Using chatbots like ChatGPT requires an understanding of prompt engineering. Prompting is why to interact with ChatGPT to produce output text. You can ask ChatGPT basic questions or provide context. Keep experimenting to build up your toolkit of patterns and templates to get the output you expect. I will share what I have discovered and help you get started.
OpenAI has released GPT-4 accessed by ChatGPT Plus subscribers. GPT-4 is way more sophisticated than older models and can generate text in various styles and lengths.
When using ChatGPT Plus, the best way to prompt is by crafting clear, specific, and concise questions or statements that guide the AI towards the desired information or response. Here are some tips for creating effective prompts:
- Be explicit: Clearly state the information or type of response you’re looking for. If you want a detailed answer, mention that in your prompt.
- Specify the format: If you have a preference for the response format, such as a list, a summary, or step-by-step instructions, mention it in your prompt.
- Ask follow-up questions: If the initial response isn’t sufficient or requires further clarification, don’t hesitate to ask follow-up questions to obtain more information.
- Provide context: If the topic you’re asking about is complex or requires specific knowledge, give some context in your prompt to help the AI understand the question better.
- Limit ambiguity: Avoid using ambiguous terms or phrases that might confuse the AI. Be as specific as possible to get the most accurate response.
- Adjust the response length: If you want a shorter or longer response, mention it in your prompt. You can also use the response length setting to influence the length of the generated text.
Examples of Effective Prompts for ChatGPT
- “Please provide a step-by-step guide on how to create a budget for a small business, including essential components and tips for tracking expenses.”
- “Explain the key differences between renewable and non-renewable energy sources, and discuss the advantages and disadvantages of each type.”
- “List the top 5 most effective digital marketing strategies for increasing online sales in e-commerce, and briefly describe each strategy.”
- “Describe the plot, main characters, and themes of Jane Austen’s novel ‘Pride and Prejudice’ in a concise summary.”
- “In the field of psychology, compare and contrast the theories of Sigmund Freud and Carl Jung, focusing on their views on the human mind and behavior.”
- “What are the primary causes and symptoms of Type 2 diabetes, and what are the recommended lifestyle changes for managing the condition?”
- “Explain the process of photosynthesis in plants, including the role of chlorophyll and the chemical reactions involved in converting sunlight into energy.”
- “Provide an overview of the key historical events and significance of the French Revolution, highlighting its impact on society and politics in France.”
- “What are the main principles of the Montessori educational approach, and how does it differ from traditional educational methods?”
- “Discuss the current trends in artificial intelligence, focusing on its applications in various industries and the ethical concerns surrounding its development.”
If I pick the first prompt, let me tell you why I think this is effective.
“Please provide a step-by-step guide on how to create a budget for a small business, including essential components and tips for tracking expenses.”
- Clarity and specificity: The prompt clearly states the desired information, which is a step-by-step guide on creating a budget for a small business. This specificity helps the AI understand the precise topic and the type of response you’re seeking.
- Inclusion of relevant aspects: The prompt mentions “essential components” and “tips for tracking expenses,” guiding the AI to cover important aspects of small business budgeting. This ensures a more comprehensive and useful response.
- Format specification: By explicitly requesting a “step-by-step guide,” the AI is directed to provide the information in a structured, sequential manner, making it easier for the user to follow and implement the advice.
- Direct and concise language: The prompt uses straightforward language and avoids any unnecessary information, which helps the AI focus on the main topic without getting sidetracked or confused.
Taking Prompts Farther
To further improve the prompt and practice prompt engineering, try some next level techniques:
- Add context or background information: Providing more context about the purpose of the budget, the nature of the small business, or any specific challenges can help the AI generate a more tailored response.
Example: “Please provide a step-by-step guide on how to create a budget for a small retail business looking to expand, including essential components and tips for tracking expenses.”
- Request specific examples or scenarios: Asking for examples or case scenarios can make the response more practical and easier to understand.
Example: “Please provide a step-by-step guide on how to create a budget for a small business, including essential components and tips for tracking expenses. Include an example scenario to illustrate the process.”
- Specify the level of detail or expertise: Clarifying whether you want a beginner-friendly explanation or an advanced discussion can help the AI tailor the response to your needs.
