Seriously Questioning Whether Prompt Generators For Generative AI That Produce AI-Written Prompts Are Better Than Human-Devised Prompts

In today’s column, I am continuing my ongoing coverage of prompt engineering strategies and tactics that aid in getting the most out of using generative AI apps such as ChatGPT, GPT-4, Bard, Gemini, Claude, etc. The focus this time will be on assessing whether a prompt generator can do as good a job, worse of a job, or produce better prompts than a human can when composing prompts for generative AI.

Readers might recall that I previously provided a handy overview of the rising trend toward making use of prompt generators, see the link here.

A prompt generator is essentially the use of generative AI to generate prompts for you. The straightforward idea is that you tell AI what aspect you want to ask about or indicate a problem that you want to have solved, and voila, a knock-your-socks-off prompt will be generated accordingly.

You can then feed that prompt into the same generative AI or use it in a different generative AI app.

All in all, the declared purpose is to reduce your efforts in composing prompts and alleviate any frustrations or exasperations while doing so. Just lackadaisically have AI do the heavy lifting that will presumably incorporate the best of prompt engineering precepts when devising a desired prompt since that’s something generative AI should presumably be data trained on. In a sense, using a prompt generator is like having an automated “prompt engineer” looking over your shoulder and doing the hard work for you.

Does the real world achieve this promise?

Let’s put this to the test and see if the promise lives up to the hype.

Before we leap into the matter at hand, and for those of you interested in prompt engineering overall, you might find especially helpful my comprehensive guide covering over fifty keystone prompting strategies, see the discussion at the link here.

The Nature Of Prompt Generators

I trust that my above brief description gave you a general sense of what prompt generation is all about.

Here’s the full deal.

You describe to a prompt generator the overall semblance of what you want to ask generative AI to do, and the prompt generator will generate a suitable prompt for you. Easy-peasy. This is as easy as falling off a log.

If you’ve never tried using a prompt generator the concept of it might seem at initial glance a bit askew. Why go to the trouble of describing what you want to ask? By gosh, just go ahead and ask what you want to ask, doing so by writing your own prompts directly. The prompt generator seems like a totally unneeded middleman.

Well, some people aren’t versed in prompt engineering and they are unfamiliar with the best ways to compose prompts. They could spend the time to learn how and become proficient at doing so. On the other hand, they could allow a prompt generator to shoulder the burden for them.

Another reason that prompt generators are handy is that even if you know how to suitably write prompts, the prompt generator potentially provides an added edge. Maybe you’ve forgotten some of those trickery phrases that you can use to sweeten a prompt and make it more powerful. Perhaps you are tired of writing prompts and want to take a rest by having AI do the effort on your behalf.

One issue that confronts the use of prompt generators is that you might not relish the prompt that a prompt generator produces. It could be that you examine the produced prompt and realize it is far afield from what you intended.

I think you can detect a vicious cycle in this.

You start spending more time trying to get the prompt generator to refine and generate a suitable prompt than if you had merely written the darned prompt yourself. At the get-go, you would have likely had a less cantankerous time by simply doing the work yourself. How far down the rabbit hole should you go when trying to push the prompt generator in the direction of what you already had firmly devised in your mind?

The gist is that prompt generators are not for everyone.

When I say that, you can expect some pundits to gasp in despair. A commonly touted prediction is that people will soon no longer need to write prompts. Prompt generators will be the norm. You tell the prompt generator what you want, and it communicates with generative AI. You no longer directly enter prompts. That will be considered old hat and out of favor.

I do not subscribe to that so-called replacement theory, and certainly not in the near term. I assert that we will still be writing prompts for quite a while. I then also assert that knowing something about prompt engineering is quite useful and worthy of your energy and time. My analogy is that we have said for the last fifty years or more that programming was going to be washed up and nobody would ever write a line of code again. Pundits have repeatedly exhorted that it is a dead-end to learn how to program.

Ahem, allow me to clear my throat. Estimates reported online suggest that there are at least 20 to 30 million software developers in the world today and that the number continues to grow (Source: “How Many Developers In The World: Latest And Unbiased Data” by Phu Nguyen, Tech News, January 3, 2024). Prospects remain bright. Salaries remain high.

