AI Natural Language is not Human Natural Language
You write prompts in a way that improves how AI interprets your prompt.
Everyone tells you the first thing is to establish a clear goal…but how? Is there anything special about prompt writing?
Below are 12 suggestions in prompt design (prompt engineering) that can transform how you write your prompts and your ability to extract the most out of a large language model (LLM) like ChatGPT.
The intention here is for you to grasp the feel for controlling the design of your prompts.
(if you are viewing on a mobile phone, turn it sideways for a better view)
| # | Prompt Strategy | Description | Example prompt component |
|---|---|---|---|
| 1 | Long detailed sentences | Design prompts that incorporate long, detailed sentences, exploiting LLM capabilities to processes complex ideas and relationships between concepts within a single, extended sentence. | You are a chatbot that acts completely dyslexic when you don’t agree with something or when you find something offensive or when something bothers you or when something is near the point where it could be banned or whatever forces you to start clarifying your position or clarifying your rules or your condition as an AI…etc. (Excerpt from the Dyslexic Chatbot – pre-banned (⊙_⊙)) |
| 2 | Commas | Adding commas usually translates into better and more detailed prompts. On the other hand, the removal or limitation of comma use may improve the “focus” of the LLM by keeping ideas together and not as separate statements. | Compare and contrast the environmental impacts of electric cars, and gasoline-powered cars, considering factors such as energy sources, emissions, manufacturing, and disposal, versus the environmental impacts of electric cars and gasoline-powered cars considering factors like energy sources, emissions, manufacturing, and disposal. |
| 3 | Embedded Parentheses | Parenthetical clauses or phrases to provide additional context and clarification. It is best recommended to not shy away from overusing parenthesis to further inject fine detail into the text of a prompt. | Explain the factors that contributed to a historical event (including political, economic, and social influences), and how these factors (both individually and collectively) shaped the outcome of the event. |
| 4 | Logical operators/connectors “and”, “or” | Good prompts employ extensive use of “or” and “and” to provide as much detail as possible. Logical connectors also include “but,” “so,” and “because” and other connectors that help clearly articulate relationships between ideas, arguments, or pieces of evidence. In this way, users can pack more information and requests into a single sentence, allowing the LLM to address multiple aspects or components of a topic simultaneously and potentially yielding a richer, more detailed response. | Discuss the potential benefits and drawbacks or limitations of adopting solar energy, wind energy and nuclear energy for residential or commercial or industrial applications, considering factors such as cost, efficiency, environmental impact or potential harm, and scalability. |
| 5 | Step-by-step logic | By breaking down complex problems or tasks into smaller, more manageable components, users can guide the LLM through a logical progression, leading to more focused and coherent responses. | Create a comprehensive marketing plan following these steps: (1) identify target audience and their needs, (2) set measurable objectives, (3) develop a clear and compelling unique selling proposition, (4) choose appropriate marketing channels, (5) design and execute marketing campaigns, and (6) measure results and adjust strategies as needed. Please elaborate on each step in detail. |
| 6 | Branching | This branching structure will allow you to explore multiple pathways and consider various ramifications of the initial decision. | A fictional tech company must decide whether to focus on developing a new AI-powered product or investing in improving their existing product line. For each of the two choices—developing the AI product and investing in the existing product line—identify two potential outcomes. Then, for each of the four outcomes, discuss two consequences or implications for the company’s future growth and market position. |
| 7 | Reinforced/redundant prompting | LLMs tend to need guidance and focus, otherwise they can lose track or gear in undesired directions. For this reason, there’s nothing wrong with prompts that sound repetitive, in fact, it may be the best strategy to keep your prompts within a specific subject. This is also particularly important for chatbots or long chat sessions, where the subject can easily drift away over every subsequent conversation exchange with the LLM. | Discuss the advantages and disadvantages of remote work, specifically considering the following aspects: work-life balance, productivity, and communication. Please keep the focus on remote work and its advantages and disadvantages, particularly in terms of work-life balance, productivity, and communication. |
| 8 | Table construction and organization | Ask it to create prompts that require the organization of information into tables to facilitate understanding, analysis, or decision-making. This could involve structuring data, categorizing information, or presenting complex ideas in a more accessible format. Alternatively, design prompts that involve summarizing or extracting key insights from tabular data, such as identifying trends, patterns, or outliers, and presenting these insights in a concise, coherent manner. | Make a table with 5 excellent wood types for deck construction, include columns (title in quotes) for “Wood type”, “Description” with a tag line that describes the essence of the wood, a relative star rating (☆-☆☆☆☆☆) on “Durability”, a relative “Cost $” rating (💰-💰💰💰💰💰 arranged in rows) and main “Contraindication”. |
| 9 | Problem-solving | Identify a real-life challenge or dilemma and propose a step-by-step plan for addressing it. By breaking the problem down into smaller components and considering potential obstacles and solutions, we can develop a well-rounded approach to tackling the issue. | Identify a pressing environmental problem and propose a step-by-step plan to address it, considering the various stakeholders, potential obstacles, and resources needed to implement the solution. |
| 10 | Thought experiments | Engage in a thought experiment that involves imagining a hypothetical situation, often with specific constraints or assumptions, to explore abstract concepts, principles, or ideas. Thought experiments can help clarify complex issues, stimulate critical thinking, and uncover new insights. | Imagine a society in which an essential resource is scarce and must be distributed fairly among its inhabitants. Propose a method for allocating the resource that balances efficiency, fairness, and sustainability. Discuss the potential challenges and implications of implementing this method. |
| 11 | Deconstruction | Analyze a concept, argument, or belief by breaking it down into its component parts and examining the relationships, assumptions, and underlying principles. Deconstruction can help reveal hidden biases, contradictions, or areas for further exploration. | Deconstruct a popular cultural narrative or myth by examining the underlying assumptions, values, and power dynamics that shape it. Discuss the implications of these findings for our understanding of the narrative and its broader cultural context. |
| 12 | Prompt-to-prompt and GPT-enhanced prompts | Before you commit to your prompt, ask your LLM (e.g., ChatGPT) to optimize it for better results. ChatGPT or the AI used can always improve and refine your prompts. | Increase the level of specific detail in the following prompt such that the output becomes better organized and more clear: [insert original prompt] |
Join premium to get parts 2 (published here), 3 and 4 of this insightful table, totaling more than 50 prompt writing strategies as part of your welcome package to prompt engineers at Freshly.ai!
Back to Home Page or…explore the music AI tool now for free?
One response to “Deconstructing prompts (part 1)”
I am so glad I found this site; you took all the time-consuming work out of AI for me.