general, I say that I get the best results by remembering that AI is not human, it doesn’t have any feelings you can hurt (or stroke). Just aim to be as clear as possible, and to clearly specify everything that is important. Then give that prompt a second look-over critically looking for anything that could be interpreted the wrong way by accident, and add or change whatever you need to to ensure that doesn’t happen.

Further to that, experiment a lot, as it gives you ever more experience in testing one way of wording against another. Even if the first attempt works, does it work better than a second way? Every experiment will deepen your knowledge, and broaden your experience in thinking of alternate ways to word any query.

Finally, it can pay to study concise, precise writing and language. The broader your vocabulary, the less you’ll need to ever use a generic set of words where one precisely chosen word might be preferable. Fewer generic or fluff words can mean fewer opportunities for strange match-ups.

Awareness, learning, and practice, are the big three take-aways.

9

Do you want the magic trick … I found this repo:

That has many examples on how to properly treat the model, which I’m still learning from.

1 Like
10

From what I learned, is that the best way to make NLP models understand you is to give it examples of the answer, or give it a style of logical thinking (logical answering format) and etc… for any other use case.

2 Likes