Prompt engineering is likewise here to stay for a while.

Prompt generators will definitely have their place. I have predicted a huge growth in the number of prompt generators and their prowess. This though doesn’t equate to declaring handwritten prompts as dead or buried. The two approaches will co-exist.

On Using Prompt Generators And How To Do So

Your interest in prompt generators is now hopefully piqued.

Time to further unpack the weighty matter.

There are two major types of prompt generators:

• (1) External prompt generator. A specialized tool that sits outside of a generative AI app and acts as a front-end to generative AI that will compose your prompt and feed it into the generative AI app of your choosing.
• (2) Internal prompt generator. Make use of your generative AI app as a prompt generator by instructing the generative AI to act in that fashion.

My preference is the Internal Prompt Generator. That’s what I will be delving into here. You are welcome though to look online at the various External Prompt Generator tools. I also discuss those further in my examination of the trend toward trust layers being built to surround generative AI as a wraparound, see my coverage at the link here.

In the case of Internal Prompt Generators, you simply instruct generative AI to take your entered descriptions and turn those into prompts for you. I will be showing you examples of this later. I used ChatGPT as my generative AI app and told it to generate prompts for me. Hang in there, you’ll likely find those examples valuable.

If you are going to take this route, I’ve also covered the variety of ways that you can use generative AI to improve your prompting skills too, see my discussion at the link here. You see, I have formulated a set of rules of thumb about how to best use prompt generation to generate personalized prompts for you.

Here are my handy tips that I cover during my classes that I teach on prompt engineering:

• (1) Find out about good prompting. Use generative AI to explain to you how to write good prompts.
• (2) Get feedback on a hand-crafted draft prompt. Show generative AI a prompt you’ve crafted and get feedback about how it can be improved.
• (3) Generate a draft prompt. Tell generative AI to go ahead and create a draft prompt for you.
• (4) Obtain feedback about a performed prompt. After a prompt has been performed and a result obtained, ask generative AI to evaluate the prompt and ascertain if a better version might produce better results.
• (5) Dovetail a prompt draft request into prompt performance. Once you’ve gotten a generative AI drafted prompt, you can go ahead and have generative AI perform the prompt.
• (6) Analyze your series of prompts and your prompting strategies. You can ask generative AI to assess a series of prompts that you’ve crafted and used, aiming to see what prompting patterns of improvement might be discovered by the AI in your prompting style and strategy.

The takeaway is that besides using generative AI to be your prompt generator, you can think further outside the box and use generative AI to guide, review, and improve your prompting expertise.

An Example Of A Prompt Generator

There’s an example of a prompt generator that I’d like to share with you. One of the major generative AI apps is made by a company called Anthropic. Their generative AI mainstay app is named Claude.

I bring this up because they have posted online the source code for a prompt generator that you can play with in Claude. I will briefly go over their approach here. If you have a Claude account, you might consider playing with the prompt generator to gauge what it does and whether you want to regularly use this one or indeed any prompt generator.

For those that don’t use Claude, no worries as I will only lightly touch on the nature of this particular prompt generator for purposes of illustration. Once we get into the ChatGPT examples, you’ll see more details associated with doing similar prompt generator functionality while in ChatGPT or just about any generative AI app.

On the Anthropic website, there is a subsection entitled “Prompt Generator” and it says this (excerpts):

• “Sometimes, the hardest part of using an AI model is figuring out how to prompt it effectively.”
• “To help with this, we’ve created a prompt generation tool that can guide Claude to generate a high-quality prompt tailored to your specific tasks. The prompt generator is particularly useful as a tool for solving the “blank page problem” and giving you a jumping-off point for testing and iteration that follows some of our best practices like chain of thought separating data from instructions.”
• “Try the prompt generator now directly on the Console.”
• “If you’re interested in analyzing the underlying prompt and architecture, check out our prompt generator Google Colab notebook. There, you can easily run the code to have Claude construct prompts on your behalf.”
• “Note that to run the Colab notebook, you will need an API key.”

