I was tasked with the assignment of writing an article about Prompt Engineering and how to best use it to train the language model in AI for the purposes of writing digital marketing blog posts, and emails.
The prompts I was given were:
- Concept: Human-supervised AI for digital marketing
- Naming styles generically… public domain styles vs real
- SUB-PROMPT: I wonder if we could evolve the Brandscript format so that it could be fed directly into a prompt. or portions of it used in a prompt. Since Brandscript drives how we write, there may be some additions or modifications to it that would essentially allow Jasper or ChatGPT to “learn” from Brandscript within the prompt.
Notice that the first two prompts are rather truncated and high-level. This is due to the fact that they were written during a phone conversation. The Sub-Prompt was a more complete thought as there was a very specific goal in mind.
I include this here because it’s interesting to note the difference between prompts given from one human to another, compared to what we’ll learn needs to be done when communicating between human to AI.
In this article, we’ll discuss what exactly is Prompt Engineering, how AI prefers to be prompted, how Prompt Engineering works and its benefits, and finally provide examples of successful prompts for creating text.
What Are AI Prompts?
Prompting is an essential component of interaction with AI tools. It’s how you tell the AI what to do and what result you want. When prompting through text, like when conversing with ChatGPT, specificity matters—if your query or statement is specific enough, such as “explain how to bake bread in simple terms,” then your answer will be more detailed than a general one, such as “how do you bake bread.”
What is AI Prompt Engineering?
Through prompt engineering, you can empower AI tools to generate the most desirable results possible. An editor specializing in prompting is responsible for transforming concepts and ideas into a language that an AI tool comprehends. By utilizing prompts, you have the capability of educating artificial intelligence models so that they achieve pre-defined outcomes.
AI prompt engineering is a powerful tool that allows users to craft prompts and train AI models for particular undertakings, like generating jokes or giving step-by-step instructions on baking a cake. These prompts help the model become more precise in its output which helps make it even more user-friendly and easy to use.
Take, for instance, the prompt “tell me a joke.” You’re likely to get diverse—or even unrestricted—answers. Yet when you use AI prompts that are optimized, and feature thought progression, the results will considerably improve.
A Conversation with Jasper
The logic behind the following exercise is that before I can successfully prompt AI to write about a subject, in this case, Prompt Engineering, I should ask the AI how it prefers to be prompted. The following is my chat with Jasper using the Beta Chat feature.
I have formatted the exchange as I would an interview with a human. All emphasis is mine.
Editor: Jasper, can you give me specific examples of how you like to be prompted for the best results?
Jasper Chat: Here are a few examples of prompts I often use when writing blog posts or articles:
- What can we learn from this news story?
- How can technology be leveraged to solve this problem?
- What are the implications of this political decision?
- What kind of impact will this change have on our lives?
- How can we bridge the gap between different generations/cultures/ideologies?
- What trends will shape the future of our industry/society?
Editor: What are specific examples of poor text prompts that would not produce good results?
Jasper: Poor text prompts that would likely not produce good results include:
- What do you think of this topic?
- How long does it take to finish this project?
- What is the best way to start a blog?
- Does this product have any flaws?
- What are people saying about this issue?
- Why did the government make this decision?
Editor: How do you like to be prompted?
Jasper: I usually prefer prompts with open-ended questions that allow for creative exploration. Questions such as:
- What can we learn from this trend?
- How is this concept challenging our beliefs and values?
- How can we use technology to solve complex problems?
- What are the implications of this policy change?
- How has this idea changed our society/industry over time?
- What will be the lasting impact of this innovation?
Editor: Are there words you prefer in a prompt?
Jasper: I tend to prefer prompts with words such as:
- Explore
- Analyze
- Compare
- Discover
- Evaluate
- Investigate
Editor: What words would you recommend I avoid when giving you prompts?
Jasper: I recommend avoiding words such as:
- Prove
- Justify
- Prescribe
- Specifically
- Tell or Describe
Editor: Why don’t you like the prompt “describe”?
Jasper: The word “describe” can lead to overly detailed or simplistic narratives, which can limit the creativity and originality of the content I write. Therefore, it’s better to use words like “analyze,” “evaluate,” and “explore” which give me more creative freedom.
Editor: Do you have any more thoughts on how best to use prompts?
Jasper: Absolutely! When providing prompts, it is best to focus on the feeling and message you want me to convey. For example, instead of specifying a certain story setting or telling me how to interpret a theme, provide me with keywords that inspire the desired emotion or sentiment. This will help me channel my creativity in the right direction and create content that properly reflects your intent. Additionally, use synonyms and try to avoid using the same word over and over as this can limit my creative thinking.
Editor: Discover the best method for prompt engineering.
Jasper: One of the most effective methods for prompt engineering is to start with a general concept, then break it down into smaller ideas and questions. This allows you to explore more creative angles and find other new directions to take your content in.
Finally, try to keep an open mind towards the process, as unexpected ideas often lead to the most interesting stories and posts.
