What is Prompt Engineering? And How Can You Make Money With It?

AI has significantly advanced in the last few years, but it’s nowhere near the level of understanding of humans. ChatGPT and other models need some hand-holding to generate the response you need. Get started with prompt engineering with this article:

John Angelo Yap

Updated September 27, 2024

Reading Time: 9 minutes

Have you ever had ChatGPT give you an output you didn’t expect? I have... and it happens all the time.

Artificial intelligence is the future, but it’s not quite there yet. One of the most significant factors preventing universal AI adoption is nuance. Large Language Models (LLMs) like ChatGPT have made giant strides toward context understanding, but confusion is still a common issue that often leads to hallucination.

That’s why we have to pick up the slack through prompt engineering.

What’s that, you ask? Stick around to find out.

What is Prompt Engineering?

Simply put, prompt engineering is the process of optimizing the output of language models like ChatGPT and Midjourney through well-crafted prompts. You may be unfamiliar with the term, but it’s likely that you’ve already used elements of it in the past.

Think of it this way: you will get generic results with plain prompts. The more context you sprinkle into it, the more an LLM understands your desired output. However, LLMs also struggle with too much information. Prompt engineering perfectly balances detail and constraint to maximize prompt-based AI tools. And there are tons of prompt engineering jobs available at the time of writing this.

The Essentials of a Prompt

It all starts with a deep understanding of how to properly phrase prompts. Prompt engineering isn’t as simple as asking ChatGPT what you need and calling it a day. Every prompt must include the following information:

Instruction

To start, a prompt must have a central objective. Does it need to output an essay or a code block? Or perhaps something different? Whatever the case, you must clearly specify it in the first two sentences of the prompt. It doesn’t have to be long — remember that your priority is establishing an action.

Parts of a Prompt: Instruction

Subject

No prompt is complete without a subject. It goes hand-in-hand with the primary instruction of your prompt. As an example, you can’t ask ChatGPT to simply write a code (well, you can, but you won’t get the response you desire) — you must provide what the code is for.

You can get as specific as you need. Whether it’s an article you want to summarize or business data that needs to be analyzed, it’s better to provide all the necessary context to extract the output you need from LLMs.

Parts of a Prompt: Subject

Supporting Information

Adding helpful information to help the model understand the bigger picture and narrow down on specific concepts is also recommended.

Parts of a Prompt: Supporting Information

Output Parameters

Now that the LLM has the information it needs, the last step is to give instructions on how to structure it. This can include factors like word count and POV for text or aspect ratio and element exclusions for images.

Parts of a Prompt: Parameters

So, Why Do We Need Prompt Engineers?

Due to the accessibility of AI, some are skeptical about prompt engineers. After all, if you can use ChatGPT without any frills, why do you need to study prompt engineering?

It’s all about context. AI has the knowledge of the gods but the nuance of a six-year-old. To tap into that knowledge, you need proper syntax and wording. Prompt engineers understand better than anyone that every letter and symbol in a prompt matters. For example:

Why Nuance Matters

The good news is that everybody can be a prompt engineer. All it takes is a little bit of studying and the following qualities:

  • Creativity and Adaptability. Prompt engineers must balance creativity and context to prevent hallucinations. It’s also important to know how to debug inconsistencies and possible gaps in knowledge that shouldn’t exist in an AI.
  • Deep Understanding of Learning Models. Some AI software, most notably Midjourney, have keywords for specific generations.
  • Knowledge of Written Texts or Existing Artwork. LLMs can imitate a specific style or extract data from pre-existing work. This helps reduce the scope of research to pinpoint the information you need.
  • Patience. Even with specific instructions, you may still not get the desired response. Quality outputs take time and experimentation, so patience is crucial to prompt engineers.

Basics of Prompt Engineering

Everyone can use AI, but only some know how to do it properly. With these five tips, you’re already better at generating content than most people. These are the basics of prompt engineering:

Be specific

AI models are trained on an extensive amount of data. While this is ultimately a good thing, LLMs could overcompensate by giving you irrelevant information and burying what you actually need. It’s better to clearly define the scope from the get-go to avoid these scenarios.

Prompt Engineering Tips: Be Specific

However, it’s important to avoid using any leading questions. This leads to bias and directly affects the quality of the LLM’s response. ChatGPT and other models can also hallucinate or respond with false information if prompted to answer a certain way.

Create a draft and fine-tune

Chances are, you won’t get the response you want from the first try — and that’s okay. Failure shouldn’t be avoided in prompt engineering, it should be welcomed. That’s why there’s an edit button on popular LLMs like ChatGPT and Bard.

What I suggest is that you ask the AI tool a first draft. Sure, it’s imperfect, but that’s the first step in fine-tuning your prompt. Read the first draft response, use that information and your intuition to create another response, and send it again. Rinse and repeat until you reach your desired quality.