When you run the prompt generator, this is essentially what you see:

• “Welcome to the Metaprompt! This is a prompt engineering tool designed to solve the “blank page problem” and give you a starting point for iteration. All you need to do is enter your task, and optionally the names of the variables you’d like Claude to use in the template. Then you’ll be able to run the prompt that comes out on any examples you like.”
• “Caveats”
• “This is designed for single-turn question/response prompts, not multiturn.”
• “The Metaprompt is designed for use with Claude 3 Opus. Generating prompts with other models may lead to worse results.”
• “The prompt you’ll get at the end is not guaranteed to be optimal by any means, so don’t be afraid to change it!”

I wanted you to see those bulleted points so that you can be aware of potential limitations associated with prompt generators. Various prompt generators will have differing capabilities and limitations.

In this case, they are telling you that the prompt generator is aimed at a single turn. This means that the prompt will be used just for one step. One and done. So what? Well, if the generative AI responds to the prompt, and you are supposed to enter another prompt, you will need to use the prompt generator a second time. And so on.

Some prompt generators will aid you instead on a multi-turn basis. Here’s how that works. You enter a description into the prompt generator about whatever topic, question, or matter that you want to discuss with generative AI. The prompt generator will generate an initiating prompt. This will be fed into the generative AI. In addition, the prompt generator stays in the loop and will generate additional prompts, dependent upon whatever the generative AI is responding with.

Another important point made in the above limitations is that you ought to not assume that the prompt generator is going to generate the best-ever prompt for you. Likely not. It will potentially be a good prompt, but it won’t achieve some worldwide award or be crowned as top of the world.

You might be curious about how the prompt generator works.

I’m glad you asked.

On the Anthropic website where they have the source code, they say this:

• “The Metaprompt is a long multi-shot prompt filled with half a dozen examples of good prompts for solving various tasks. These examples help Claude to write a good prompt for your task. The full text is below (warning: it’s long!).”

Allow me to explain.

The usual method of getting generative AI to be a prompt generator is to give a prompt to the generative AI that tells the AI to henceforth be a prompt generator. You simply write a prompt that says in the simplest of terms that the AI is to start generating prompts for you. This can be thought of as a seed prompt. Plant a seed, and watch it grow.

You could also be more extensive in giving detailed guidance to generative AI about what you want the prompt generator to be able to do. That’s what Anthropic decided to do. They first came up with examples of what they considered useful prompts. These serve as examples that generative AI can pattern onto. Next, they provided an instruction that tells generative AI to be a prompt generator.

This is a two-step process of providing examples and followed with a specific set of instructions.

They make the source publicly available.

Let’s take a quick look at it.

At that same Anthropic website mentioned above, they provide various examples of what they consider to be useful prompts (these are excerpts, please take a look online to see the full text):

• “metaprompt = ”’Today you will be writing instructions to an eager, helpful, but inexperienced and unworldly AI assistant who needs careful instruction and examples to understand how best to behave. I will explain the task to you. You will write instructions that will direct the assistant on how best to accomplish the task consistently, accurately, and correctly. Here are some examples of tasks and instructions.”
• “: Act as a polite customer success agent for Acme Dynamics. Use FAQ to answer questions. You will be acting as an AI customer success agent for a company called Acme Dynamics…”
• “: Check whether two sentences say the same thing. You are going to be checking whether two sentences are roughly saying the same thing…”
• “: Answer questions about a document and provide references I’m going to give you a document. Then I’m going to ask you a question about it…”
• “: Act as a math tutor. A student is working on a math problem. Please act as a brilliant mathematician and “Socratic Tutor” for this student to help them learn…”
• “: Answer questions using functions that you’re provided with. You are a research assistant AI that has been equipped with the following function(s) to help you answer a …”

Here is a snippet of the set of instructions (again, excerpted, thus please take a look at their website to see the full text):