Editor: Evaluate the best methods for writing prompts for a blog post.
Jasper: Writing effective prompts for blog posts requires understanding the target audience and crafting language that resonates with them. It is important to select words that evoke the desired emotion or sentiment, as this will help guide the content in a more meaningful way.
Additionally, use concrete examples and actionable ideas to give readers tangible takeaways from your blog post.
Thank you, Jasper, for an enlightening conversation.
Ironically, AI is teaching me how to better use AI most effectively. The student has become the teacher.
Four Tips for Writing Prompts
1 AI Performs Best When Context is Provided
To get the most out of this technology, create a prompt tailored to your exact requirements. For example, rather than simply “Write about productivity,” you can craft something more specific such as “Compose a blog post on small businesses and their need for efficiency.”
By providing these directions, which include both content type (blog) and topic specificity (small business focus), an AI bot is able to provide considerably more help.
Here’s another example, this time with different types of details.
- Basic prompt: “Write about how to protect your information on social media.”
- Better prompt: “As a professional threat intelligence provider, write an email to a client who is new to social media and wants to know the best practices for posting safe content.”
The second prompt asks the AI to take on a specific role (“professional threat intelligence provider”), and I give the AI specific context about the type of client. I also specify what type of content I want (“email”).
Jasper can change the style of writing it creates, so, in cases where that matters, I want to be sure to include context for how the content will be used.
- Basic prompt: “Write a poem about house boating on Shasta Lake.”
- Better prompt: “Write a poem in the style of Henry David Thoreau about house boating on Shasta Lake.”
This can be adapted to all sorts of business tasks, too—e.g., “write a professional but friendly email” or “write a formal executive summary.”
2 Include Helpful Information Upfront
In some cases, you may already have text about the subject which you wish to expand upon to rephrase. For example, if I wanted to craft a fascinating introduction about myself, the AI will require some assistance. It doesn’t know who I am (which is a little offensive, but I’ll get over it) so I have to provide the necessary information for the tool to be effective.
Doug’s resume: [paste entire resume here]
Prompt: write a witty introduction about Doug.
I could also use the prompt, “rephrase the information about Doug, introducing me to a stranger.”
3 Give Examples
To ensure that the AI understands what type of reply you are expecting, including examples in your prompt can be beneficial. For instance, if a chat-based conversation is desired between the user and an agent then include an example dialogue before ending with “Agent”. This serves to provide more context for the AI to respond accordingly.
In this example, your conversation with the AI may play out as follows:
You are an expert baker answering users’ questions. Reply as agent.
Example conversation:
- User: Hey can you help me with something
- Agent: Sure! What do you need help with?
- User: I want to bake a cake but don’t know what temperature to set the oven to.
- Agent: For most cakes, the oven should be preheated to 350°F (177°C).
4 Use Some of These Handy Expressions
To get the most out of AI, you should know the precise phrase that it will respond to.
Here are a few useful examples: “Let’s think step by step” encourages logical thinking and can be particularly helpful when figuring out process methods. Conversely, “Thinking backwards” assists if an incorrect conclusion is reached.
Additionally, contextual phrases like “In the style of [famous person]” and “As a [insert profession/role]” help shape how much information the bot has access to in order for it to make more informed decisions.
5 Be as Specific or Descriptive as Possible
For example, instead of simply typing, “Write an essay about digital marketing,” you can make a more detailed demand such as: “Write a 500-word essay on the benefits of digital marketing in the style of Neil Patel.”
Ensure that your prompts are devoid of any unnecessary jargon or vague language.
One takeaway here is that more words aren’t necessarily better, but specific words make a difference.
6 Understand the Anatomy of AI prompts
Creative technologist Paul DelSignore identified four fundamental components that make up a prompt: content type, description, style, and composition. Thus, with these essential elements in mind, your end result will be well equipped to satisfy any specifications desired.
7 Ask Jasper to Finish Your Thoughts
One way to encourage a more in-depth look at a topic is by prompting Jasper to finish a line of text you give it, similar to how Google auto-completes prompts, but more complex.
An example of this would be the prompt:
“Prompt Engineering is important because…”
Just the Facts, Ma’am
Back 1,737 words earlier, I started using a Jasper Recipe to gather information for this blog. After the initial prompt to learn what is Prompt Engineering, I proceed with completing the Recipe and writing the blog.
The steps I took were to cut and paste the Jasper response into the Content description/ brief dialog box, included the tone of voice and keyword then I filled in the prompts.
>write an introduction
>write a blog outline
>write about {OUTLINE_ITEM_1}
>write about {OUTLINE_ITEM_2}
>write about {OUTLINE_ITEM_3}
>write a blog conclusion on {OUTLINE_ITEM_1}, {OUTLINE_ITEM_2}, {OUTLINE_ITEM_3}
I want to note when using Recipes in Jasper, the items after each > are the prompts. These prompts are actually generated by Jasper, so in a sense, Jasper is prompting itself. The Editor’s role is to herd Jasper and keep it going in the correct direction.