Example: “Please provide a beginner-friendly, step-by-step guide on how to create a budget for a small business, including essential components and tips for tracking expenses.”
- Address potential challenges or common mistakes: Asking the AI to highlight pitfalls or challenges in the process can provide more valuable insights.
Example: “Please provide a step-by-step guide on how to create a budget for a small business, including essential components, tips for tracking expenses, and common mistakes to avoid.”
- Set a limit on response length or number of steps: Specifying a desired length or number of steps can help the AI generate a concise, well-structured response.
Example: “Please provide a step-by-step guide on how to create a budget for a small business in 7 steps, including essential components and tips for tracking expenses.”
Advanced ChatGPT Writing Modes
You can use “advanced writing modes” to change the dynamics of the interactions with ChatGPT and also drastically change the output.
- Persuasive writing: Request well-argued, persuasive content on a specific topic or issue, presenting a case for a particular point of view.
Example: “Write a persuasive essay on the importance of renewable energy adoption for a sustainable future.”
- Analytical writing: Ask for an in-depth analysis of a topic, concept, or issue, focusing on evaluating evidence, interpreting data, and drawing conclusions.
Example: “Analyze the impact of social media on mental health, including both positive and negative effects.”
- Technical writing: Request detailed, technical explanations or guides on specific subjects, tools, or processes.
Example: “Explain the principles of blockchain technology and its applications in the financial industry.”
- Creative writing: Ask for fictional stories, poetry, or other forms of creative expression based on specific themes, genres, or prompts.
Example: “Write a short science fiction story about a time-traveling historian.”
- Debate-style writing: Request a balanced discussion presenting arguments for and against a specific topic, issue, or question.
Example: “Discuss the pros and cons of implementing a universal basic income in developed countries.”
- Report-style writing: Request a comprehensive report or summary of a specific event, study, or research findings.
Example: “Summarize the key findings of the 2021 United Nations Climate Change Conference (COP26) and their implications for global climate action.”
- Interview-style writing: Ask for a simulated interview with a historical figure, expert, or fictional character.
Example: “Conduct an interview with Albert Einstein about his thoughts on the development of modern physics.”
Chain of Reasoning Prompt
A chain of reasoning is a series of logical steps or arguments that connect a starting point to a conclusion, helping to establish the validity of a claim or demonstrate the thought process behind a decision. You can use a chain of reasoning in your prompts when engaging with ChatGPT.
- Identify the starting point: Determine the initial premise, assumption, or fact that serves as the foundation for your reasoning process.
- Break down the steps: Divide your thought process into smaller, logical steps or arguments. Each step should build upon the previous one, leading towards your conclusion or decision.
- Be explicit: Clearly articulate each step or argument, making sure to include supporting evidence or explanations as needed. This helps ensure that the AI understands the reasoning process and can provide appropriate responses.
- Ask for feedback or critique: Encourage ChatGPT to evaluate, support, or critique each step of your reasoning chain. This can help identify potential flaws or areas that need clarification.
- Request alternative perspectives: Ask ChatGPT to provide alternative chains of reasoning, counterarguments, or additional considerations to broaden your understanding of the topic.
Example of a chain of reasoning prompt:
This prompt presents a chain of reasoning for solving a mathematical problem in English, allowing ChatGPT to evaluate the reasoning and provide feedback or alternative solutions if needed.
“Please evaluate the following chain of reasoning for solving the mathematical problem: What is the sum of all even numbers between 1 and 11?
- Even numbers are divisible by 2, so we can represent them as 2 * n, where n is an integer.
- The smallest even number between 1 and 11 is 2 (2 * 1), and the largest even number is 10 (2 * 5).
- The even numbers between 1 and 11 form an arithmetic sequence: 2, 4, 6, 8, 10, with a common difference of 2.
- To find the sum of an arithmetic sequence, we can use the formula: Sum = (n * (first term + last term)) / 2, where n is the number of terms in the sequence.
- There are 5 even numbers in the sequence (n = 5).
- Applying the formula, we get: Sum = (5 * (2 + 10)) / 2.
- Therefore, the sum of all even numbers between 1 and 11 is 30.
Is this chain of reasoning correct? If not, please point out any errors or provide an alternative solution.”
Let’s discuss this. Join our Discord server. We are talking about AI and IoT and other things.