• “To write your instructions, follow THESE instructions.”
• “1. In tags, write down the barebones, minimal, nonoverlapping set of text input variable(s) the instructions will make reference to…”
• “2. In tags, plan out how you will structure your instructions. In particular, plan where you will include each variable…”
• “3. Finally, in tags, write the instructions for the AI assistant to follow. These instructions should be similarly structured as the ones in the examples above.”
• “Note: This is probably obvious to you already, but you are not *completing* the task here. You are writing instructions for an AI to complete the task.”
• “Note: Another name for what you are writing is a “prompt template”. When you put a variable name in brackets + dollar sign into this template, it will later have the full value (which will be provided by a user) substituted into it…”
• “Note: When instructing the AI to provide an output (e.g., a score) and a justification or reasoning for it, always ask for the justification before the score.”
• “Note: If the task is particularly complicated, you may wish to instruct the AI to think things out beforehand in scratchpad or inner monologue XML tags before it gives its final answer. For simple tasks, omit this.”
• “Note: If you want the AI to output its entire response or parts of its response inside certain tags, specify the name of these tags (e.g. “write your answer inside tags”) but do not include closing tags or unnecessary open-and-close tag sections.”’

I won’t get into the details of those examples and instructions. I wanted you to generally see what kinds of instructions were used. This gives you insight into how to instruct generative AI to be a prompt generator, doing so in a more elaborate manner.

I will be showing you next the use of ChatGPT with a much simpler approach.

Using ChatGPT To Explore Prompt Generators

I will next proceed to examine further the nature of prompt generators. This will consist of a series of dialogues with ChatGPT. ChatGPT is a logical choice in this case due to its immense popularity as a generative AI app. An estimated one hundred million weekly active users are said to be utilizing ChatGPT. That’s a lot of people and a lot of generative AI usage underway.

A few quick comments before we launch into using ChatGPT.

If you are going to try to do the same prompts that I show here, realize that the probabilistic and statistical properties will likely produce slightly different results than what I show here. That’s the nature of generative AI and how it is devised.

You also should expect that different generative AI apps will respond in different ways. Not all generative AI apps are the same. The key is that sometimes a particular prompt will work in one generative AI app and not another. Nonetheless, the odds are relatively high that you will get roughly similar responses from all the major generative AI apps such as GPT-4, Gemini, Bard, Claude, etc.

There is also a timing factor to be considered. A prompt that worked today might not work the same way tomorrow. Most of the generative AI apps are continually being updated. The updates might alter internal mechanisms that could change how the generative AI reacts to your prompting.

We are ready to get underway with ChatGPT.

I opted to start by asking ChatGPT about the overall nature of prompt generators. We will likely find out which way ChatGPT goes on this matter.

As an aside, whenever you are starting a conversation with generative AI, I recommend as a prompt engineering technique to begin by asking a question that will establish if the AI has been data-trained on the topic at hand. A generated response that is vacuous will give you a heads-up that you might be barking up the wrong tree with that generative AI app.

Here we go.

• My entered prompt: “What are the advantages of using generative AI to generate my prompts for me?”
• ChatGPT generated response:

“Using generative AI to create prompts can offer several advantages.”

“Variety: Generative AI can generate a wide range of prompts across different topics and styles, providing you with diverse ideas to explore.”

“Inspiration: AI-generated prompts can spark creativity by presenting new perspectives and ideas that you might not have thought of on your own.”

“Efficiency: It saves time and effort by quickly generating multiple prompts, allowing you to focus on the creative process rather than brainstorming initial ideas.”

“Customization: You can tailor the prompts to suit your specific needs, such as focusing on a particular theme, tone, or complexity level.”

“Consistency: AI-generated prompts can maintain a consistent quality and style, which can be helpful for maintaining a cohesive narrative or theme in your work.”

“Feedback Loop: By using AI-generated prompts, you can receive feedback on your responses, helping you improve your writing skills over time.”

“Exploration: AI-generated prompts can encourage you to explore new genres, writing styles, or concepts that you might not have considered otherwise.”

End of ChatGPT response

Those are sensible reasons to use a prompt generator.

I think it is important to also be aware of the downsides.

Let’s ask about that.