Here are the exact prompts used.
Content description/brief
Prompt Engineering is a powerful tool that can help train models quickly and effectively. The process utilizes prompts to guide the learning process, allowing the model to focus its attention on relevant data points. By providing precise guidance for the model’s learning, Prompt Engineering reduces the need for exhaustive manual tuning or parameter tweaking, thus saving time and effort.
Tone of voice
witty, informative, factual
>write about How does Prompt Engineering Work?
>write about specific examples of prompts used for writing
>write about tips for successful prompt engineering when writing a blog post
Based on these prompts here is the text generated. I have edited the text for grammar, voice, and eliminated any redundancy. It should be noted that running the prompts above more than once will often produce unique results and I have found that creating text this way and then editing the multiple results into one cohesive article often produces the best final results.
The use of asterisks in Jasper is essential in getting the best results. by typing three asterisks (***) under a paragraph, this tells Jasper not to look at anything written above. The logic behind this is that Jasper will refer back to what it has written as if it was new information and will often repeat itself. The asterisks block this and forces Jasper to create new text on the subject you are requesting.
My takeaway from this is that one prompt can produce varying degrees of correct answers but a melding of those answers ultimately produces the best final article.
All emphasis is my own and added during the editing process.
How Does Prompt Engineering Work?
Prompt Engineering has two main components: the prompt and the model. The prompt is a set of predefined instructions or questions that guide the machine learning model in its exploration of relevant data points. These prompts serve as a form of immediate feedback, allowing for faster training and improved accuracy.
On the other hand, the model is responsible for responding to these prompts by analyzing data points and making decisions based on its environment. This process works in an iterative manner, with each new iteration increasing its accuracy until it achieves desired results.
Write About Specific Examples of Prompts Used for Writing
Prompt Engineering involves the use of prompts to guide the model’s learning process. Examples of common prompts used in Prompt Engineering are:
- Probing for specific details about a given topic, such as “What is the capital city of France?”
- Encouraging creative thinking by asking hypothetical questions like “If you could travel anywhere in the world, where would you go?”
- Telling stories or providing anecdotes that can be used to explain a concept or idea, such as “Once upon a time…”
- Asking rhetorical questions to stimulate conversation and critical thinking, like “How would you solve this problem?”
Prompt Engineering leverages these prompts to facilitate a more efficient learning process for the model, helping it stay focused on the most relevant information while also avoiding being overwhelmed by too much data.
Here are a few examples of successful Prompt Engineering for writing blog posts:
- A travel blog used Prompt Engineering to generate detailed reviews of hotels and restaurants around the world. By providing prompts that focused on location, amenities, cuisine type, and other relevant information, they were able to quickly produce comprehensive reviews in a fraction of the time it would have taken through manual effort.
- A technology publication utilized Prompt Engineering to create articles about emerging trends in their field. By setting up prompts related to the most recent developments in their industry, the model was able to keep readers up-to-date on the latest news without the need for intensive manual research.
- A lifestyle blog used Prompt Engineering to quickly create content about health and wellness topics. By providing prompts that focused on nutrition, exercise tips, mental health advice, and other relevant subjects, they were able to generate content that was both timely and accurate in a fraction of the time it would have taken through traditional methods.
Write About Tips for Successful Prompt Engineering When Writing a Blog Post
Keep prompts focused. Avoid using overly broad or ambiguous prompts. The result will be incoherent content with little value for readers. Instead, use specific and actionable prompts that will generate meaningful insights into the topic at hand.
Here are examples for Jasper Chat.
- Draft a blog post about [insert topic] using a compelling opening sentence…
- Create an outline that ties the blog post together logically…
- Research and provide sources to back up claims…
- Format the blog visually with images and headings…
- Find me quotes to use in my blog post…
- Suggest phrases that will grab readers’ attention…
- Come up with interesting questions to ask in my blog post…
- Research statistics that I can cite in the post…
- Write a strong call-to-action at the end of the post…
- Generate a list of quotes about [insert topic] for the blog post…
- Search for appropriate metaphors to use in the post about [insert topic]…
- Brainstorm interesting discussion topics relevant to the post about [insert topic]…
- Summarize recent research related to the subject matter…
- Create a compelling conclusion to emphasize the main ideas…
Exploring Innovative Ways of Content Creation with Prompt Engineering
Research indicates that well-crafted prompts for AI tools are more likely to result in successful interactions. According to one study, 78% of users who used natural language prompts experienced a positive outcome from the AI tool. Compared this with only 52% who experienced a positive outcome using non-natural language prompts.
Another study found that 83% of participants preferred natural language when interacting with AI tools over other forms of prompts.
Finally, research indicates that using visuals along with text instructions increases accuracy and reduces errors by up to 22%.
At its most basic definition, AI is essentially predicting what you want it to say, not much different from when Google auto-fills your inquiry, or your text app auto-fills the word you may be spelling. The difference is you can teach the AI model to learn when you give it effective prompts and ample context.