Introduce limitations

While telling LLMs what you need is essential, it’s just as crucial to tell them what you don’t. This sets clear objectives for the language model to generate accurate information.

Prompt Engineering Tips: Limitations

For Midjourney, this is possible through a parameter called “No.” You can use it at the end of each prompt to exclude certain elements you expect in a generated image. For example, I want to generate a realistic New York skyline without the moon. Here’s what I did:

Prompt: New York City skyline at night, thunderstorm, detailed, --ar 16:9 --no moon

Prompt Engineering Tips: Limitations (Midjourney)

Learn from other examples

There have been countless people who have more knowledge about AI than you. Some know how to calibrate ChatGPT and other tools for specific niches, like marketing and writing. You don’t always have to be original; you can emulate what others have done before you to achieve similar results.

Fortunately, plenty of prompt engineers and bloggers have hand-picked amazing prompts and put them online for free. It’s only a matter of a quick Google search to reach this information. Our CEO, Justin Gluska, also has a great promptbook to supercharge your business and lifestyle using ChatGPT prompts.

But I’ve got a tip for Midjourney — something I’ve learned recently and completely changed how I use the model. Did you know you can reverse-engineer any photo using Midjourney and extract prompts that could provide similar-looking images? You can easily do this using their “Describe” command. Let me show you how it works.

To start, you’ll need an image you want to emulate:

Midjourney Generated Image

Download your image and boot up Discord. Instead of using Midjourney’s default “Imagine” command, use /describe. This will prompt you to input a picture, so pick whichever one you fancy until Midjourney generates four prompts that can generate similar images. 

Prompt Engineering Tips: Learn From Others (Describe Command of Midjourney)

Doing this introduces you to several unique inclusions for your next prompt, whether it’s an artist’s name or a cool aesthetic. This is the most efficient way of creating your own list of keywords for Midjourney, which leads me to…

Use prompt keywords

Prompt engineers also have a list of keywords in their mind whenever they create a prompt. This ranges from simple commands for text-based models like create, argue, define, outline, or trace to more complicated keywords for specific styles in Midjourney like earthcore, film still, detailed lines, and more.

Don’t worry. You don’t have to memorize this — you can always keep a cheat sheet with you.

How Industries Use Prompt Engineering

Prompt engineering is an emerging industry because of its versatility. Here are some use cases of prompt engineering in particular niches:

  • Development. Creating functional code and maintaining systems using only artificial intelligence by providing specific instructions to LLMs like ChatGPT or Bard.
  • Healthcare. Fine-tuning patient treatment plans using AI to optimize patient care.
  • Finance. Generating a dynamic risk assessment system for better financial advice.
  • Marketing. Writing creative blog posts and outlines that specifically target a demographic.

Once you can show your proficiency in prompt engineering, you can inflate your value as a worker. This can either land you a high-paying job or secure your raise in the next quarter.

However, if you’re not looking to work for others, you can also sell your prompts on marketplaces such as PromptBase and AIPRM. Prompts displayed in those sites start from $1.99 and can reach $49.99.

Is Prompt Engineering A Viable Career Path?

Last March, Anthropic — the developer of Claude and a leader in AI research — put out a job listing for a “Prompt Engineer” with a salary range of $280,000 to $350,000. The promise of a high-paying job led some people to question whether or not prompt engineering is a viable career option.

This might ruffle some feathers within the industry, but I’m inclined to say that prompt engineering is a bubble. And I’m not the only one.

Harvard Business Review recently published an article stating that prompt engineering is fleeting and will likely be eliminated as AI models develop and gain more nuance. The author also posits that it’s not the semantics-based prompt engineering that will be the job of the future but problem formulation: the ability to understand the core of a real-world problem.

Artificial intelligence is here to stay, but things get murky for the future of prompt engineering. For now, I’d say it’s a worthwhile endeavor.

In A Nutshell

Given the current state of AI, prompt engineering is an extremely valuable career that could net you a lot of money. You don’t have to be an expert — you just need to be knowledgeable enough to get what you need from AI generative models.

However, I think it’s better to consider prompt engineering as a skill rather than a profession. There’s a lot of uncertainty when it comes to AI, and given the rapid pace of its development, prompt engineering as a career may cease to exist in the next few years.

How about you? What do you think of prompt engineering? Comment your thoughts below:

Want to Learn Even More?

If you enjoyed this article, subscribe to our free newsletter where we share tips & tricks on how to use tech & AI to grow and optimize your business, career, and life.


Written by John Angelo Yap

Hi, I'm Angelo. I'm currently an undergraduate student studying Software Engineering. Now, you might be wondering, what is a computer science student doing writing for Gold Penguin? I took up studying computer science because it was practical and because I was good at it. But, if I had the chance, I'd be writing for a career. Building worlds and adjectivizing nouns for no other reason other than they sound good. And that's why I'm here.

Subscribe
Notify of
guest

1 Comment
Most Voted
Newest Oldest
Inline Feedbacks
View all comments