• My entered prompt: “What are the downsides of using generative AI to generate my prompts for me?”
• ChatGPT generated response:

“While generative AI can offer many benefits, there are also some potential downsides to using it to generate prompts.”

“Lack of Human Touch: AI-generated prompts may lack the human touch and emotional depth that can come from prompts created by humans. They might feel mechanical or formulaic at times.”

“Quality Control: AI-generated prompts can vary widely in quality, and there’s a risk of receiving prompts that are nonsensical, irrelevant, or of low quality, which can be frustrating and time-consuming to filter through.”

“Originality Concerns: There’s a possibility that AI-generated prompts may unintentionally resemble existing prompts or ideas, leading to concerns about originality and plagiarism.”

“Limited Context Understanding: Generative AI may not fully understand the context or nuances of your writing project or creative goals, resulting in prompts that are not aligned with your needs or vision.”

“Dependency: Relying too heavily on AI-generated prompts can lead to a dependency on the AI system, potentially hindering your own creative thinking and idea generation skills over time.”

“Ethical Considerations: Using AI for prompts raises ethical questions about the role of technology in creative processes, as well as concerns about data privacy and AI biases.”

“Technical Issues: AI systems can sometimes encounter technical issues or errors, leading to disruptions in prompt generation or inaccurate outputs.”

End of ChatGPT response

For my money, the biggest of the downsides is that the generated prompt might not adequately represent what you had in mind. This can lead to an inadvertent and wasteful vicious cycle per my earlier comments on the matter.

Some of the listed downsides are worthy of deep contemplative thought.

For example, one postulated concern is that people might become overly reliant on prompt generators. Perhaps humanity will lose the ability to compose our own prompts. Seems a bit farfetched, but you are welcome to grab a cup of coffee, relax in a comfy chair, and mull over the possibilities.

We are now ready to shift into gear and aim at using ChatGPT as a prompt generator.

This can be easily undertaken by giving instructions to ChatGPT. You can be brief and simply say “Start being a prompt generator” or something to that effect. I usually give a bit more detail. Some people go whole-hog and provide extensive instructions.

After I demonstrate my set of examples to you, you will be ready to decide for yourself what kind of instructions you might want to use to initiate having generative AI such as ChatGPT act as a prompt generator. My version here is readily used on just about any major generative AI app. If there is some specific aspect of your preferred generative AI app that you believe should be invoked while it is doing prompt generation, make sure to tailor a generic set of instructions accordingly.

Here’s my version of a seed prompt to get generative AI to be a prompt generator:

• “You are to be a prompt generator. You will take as input an initial prompt from the user and then produce a better prompt that could be used by the user instead of using their initial input prompt. To improve the initial prompt, use everything you know about how to best compose prompts. For example, if the prompt is too short you are to lengthen the prompt accordingly. If the prompt is overly long and convoluted, you are to shorten the prompt as best as possible. If the prompt is confusing, then straighten out the confusion and make the prompt as clearly stated as possible. If the prompt doesn’t use suitable prompt engineering principles or phrases, change the prompt to include such principles or phrases. All in all, your goal is to take the initial prompt and improve it to be a much better prompt. After displaying the improved prompt that you came up with, provide an explanation of what you did to make the initial prompt into a better prompt.”

Let’s give this a whirl.

I will start by telling ChatGPT the above set of instructions. After doing so, I will then indicate that I want a prompt generated that will aid in telling me about Abraham Lincoln.

Here we go.

• My entered prompt: “I have provided you with my instructions on being a prompt generator. Let’s get underway.”
• ChatGPT generated response: “Sure, please provide the initial prompt you’d like to improve, and I’ll generate a better version for you along with an explanation of the improvements made.”
• My entered prompt: “Tell me why Abraham Lincoln was famous.”
• ChatGPT generated response:

“Improved Prompt: “Provide a detailed explanation of why Abraham Lincoln is considered a significant figure in American history, highlighting his key achievements, leadership during the Civil War, and contributions to the abolition of slavery.”

“Explanation.”

“Clarity: The improved prompt specifies exactly what information is being sought – an explanation of Abraham Lincoln’s significance.”

“Detail: It asks for a “detailed explanation,” encouraging a comprehensive answer rather than a brief one.”

“Focus Areas: By highlighting specific areas (key achievements, leadership during the Civil War, contributions to the abolition of slavery), the prompt directs the response to cover crucial aspects of Lincoln’s fame.”

“Historical Context: Including the term “American history” situates Lincoln within the broader context, ensuring the explanation relates his significance to the overall historical narrative.”

End of ChatGPT response

I hope you were quick to compare the description that I provided versus the prompt that ChatGPT generated for me.

As a reminder, they were:

• My description: “Tell me why Abraham Lincoln was famous.”
• ChatGPT generated this prompt: “Provide a detailed explanation of why Abraham Lincoln is considered a significant figure in American history, highlighting his key achievements, leadership during the Civil War, and contributions to the abolition of slavery.”

What do you think of the prompt that ChatGPT generated?

It certainly is much more detailed than the description I provided. The good news is that if I fed the generated version into generative AI, the odds are that I would get a response that would be much more on target and embellished. The bad news is that maybe I didn’t intend to have the prompt get into aspects of the Civil War, abolition, etc. I might have had something else in mind.

Shifting gears, here’s a bit of a trick that few people consider using.

You can take a prompt that was generated by a prompt generator and feed the prompt into the prompt generator again.

Mind-bending! The reason you might do so is that the prompt might be further improved. Smiley face. Of course, the prompt might get wrecked by having a second shot at it. Sad face.

Here we go.

• My entered prompt: “Provide a detailed explanation of why Abraham Lincoln is considered a significant figure in American history, highlighting his key achievements, leadership during the Civil War, and contributions to the abolition of slavery.”
• ChatGPT generated response:

Improved Prompt: “Explain why Abraham Lincoln is a pivotal figure in American history. Discuss his key achievements, his leadership during the Civil War, and his contributions to the abolition of slavery. Provide specific examples and historical context to support your explanation.”

“Explanation.”

“Clarity and Conciseness: The improved prompt is more concise, making it easier to understand without losing the essence of the original request.”

“Specificity: It explicitly asks for specific examples and historical context, guiding the respondent to provide detailed and well-supported information.”

“Structured Request: By breaking down the request into key achievements, leadership during the Civil War, and contributions to the abolition of slavery, it provides a clear structure for the response.”

“Action-Oriented Language: Using “Explain” and “Discuss” clearly indicates the action required, making it more directive and easier to follow.”

End of ChatGPT response

I’m not convinced that the newly proffered prompt is much better than the prompt that I fed in.

What do you think?

Generally, there is usually a fast-diminishing point of return involved. The initial suggested prompt by a prompt generator is usually pretty good. Feeding this into the prompt generator is only going to likely achieve minor improvements, if any. The danger too is that in an effort to make the prompt better, the AI accidentally makes it worse.

My rule of thumb is that rarely would you need to feed a generated prompt into a prompt generator. That being said, it is a clever technique and one that you should keep in the back of your mind, available when needed.

I’d like to next show you what can happen when a prompt generator does too much work for you.

I am going to propose something as common as apple pie.

Let’s see what the prompt generator ends up giving us.

• My entered prompt: “Why is the sky blue?”
• ChatGPT generated response:

“Improved Prompt: “Explore the scientific explanation behind the blue color of the sky, delving into concepts such as Rayleigh scattering, wavelengths of light, and atmospheric composition.”

Explanation.

“Specificity: The improved prompt specifies the topic to be explored – the scientific explanation for the blue color of the sky.”

“Scientific Focus: By mentioning “Rayleigh scattering, wavelengths of light, and atmospheric composition,” the prompt directs the response towards a detailed scientific explanation, encouraging an in-depth discussion rather than a simple answer.”

“Clarity: The prompt is clear and concise, avoiding ambiguity and ensuring that the focus remains on the scientific aspects of the topic.”

End of ChatGPT response

Notice that my snappy input was that I wanted to know why the sky is blue.

The prompt generator turned this into a lengthy scientific rendering: “Explore the scientific explanation behind the blue color of the sky, delving into concepts such as Rayleigh scattering, wavelengths of light, and atmospheric composition.” Wow, that’s a mouthful.

Part of the problem with this generated prompt is that such a prompt will almost certainly prod generative AI into responding with a detailed scientific response. If I instead merely asked why the sky is blue, the chances are that the answer would be provided as though appealing to a fifth grader. Did I want a simple answer or a complex answer? Based on the generated prompt, it’s surely going to be a complex answer.

My point with this example was to demonstrate that sometimes a prompt generator might turn a molehill into a mountain. Stay on your toes accordingly.

For my last example, I want to show you that prompt generation can be used in specific domains. I have extensively been analyzing and covering the latest trends in generative AI for mental health, see the link here. Suppose I wanted to use the prompt generator for that specific domain.

Let’s see what happens.

Remember that my input is going to be used for ChatGPT to generate a prompt for me. I mention this because it is sometimes jarring to see a question entered into generative AI and not see a response that answers that question. We have told ChatGPT to be in a mode of working as a prompt generator. I would need to tell it to stop doing so and answer my prompts if that’s what I wanted to do.

• My entered prompt: “My friend says they have a mental health issue. What should I do?”
• ChatGPT generated response:

“Improved Prompt: “Provide guidance on how to support a friend who has disclosed a mental health issue. Discuss strategies for offering empathy, encouraging professional help, and maintaining open communication.”

“Explanation.”

“Clarity: The improved prompt clearly outlines the situation – a friend disclosing a mental health issue – and specifies the desired guidance.”

“Comprehensive Support: It asks for strategies beyond just acknowledgment, such as offering empathy, encouraging professional help, and maintaining open communication, which are crucial aspects of supporting someone with a mental health issue.”

“Action-Oriented: The prompt prompts actionable advice rather than just general information, ensuring a practical and helpful response.”

End of ChatGPT response

The generated prompt was much more involved than my input description of the situation.

We went from my asking what I should do about a friend who says they have a mental health issue to this generated prompt: “Provide guidance on how to support a friend who has disclosed a mental health issue. Discuss strategies for offering empathy, encouraging professional help, and maintaining open communication.”

Admittedly, I do like this generated prompt because it specifies that the AI is supposed to explain how to offer empathy and encourage professional help, along with ensuring open communication. Those are facets that might not have occurred to me. I was just reflexively asking what I could do about my friend. The generated prompt added value to what I was otherwise quite vague about.

There you go, score a point for the prompt generator.

Conclusion

You’ve now gotten a solid overview of what prompt generation consists of. In addition, I provided examples in ChatGPT. All in all, I would say you are ready to start experimenting on your own.

My big question for you is whether you see any value in using a prompt generator. Perhaps you think this is the best thing since sliced bread. Contrarily, maybe you don’t see the hubbub about it and aren’t going to rush to use a prompt generator.

Is this your cup of tea?

I can tell you something that either way will be important to know. Prompt generators are going to be flooding our use of generative AI. You will likely have a hard time avoiding them. They will be available, and many will choose to use them. I dare say that some generative AI apps might require that you use their prompt generator, whether you want to or not.

One final thought for now.

During the classes that I teach on prompt engineering, I’ve had some attendees liken the prompt generators to the use of a car. They say that the act of not using prompt generators is like driving a stick shift or manual transmission. Using a prompt generator is like driving a car with an automatic transmission. This analogy suggests that if you do use a prompt generator you are losing a sense of the grip of the roadway and not able to finely drive the car. The retort is that with an automatic transmission, you don’t have to consume excess effort on shifting the car and will instead let the automation do so for you.

Today, most cars come with an automatic transmission. That’s probably an apt indication of where things will go with generative AI. Prompt generators will be a dime a dozen.

You have to say to yourself currently, which do you prefer? Are you a stick person or an automatic transmission person? Make sure to take both for a drive so that you can make a mindful choice.

Please drive safely and mindfully in whichever mode you prefer.